CHAPTER 11 Systems of Nonlinear Differential Equations Contents 11.1 Autonomous Systems 11.2 Stability of Linear Systems 11.3 Linearization and Local Stability 11.4 Autonomous Systems as Mathematical Models 11.5 Periodic Solutions, Limit Cycles, and Global Stability Ch11_2 11.1 Autonomous Systems Introduction A system of first-order differential equations is called autonomous, when it can be written as dx1 g1 ( x1 , x2 , , xn ) dt dx2 g 2 ( x1 , x2 , , xn ) dt (1) dxn g n ( x1 , x2 , , xn ) dt Ch11_3 Example 1 dx1 x1 3 x2 t 2 dt dx2 x1 sin( x2t ) dt The above is not autonomous, since the presence of t on the right-hand side. Ch11_4 Example 2 Consider d 2 g sin 0 2 l dt If we let x = , y = , then x y g y sin x l is a system of first-order. Ch11_5 Vector Field Interpretation A plane autonomous system can be written as dx P ( x, y ) dt dy Q ( x, y ) dt The vector V(x, y) = (P(x, y), Q(x, y)) defines a vector field of the plane. Ch11_6 Example 3 A vector field for the steady-state flow of a fluid around a cylinder of radius 1 is given by x2 y 2 2 xy V ( x, y ) V0 1 2 , 2 2 2 2 2 (x y ) (x y ) where V0 is the speed of the fluid far from the cylinder. Ch11_7 Example 3 (2) If a small cork is released at (−3, 1), the path X(t) = (x(t), y(t)) satisfies dx x2 y 2 V0 1 2 2 2 dt (x y ) 2 xy dy V0 2 2 2 dt (x y ) subject to X(0) = (−3, 1). See Fig 11.1. Ch11_8 Fig 11.1 Ch11_9 Types of Solutions (i) A constant solution x(t) = x0, y(t) = y0 (or X(t) = X0 for all t). The solution is called a critical or stationary point, and the constant solution is called an equilibrium solution. Notice that X(t) = 0 means P ( x, y ) 0 Q ( x, y ) 0 Ch11_10 (ii) A solution defines an arc – a plane curve that does not cross itself. See Fig 11.2(a). Referring to Fig11.2(b), it can not be a solution, since there would be two solutions starting from point P. Fig 11.2 Ch11_11 (iii) A periodic solution – is called a cycle. If p is the period, then X(t + p) = X(t). See Fig 11.3. Ch11_12 Example 4 Find all critical points of the following: 2 2 6 (c) x 0.01x (100 x y ) x x y (a) x x y (b) y x 2 y y x y y 0.05 y (60 y 0.2 x) Solution (a) x y 0 x y 0 then y = x. There are infinitely many critical points. Ch11_13 Example 4 (2) (b) x2 y 2 6 0 x y0 2 Since x2 = y, then y2 + y – 6 = (y + 3)(y – 2) = 0. If y = – 3, then x2 = – 3, there are no real solutions. If y = 2, then x 2 . The critical points are ( 2, 2) and ( 2, 2) . Ch11_14 Example 4 (3) (c) From 0.01x(100 – x – y) = 0, we have x = 0 or x + y = 100. If x = 0, then 0.05y(60 – y – 0.2x) = 0 becomes y(60 – y) = 0. Thus y = 0 or y = 60, and (0, 0) and (0, 60) are critical points. If x + y = 100, then 0 = y(60 – y – 0.2(100 – y)) = y(40 – 0.8y). We have y = 0 or y = 50. Thus (100, 0) and (50, 50) are critical points. Ch11_15 Example 5 Determine whether the following system possesses a periodic solution. In each case, sketch the graph pf the solution satisfying X(0) = (2, 0). (a) x 2 x 8 y (b) x x 2 y y x 2 y y 1 / 2 x y Solution (a) In Example 6 of Section 10.2, we have shown x c1 (2 cos 2t 2 sin 2t ) c2 (2 cos 2t 2 sin 2t ) y c1 ( cos 2t ) c2 sin 2t Ch11_16 Example 5 (2) Thus every solution is periodic with period . The solution satisfying X(0) = (2, 0) is x = 2 cos 2t + 2 sin 2t, y = – sin 2t See Fig 11.4(a). Ch11_17 Example 5 (3) (b) Using the similar method, we have x c1 (2et cos t ) c2 (2et sin t ) y c1 (et sin t ) c 2 (et cos t ) Since the presence of et, there are no periodic solutions. The solution satisfying X(0) = (2, 0) is x 2et cos t , y et sin t See Fig 11.4(b). Ch11_18 Fig 11.4(b) Ch11_19 Changing to Polar Coordinates Please remember that the transformations are r2 = x2 + y2 and = tan–1(y/x), dr 1 dx dy d 1 dx dy x y , 2 y x dt r dt dt dt r dt dt Ch11_20 Example 6 Find the solution of the following system 2 2 x y x x y y x y x 2 y 2 satisfying X(0) = (3, 3). Solution dr 1 [ x( y xr ) y ( x yr )] r 2 dt r d 1 2 [( y )( y xr ) x( x yr )] 1 dt r Ch11_21 Example 6 (2) Since (3, 3) is (3 2, / 4) in polar coordinates, then X(0) = (3, 3) becomes r (0) 3 2 and (0) =π/4. Using separation of variables, we have the solution is 1 r , t c2 t c1 for r 0. Applying the initial conditions, we have 1 r , t t 2/6 4 Ch11_22 Example 6 (3) The graph of 1 r 2/6 /4 is shown in Fig 11.5. Ch11_23 Fig 11.5 Ch11_24 Example 7 Consider the system in polar coordinates: dr d 0.5(3 r ), 1 dt dt Find and sketch the solutions satisfying X(0) = (0, 1) and X(0) = (3, 0) in rectangular coordinates. Solution By separation of variables, we have r 3 c1e 0.5t , t c2 Ch11_25 Example 7 (2) If X(0) = (0, 1), then r(0) = 1 and (0) = /2. Thus c1 = –2, c2 =/2. The solution curve is the spiral . Notice that as t →, r 3 2e0.5( / 2) increases without bound and r approaches 3. If X(0) = (3, 0), then r(0) = 3 and (0) = 0. Thus c1 = c2 = 0 and r = 3, = t. We have the solution is x = r cos = 3 cos t and y = r sin = 3 sin t. It is a periodic solution. See Fig 11.6. Ch11_26 Fig 11.6 Ch11_27 11.2 Stability of Linear Systems Some Fundamental Questions Suppose X1 is a critical point of a plane autonomous system and X = X(t) is a solution satisfying X(0) = X0. Our questions are when X0 is near X1: (i) Is limt X(t) = X1? (ii) If the answer of (i) is “no”, does it remain close to X1 or move away from X1? See Fig 11.7 Ch11_28 Fig 11.7 Ch11_29 Referring to Fig11.7(a) and (b), we call the critical point locally stable. However, if an initial value results in behavior similar to (c) can be found in nay given neighborhood, we call the critical point unstable. Ch11_30 Stability Analysis Consider x = ax + by y = cx + dy We have the system matrix as a b A c d To ensure that X0 = (0, 0) is the only critical point, we assume the determinant = ad – bc 0. Ch11_31 Then det (A – I) = 0 becomes 2 − + = 0 where = a + d. Thus ( 2 4 ) / 2 Ch11_32 Example 1 Find the eigenvalues of the system x x y y cx y in terms of c, and use a numerical solver to discover the shape of solutions corresponding to the case c = ¼ , 4, 0 and −9. Ch11_33 Example 1 (2) Solution Since the coefficient matrix is 1 1 c 1 then we have = −2, and = 1 – c. Thus 2 4 4(1 c) 1 c 2 Ch11_34 Example 1 (3) If c = ¼ , = −1/2 and −3/2. Fig 11.8(a) shows the phase portrait of the system. When c = 4, = 1 and 3. See Fig 11.8(b). Ch11_35 Example 1 (4) When c = 0, = −1. See Fig 11.8(c). When c = −9, = −1 3i. See Fig 11.8(d). Ch11_36 Case I: Real Distinct Eigenvalues According to Sec 10.2, the general solution is X(t ) c1K1e 1t c2 K 2e 2t e1t (c1K1 c2 K 2e( 2 1 )t ) (a) Both eigenvalues negative: Stable Node It is easier to check that under this condition, X(t) 0 as t See Fig 11.9. Ch11_37 Fig 11.9 Ch11_38 (b) Both eigenvalues positive: Unstable Node It is easier to check that under this condition, |X(t)| becomes unbounded as t See Fig 11.10 Ch11_39 Fig 11.10 Ch11_40 (c) Eigenvalues have opposite signs (2 < 0 < 1): Saddle Point On if c1 = 0, will approach 0 along the line determined by K2 as t . This unstable solution is called a saddle point. See Fig 11.11. Ch11_41 Fig 11.11 Ch11_42 Example 2 Classify the critical point (0, 0) of each system X = AX as either a stable node, an unstable node, or a saddle point. (a) (b) 2 3 10 6 A A 2 1 15 19 Solution (a) Since the eigenvalues are 4, −1, (0, 0) is a saddle point. The corresponding eigenvectors are respectively 1 3 K2 K1 1 2 Ch11_43 Example 2 (2) If X(0) lies on the line y = −x, then X(t) approaches 0. For any other initial conditions, X(t) will become unbounded in the direction determined by K1. That is, y = (2/3)x serves an asymptote. See Fig 11.12. Ch11_44 Fig 11.12> Ch11_45 (b) Since the eigenvalues are − 4, −25, (0, 0) is a stable node. The corresponding eigenvectors are respectively 1 K1 1 2 K2 5 See Fig 11.13. Ch11_46 Fig 11.13 Ch11_47 Case II: A Repeated Real Eigenvalue According to Sec 10.2, we have the following conditions. (a) Two linearly independent eigenvectors The general solution is X(t ) c1K1e1t c2K 2e1t (c1K1 c2K 2 )e1t If 1 < 0, then X(t) approaches 0 along the line determined by c1K1 + c2K2 and the critical point is called a degenerate stable node. Fig 11.14(a) shows the graph for 1 < 0 and the arrows are reversed when 1 > 0, and is called a degenerate unstable node. Ch11_48 Fig 11.14 Ch11_49 (b) A single linearly independent eigenvector When we only have a single eigenvector, the general solution is X(t ) c1K1e1t c2 (K1te 1t Pe1t ) c1 c2 te (c2K1 K1 P) t t If 1 < 0, then X(t) approaches 0 in one of directions determined by K1(See Fig 11.14(b)). This critical point is again called a degenerate stable node. If 1 > 0, this critical point is again called a degenerate unstable node. 1t Ch11_50 Case III: Complex eigenvalues (2 – 4 < 0) (a) Pure imaginary roots (2 – 4 < 0, = 0) We call this critical point a center. See Fig 11.15 Ch11_51 (b) Nonzero real part (2 – 4 < 0, 0) real part > 0: unstable spiral point (Fig 11.16(a)) real part < 0: stable spiral point (Fig 11.16(b)) Ch11_52 Example 3 Classify the critical point (0, 0) of each system 3 18 1 2 (a) A (b) A 2 9 1 1 Solution (a) The characteristic equation 2 + 6 + 9 = ( + 3)2= 0 so (0, 0) is a degenerate stable node. (b) The characteristic equation 2 + 1 = 0 so (0, 0) is a center. Ch11_53 Example 4 Classify the critical point (0, 0) of each system 3.10 1.01 (a) A 1.10 1.02 axˆ abxˆ (b) A cdyˆ dyˆ for positive constants. Solution (a) = −0.01, = 2.3789, 2 − 4 < 0: (0, 0) is a stable spiral point. Ch11_54 Example 4 (2) (b) (axˆ dyˆ ) 0, adxˆyˆ (1 bc) if bc 1, 0 : a saddle point if bc 1, 0 : either satble, degenerate stable, or stable spiral Ch11_55 THEOREM 11.1 Stability Criteria for Lonear Systems For a linear plane autonomous system X’ = AX with det A 0, let X = X(t) denote the solution that satisfies the initial condition X(0) = X0, where X0 0. (a) limt→X(t) = 0 if and only if the eigenvalues of A have negative real parts. This will occur when ∆ > 0 and < 0. (b) X(t) is periodic if and only if the eigenvalues of A are pure imaginary. This will occur when ∆ > 0 and = 0. (c) In all other cases, given any neighborhood of the region, there is at least one X0 in the neighborhood for which X(t) becomes unbounded as t increases. Ch11_56 11.3 Linearization and Local Stability DEFINITION 11.1 Stable Critical Points Let X1 be a critical point of an autonomous system, and let X = X(t) denote the solution that satisfies the initial condition X(0) = X0, where X0 X1. We say that X1 is a stable critical point when, given any radius ρ > 0, there is a corresponding radius r > 0 such that if the initial position X0 satisfies │X0 – X1│< r, then the corresponding solution X(t) satisfies │X(t) – X1│ < ρ for all t > 0. If, in addition limt→X(t) = X1 whenever │X0 – X1│< r, we call X1 an asymptotically stable critical point. Ch11_57 This definition is shown in Fig 11.20(a). To emphasize that X0 must be selected close to X1, we also use the terminology locally stable critical point. Ch11_58 DEFINITION 11.2 Unstable Critical Points Let X1 be a critical point of an autonomous system, and let X = X(t) denote the solution that satisfies the initial condition X(0) = X0, where X0 X1. We say that X1 is an unstable critical point if there is a disk of radius ρ > 0 with the property that, for any r > 0, there is at least one initial position X0 satisfies │X0 – X1│< r, yet the corresponding solution X(t) satisfies │X(t) – X1│ ρ for all t > 0. Ch11_59 If a critical point X1 is unstable, no matter how small the neighborhood about X1, an initial position X0 can be found that the solution will leave some disk at some time t. See Fig 11.20(b). Ch11_60 Example 1 Show that (0, 0) is a stable critical point of the system x' y x x 2 y 2 y' x y x2 y 2 Solution In Example 6 of Sec 11.1, we have shown r = 1/(t + c1), = t + c2 is the solution. If X(0) = (r0, 0), then r = r0/(r0 t + 1), = t +0 Note that r < r0 for t > 0, and r approaches (0, 0) as t increase. Hence the critical point (0, 0) is stable and is in fact asymptotically stable. See Fig 11.21. Ch11_61 Fig 11.21 Ch11_62 Example 2 Consider the plane system dr 0.05r (3 r ) dt d 1 dt Show that (x, y) = (0, 0) is an unstable critical point. Ch11_63 Example 2 (2) Solution Since x = r cos and y = r sin , we have dx d dr r sin cos dt dt dt dy d dr r cos sin dt dt dt Since dr/dt = 0.05r(3-r), then r = 0 implies dr/dt = 0. Thus when r = 0, we have dx/dt = 0, dy/dt = 0. We conclude that (x, y) = (0, 0) is a critical point. Ch11_64 Example 2 (3) Solving the given differential equation with r(0) = r0 and r0 0, we can have 3 r 1 c0e 0.15t where c0 (3 r0 ) / r0 . Since 3 lim t 3 0.15t 1 c0e No matter how close to (0, 0) a solution starts, the solution will leave (0, 0). Thus (0, 0) is an unstable critical point. See Fig 11.22. Ch11_65 Fig 11.22 Ch11_66 Linearization If we write the systems in Example 1 and 2 as X = g(X). The process to find a liner term A(X – X1) that most closely approximates g(X) is called linearization. Ch11_67 THEOREM 11.2 Stability Criteria for Linear Systems Let x1 be a critical point of the autonomous differential equation x = g(x), where g is differentiable at x1. (a) If g(x1) < 0, then x1 is an asymptotically stable critical point. (b) If g(x1) > 0, then x1 is an unstable critical point. Ch11_68 Example 3 5 Both x and x are critical points of 4 4 x' cos x sin x. predict the behavior of solutions near these two critical points. Since g ' ( x) sin x cos x g ' ( 4) 2 0, g ' (5 4) 2 0 Therefore x = /4 is an asymptotically stable critical point but x = 5/4 is unstable. See Fig 11.23. Ch11_69 Fig 11.23 Ch11_70 Example 4 Without solving explicitly, analyze the critical points of the system x = (r/K)x(K – x), where r and K are positive constants. Solution We have two critical points x = 0 and x = K. Since r g ' ( x) ( K 2 x) K g ' (0) r 0, g ' ( K ) r 0 Therefore x = K is an asymptotically stable critical point but x = 0 is unstable. Ch11_71 Jacobian Matrix An equation of the tangent plane to the surface z = g(x, y) at X1 = (x1, y1) is g g z g ( x1 , y1 ) ( x1 , y1 ) ( x x1 ) ( x1 , y1 ) ( y y1 ) x y Similarly, when X1 = (x1, y1) is a critical point, then P (x1, y1) = 0, Q (x1, y1) = 0. We have P P x ' P ( x, y ) ( x1 , y1 ) ( x x1 ) ( x1 , y1 ) ( y y1 ) x y Q Q y ' Q ( x, y ) ( x1 , y1 ) ( x x1 ) ( x1 , y1 ) ( y y1 ) x y Ch11_72 The original system X = g(X) may be approximated by X = A(X – X1), where P ( x1 , y1 ) x A Q x ( x1 , y1 ) P ( x1 , y1 ) y Q (x ,y ) y 1 1 This matrix is called the Jacobian Matrix at X1 and is denoted by g(X1). Ch11_73 THEOREM 11.3 Stability Criteria for Plane Autonomous Systems Let X1 be a critical point of the autonomous differential equation X’ = g(X), where P(x, y) and Q(x, y) have continuous first partials in a neighborhood of X1. (a) If the eigenvalues of A = g’(X1) have negative real part, then X1 is an asymptotically stable critical point. (b) If A = g’(X1) has an eigenvalue with positive real part, then X1 is an unstable critical point. Ch11_74 Example 5 Classify the critical points of each system. (a) x’ = x2 + y2 – 6 (b) x’ = 0.01x(100 – x – y) y’ = x2 – y y’ = 0.05y(60 – y – 0.2x) Solution (a) The critical points are ( 2, 2) and ( 2, 2), 2x 2 y g ' ( X) 2 x 1 Ch11_75 Example 5 (2) 2 2 4 A1 g ' (( 2, 2)) 2 2 1 2 2 4 A 2 g ' (( 2, 2)) 2 2 1 Since the determinant of A1 is negative, A1 has a positive real eigenvalue. Therefore ( 2, 2) unstable. A1 has a positive determinant and a negative trace. Both the eigenvalues have negative real parts. Therefore ( 2, 2) is stable. Ch11_76 Example 5 (3) (b) The critical points are (0, 0), (0, 60), (100, 0), (50, 50). The Jacobiam matrix is 0.01x 0.01(100 2 x y ) g ' ( X) 0.01y 0.05(60 2 y 0.2 x) 1 0 A1 g ' ((0, 0)) 0 3 0 .4 0 A 2 g ' ((0, 60)) 0 .6 3 Ch11_77 Example 5 (4) 1 1 A3 g ' ((100,0)) 0 2 0.5 0.5 A 4 g ' ((50,50)) 0.5 2.5 Checking the signs of the determinant and trace of each matrix, we conclude that (0, 0) is unstable; (0, 60) is unstable; (100, 0) is unstable; (50, 50) is stable. Ch11_78 Example 6 Classify each critical point of the system in Example 5(b). Solution For the matrix A1 corresponding to (0, 0), = 3, = 4, 2 – 4 = 4. Therefore (0, 0) is an unstable node. The critical points (0, 60) and (100, 0) are saddles since < 0 in both cases. For A4, > 0, < 0, (50, 50) is a stable node. Ch11_79 Example 7 Consider the system x + x – x3 = 0. We have x = y, y = x3 – x. Find and classify the critical points. Solution x x x( x 1) 0, the critical points are 3 2 (0, 0), (1, 0), (-1,0). Ch11_80 Example 7 (2) The corresponding matrices are 0 1 A1 g ' ((0,0)) 1 0 0 1 A 2 g ' ((1,0)) g ' (( 1,0)) 2 0 Since det A 2 0, (1, 0) and (-1, 0) are both saddle points. The eigenvalue s of A1 are i and the status of (0, 0) is in doubt. Ch11_81 Example 8 Use the phase-plane method to classify the sole critical point (0, 0) of the system x = y2 y = x2 Solution The determinant of the Jacobian matrix 0 2y g ' ( X) 2x 0 is 0 at (0, 0), and so the nature of (0, 0) is in doubt. Ch11_82 Example 8 (2) Using the phase-plane method, we get dy dy / dt x 2 2 dx dx / dt y 2 2 3 3 y dy x dx , or y x c If X(0) (0, y0 ), y x y0 or y x y0 . 3 3 3 3 3 3 Fig 11.26 shows a collection of solution curves. The critical point (0, 0) is unstable. Ch11_83 Fig 11.26 Ch11_84 Example 9 Use the phase-plane method to determine the nature of the solutions to x + x − x3 = 0 in a neighborhood of (0, 0). Solution If we let dx / dt y then dy / dt x3 3 x. dy dy / dt x 3 x dx dx / dt y 3 2 4 2 y x x ydy ( x 3x)dx, or 2 4 2 c, 2 2 ( x 1) 2 thus y c0 . 2 3 Ch11_85 Example 9 (2) ( x0 1) 2 If X(0) ( x0 ,0), 0 x0 1, then c0 , 2 2 2 2 2 2 2 ( x 1 ) ( x 1 ) ( 2 x x )( x x ) 2 0 0 0 y 2 2 2 2 2 2 Note that y = 0 when x = −x0 and the right-hand side is positive when −x0 < x < x0. So each x has two corresponding values of y. The solution X = X(t) that satisfies X(0) = (x0, 0) is periodic, and (0, 0) is a center. See Fig 11.27. Ch11_86 Fig 11.27 Ch11_87 11.4 Autonomous Systems as Mathematical Models Nonlinear Pendulum Consider the nonlinear second-order differential equation d 2 g sin 0 2 l dt When we let x = , y = , then we can write x y g y sin x l Ch11_88 The critical points are (k, 0) and the Jacobian matrix is 0 1 g ' (( k ,0)) k 1 g 0 (1) l If k = 2n + 1, < 0, and so all critical points ((2n +1), 0) are saddle points. Particularly, the critical point (, 0) is unstable as expected. See Fig 11.28. Ch11_89 Fig 11.28 Ch11_90 When k = 2n, the eigenvalues are pure imaginary, and so the nature of these critical points remains in doubt. Since we assumed that there are no damping forces, we expect that all the critical points ((2n, 0) are centers. From dy dy / dt g sin y 2g 2 , then y cos x c dx dx / dt l y l 2g 2 If X(0) ( x0 , 0), then y (cos x cos x0 ) l Ch11_91 Note that y = 0 when x = −x0, and that (2g/l)(cos x – cos x0) > 0 for |x| < |x0| < . Thus each such x has two corresponding values of y, and so the solution X = X(t) that satisfies X(0) = (x0, 0) is periodic. We may conclude that (0, 0) is a center. See Fig 11.29. Ch11_92 Fig 11.29 Ch11_93 Example 1 A pendulum is an equilibrium position with = 0 is given an initial velocity of 0 rad/s. Determine under what conditions the resulting motion is periodic. Solution The initial condition is X(0) = (0, 0). 2g From y (cos x c), it follows that l 2g l 2 2 y (cos x 1 0 ) l 2g 2 Ch11_94 Example 1 (2) To establish that the solution X(t) is periodic it is sufficient to show that there are two x-intercepts x = x0 between − and and that the right-hand side is positive for |x| < |x0|. Each such x then has two corresponding values of y. If y = 0, cos x = 1 – (l/2g)02, and this equation has two solutions x = x0 between − and , provided 1 – (l/2g)02 > −1. Note that (l/2g)(cos x – cos x0) is positive for |x| < |x0|. The restriction on the initial angular velocity may be written as g 0 2 l Ch11_95 Nonlinear Oscillations: The Sliding Bead See Fig 11.30. The tangential force F has the magnitude mg sin , thus Fx = − mg sin cos . Since tan = f (x), then f ' ( x) Fx mg sin cos mg 1 [ f ' ( x)]2 dx Assume there is a damping force D, and Dx , dt f ' ( x) From Newton' s law : mx" mg x' 2 1 [ f ' ( x)] Ch11_96 and the corresponding plane autonomous system is x' y f ' ( x) y' g y 2 1 [ f ' ( x)] m If X1 = (x1, y1) is a critical point, then y1 = 0 and f (x1) = 0. The bead must be at rest at a point on the wire where the tangent line is horizontal. Ch11_97 The Jacobian matrix at X1 is 0 1 , and so g' ( X1 ) gf " ( x1 ) m , gf " ( x1 ), 4 m 2 2 m 2 4 gf " ( x1 ). Ch11_98 We can make the following conclusions. (i) f ”(x1) < 0 : A relative maximum occurs at x = x1 and since < 0, an unstable saddle point occurs at X1 = (x1, 0). (ii) f ”(x1) > 0 and > 0: A relative minimum occurs at x = x1 and since < 0 and > 0, X1 = (x1, 0) is a stable critical point. If 2 > 4gm2f (x1), the system is overdamped and the critical point is a stable node. Ch11_99 If 2 < 4gm2f (x1), the system is underdamped and the critical point is a stable spiral point. The exact nature of the stable critical point is still in doubt if 2 = 4gm2f (x1). (iii) f ”(x1) > 0 and the system is undamped ( = 0): In this case the eigenvalues are pure imaginary, but the phase plane method can be used to show that the critical point is a center. Thus solutions with X(0) = (x(0), x(0)) near X1 = (x1, 0) are periodic. Ch11_100 Example 2 A 10-gram bead slides along the graph z = sin x. The relative minima at x1 = −/2 and x2 = 3/2 are stable critical points. See Fig 11.31. Ch11_101 Fig 11.32 Fig 11.32 shows the motions when the critical points are stable spiral points. Ch11_102 Fig 11.33 Fig 11.33 shows a collection of solution curves for the undamped case. Ch11_103 Lotka-Volterra Predator-Prey Model Recall the predator-prey model: x' ax bxy x( a by ) y ' cxy dy y (cx d ) The critical points are (0,0) and (d/c,a/b), then a 0 A1 g ' ((0,0)) and 0 d bd / c 0 A 2 g ' (( d / c, a / b)) . 0 ac / b Ch11_104 Fig 11.34 The critical point (0, 0) is a saddle point. See Fig 11.34. Ch11_105 Since A2 has the pure imaginary eigenvalues, the critical point may be a center. Since dy y (cx d ) , then dx x( a by ) a by cx d y dy x dy, a ln y by cx d ln x c1 , or d cx (x e a by )( y e ) c0 Ch11_106 Fig 11.35 Typical graphs are shown in Fig 11.35. Ch11_107 1. If y = a/b, the equation F(x)G(y) = c0 has exactly two solutions xm and xM that satisfy xm < d/c < xM. 2. If xm < x1 < xM and x = x1, then F(x)G(y) = c0 has exactly two solutions y1 and y2 that satisfy y1 < a/b < y2. 3. If x is outside the interval [xm, xM], then F(x)G(y) = c0 has no solutions. The graph of a typical periodic solution is shown in Fig 11.36. Ch11_108 Fig 11.36 Ch11_109 Example 3 If we let a = 0.1, b = 0.002, c = 0.0025, d = 0.2, the critical point in the first quadrant is (d/c, a/b) = (80, 50), and we know it is a center. See Fig 11.37. Ch11_110 Fig 11.37 Ch11_111 Lotka-Volterra Competition Model Consider the model: r1 x' x( K1 x 12 y ) K1 r2 y' y ( K 2 y 21 x) (1) K2 This system has critical points at (0, 0), (K1, 0) and (0, K2). Ch11_112 Example 4 Consider the model x' 0.004 x(50 x 0.75 y ) y ' 0.001y (100 y 3.0 x) Find and classify all critical points. Solution Critical points are (0, 0), (50, 0), (0, 100), (20, 40). Since 1221 = 2.25 > 1, and so the critical point (20, 40) is a saddle point. The Jacobian matrix is Ch11_113 Example 4 (2) 0.003x 0.2 0.008x 0.003 y g ' ( X) 0.003 y 0.1 0.002 y 0.003x 0.2 0 A1 g ' ((0,0)) 0 0. 1 0.2 0.15 A 2 g ' ((50,0)) 0.05 0 0.08 0.12 A3 g' (( 20,40)) 0.06 0.04 0 0.1 A 4 g' ((0,100)) 0.3 0.1 Ch11_114 Example 4 (3) Therefore (0, 0) is unstable, whereas both (50, 0) and (0, 100) are stable nodes and (20, 40) is a saddle point. Ch11_115 11.5 Periodic Solutions, Limit Cycles and Global Stability THEOREM 11.4 Cycles and Critical Points If a plane autonomous system has a periodic solution X = X(t) in a simply connected region R, then the system has at least one critical point inside the corresponding simple closed curve C. If there is a single critical point inside C, then that critical point cannot be a saddle point. COROLLARY If a simply connected region R either contains no critical points of a plane autonomous system or contains a single saddle point, then there are no periodic solutions in R. Ch11_116 Example 1 Show that the plane autonomous system x’ = xy y’ = −1 – x2 – y2 has no periodic solutions. Solution If (x, y) is a critical point, then either x = 0 or y = 0. If x = 0, then −1 – y2 = 0, y2 = –1. Likewise, y = 0 implies x2 = –1. Thus this system has no critical points and has no periodic solutions. Ch11_117 Example 2 Show that x' 0.004 x(50 x 0.75 y ) y ' 0.001y (100 y 3.0 x) has no periodic solutions in the first quadrant. Solution From Example 4 of Sec 11.4, we knew only (20, 40) lies in the first quadrant and (20, 40) is a saddle point. By the corollary, there are no periodic solutions in the first quadrant. Ch11_118 THEOREM 11.5 Bendixson Negative Criterion If div V = P/y + Q/ y does not change sign in a connected region R, then the plane autonomous system has no periodic solution in R. Ch11_119 Example 3 Investigating possible periodic solutions of each system. (a) x' x 2 y 4 x3 y 2 , y ' x 2 y yx 2 y 3 (b) x' y x(2 x 2 y 2 ), y ' x y (2 x 2 y 2 ) Solution (a) div V P / x Q / y 1 12 x 2 2 x 2 3 y 2 3, and so there are no periodic solutions. Ch11_120 Example 3 (2) (b) div V (2 3x 2 y 2 ) (2 x 2 3 y 2 ) 4 4( x 2 y 2 ) If R is the interior of the given circle, div V > 0 and so there are no periodic solutions inside the disk. Note that div V < 0 on the exterior of the circle. If R is any simply connected subset of the exterior, then there are no periodic solutions in R. If there is a periodic solution in the exterior, it must enclose the circle x2 + y2 = 1. Ch11_121 Example 4 The sliding bead in Sec 11.4 satisfies f ' ( x) mx" mg x' 2 1 [ f ' ( x)] Show that there are no periodic solutions. Solution f ' ( x) x' y, y ' g y 2 1 [ f ' ( x)] m P Q div V 0 x y m Ch11_122 THEOREM 11.6 Dulac Negative Criterion If (x, y) has continuous first derivatives in a simply ( P ) ( Q ) connected region R and does not change x y sign in R, then the plane autonomous system has no periodic solution in R. Ch11_123 Example 5 Show that x" x 2 ( x') 2 x x' has no periodic solutions. Solution x' y, y ' x 2 y 2 x y. Letting δ ( x, y ) e axby , (P ) (Q) x y e axby (ay 2 y 1) e axbyb( x 2 y 2 x y ) Ch11_124 Example 5 (2) If we set a = −2, b = 0, then (P ) (Q ) e ax by x y which is always negative. The system has no periodic solutions. Ch11_125 Example 6 Use (x, y) = 1/(xy) to show r1 x' x( K1 x 12 y ) K1 r2 y' y ( K 2 y 21 x) K2 have no periodic solutions in the first quadrant. Ch11_126 Example 6 (2) Solution r1 K1 x r2 K 2 y P ( 12 ), Q ( 21 ) K1 y y K2 x x (P ) (Q ) r1 1 r2 1 ( ) ( ) x y K1 y K 2 x For (x, y) in the first quadrant, the last expression is always negative. Ch11_127 DEFINITION 11.3 Invariant Region A region R is called an invariant region for a plane autonomous system if, whenever X0 is in R, the X = X(t) satisfying X(0) = X0 remains in R. Fig 11.40 shows two standard types of invariant regions. Ch11_128 Fig 11.40 Ch11_129 THEOREM 11.7 Normal Vector and Invariant Regions If n(x, y) denote a normal vector on the boundary that point inside the region, then R will be an invariant region for the plane autonomous system provided V(x, y)‧n(x, y) 0 for all points (x, y) on the boundary. Ch11_130 Example 7 Find a circular region with center at (0, 0) that serves an invariant region for the system x' y x3 y' x y3 Solution For the circle x2 + y2 = r2, n = (−2x, −2y) is a normal vector that points toward the interior of the circle. Since V n ( y x3 , x y3 ) (2 x,2 y) 2( x 4 y 4 ) we may conclude that Vn 0 on the circle x2 + y2 = r2. By Theorem 11.7, the circular region x2 + y2 r2 serves as an invariant region for the system for any r > 0. Ch11_131 Example 8 Find an annular region bounded by circles that serves as an invariant region for the system 2 2 5 x' x y 5 x( x y ) x y' x y 5 y( x2 y 2 ) y5 Solution As in Example 7, the normal vector n1 = (−2x, −2y) points inside the circle x2 + y2 = r2, while the normal vector n2 = − n1 points outside the circle. V n1 2(r 2 5r 4 x6 y 6 ) Ch11_132 Example 8 (2) If r = 1, V‧n1 = 8 – 2(x6 + y6) 0. If r = 1/4, V‧n1 – 2(r2 – 5r4) < 0 and so V‧n2 > 0. The annular 1/16 x2 + y2 1 is an invariant region. Ch11_133 Example 9 The Van der Pol equation is a nonlinear second-order differential equation that arise in electronics, x' y y ' ( x 2 1) y x Fig 11.41 shows the vector field for = 1, together with the curves y = 0 and (x2 – 1)y = −x along which the vectors are vertical and horizontal, respectively. Ch11_134 Fig 11.41 It is not possible to find a simple invariant region whose boundary consists of lines or circles. Ch11_135 THEOREM 11.8 Poincare-Bendixson I Let R be an invariant region for a plane autonomous system and suppose that R has no critical points on its boundary. (a) If R is a Type I region that has a single unstable node or an unstable spiral point in its interior, then there is at least one periodic solution in R. (b) If R is a Type II region that contains no critical points of the system, then there is at least one periodic solution in R. In either of the two cases, if X = X(t) is a nonperiodic solution in R, then X(t) spirals toward a cycle that is a solution to the solution to the system. This periodic solution is called a limit cycle. Ch11_136 Example 10 Use Theorem 11.8 to show that x' y x(1 x 2 y 2 ) y ( x 2 y 2 ) y ' x y (1 x 2 y 2 ) x( x 2 y 2 ) has at least one periodic solution. Solution We first construct an invariant region bounded by circles. If n1 = (−2x, −2y) then V n1 2r 2 (1 r 2 ) Ch11_137 Example 10 (2) If we let r = 2 and then r = ½, we conclude that the annular region R: ¼ x2 + y2 4 is invariant. If (x1, y1) is a critical point, then V‧n1 = (0, 0)‧n1 = 0. Therefore r = 0 or r = 1. If r = 0, then (x1, y1) = (0, 0) is a critical point. If r = 1, the system reduces to −2y = 0, 2x = 0 and we have reached a contradiction. Therefore (0, 0) is the only critical point and is not in R. Thus the system has at least one periodic solution in R. Ch11_138 Example 11 Show that the Van der Pol equations has a periodic solution when > 0. Solution We found that the only critical point is (0, 0) and the Jacobian matrix is 0 1 g ' ((0, 0)) , then 1 , 1, 2 4 2 4 Ch11_139 Example 11 (2) Since > 0, the critical point is either an unstable spiral point or an unstable node. Bu part (i) of Theorem 11.8 the system has at least one periodic solution in R. See Fig 11.42. Ch11_140 Fig 11.42 Ch11_141 THEOREM 11.8 Poincare-Bendixson II Let R be a Type I invariant region for a plane autonomous system that has no periodic solution in R. (a) If R has a finite number of nodes or spiral points, then given any initial position X0 in R, limt→X(t) = X1 for some critical point X1. (b) If R has a single stable node or stable spiral point X1 in its interior and no critical points on its boundary, the limt→X(t) = X1 for all initial position X0 in R. Ch11_142 Example 12 Investigate global stability for the system in Example 7. x' y x3 y' x y3 Solution It is not hard to show that the only critical point is (0, 0) and the Jacobian matrix is 0 1 g' ((0, 0)) , and 0, 1 1 0 Ch11_143 Example 12 (2) (0, 0) may be either a stable or an unstable spiral. Theorem 11.9 guarantees that lim X(t ) X1 for some critical points X1. t Since (0, 0) is the only critical point, we must have lim X(t ) (0, 0) for any initial position X(0) in the plane. t The critical point is therefore a globally stable spiral point. See Fig 11.43. Ch11_144 Fig 11.43 Ch11_145