CHAPTER 3. Systems of Differential Equations, Phase Plane, Qualitative Methods Sec.3.1 Introductory Examples Example 2. Spend time on Fig. 76 until you feel that you fully understand the difference between (b) (the usual representation in calculus) and (c), because trajectories will play an important role throughout this, ‘chapter. Try to understand the reasons for the following. The trajectory stars atthe origin. It reaches its highest point where y, has a maximum (before ¢ = 1). Ithas a vertical tangent where /, has a maximum, short after ¢ = 1. Ast increases from there to 1 = 5, the trajectory goes downward until it almost reaches the fy-axis at 3; this point is a limit as t ~ oo. In terms of # the trajectory goes up faster than it comes down, Problem Set 3.1. Page 158 5. Electrical network. The problem amounts to the determination of the two arbitrary constants in a general solution of a system of two differential equations in two unknown functions /, and /,. representing the ‘currents in an electrical network shown in Fig. 76 in Sec. 3.1. You will see that this is quite similar to the ‘corresponding task for a single second-order differential equation. That solution is given by (6), in components NO = 2eve M+ eye 43, (= cre? + 0.8c,e™, Setting ¢ = 0 and using the given initial conditions /,(0) = 9, 2(0) = 0 gives 1(0) = 2c) $0.43 =9 @ 13(0) = ¢ + 0.8¢2 = 0. ) From (b) you have c, = ~0.8¢. Substituting this nto (a) and simplifying gives 2(-O.8¢3) +c = 6, hence ~0.6¢2 = 6 oF ¢2 = -10, and cy = 8. The answer is (note that 2c, = 16) 10) = 166 ~ 106-98" +3 Ayo) = 87-86%, ‘These currents are shown in the figure. (1) has the limit 3, as expected. /9(1) comes out negative: this ‘means it is directed opposite to the arrows shown in Fig. 76(a). which had been assumed arbitrarily at the beginning of the process of modeling; this had to be done because atthe beginning, one does not know in ‘what directions the currents will actually flow. Section 3.1. Problem 5. Currents 1, (upper curve) and its limit 3 and /y 9. Conversion of single differential equations to a system is an important process, which allways follows the pattern shown in formulas (9) and (10) of See. 3.1. The present equation y" ~ 9y = O can be readily24 Ordinary Differential Equations Part A solved. A general solution is y = c) e% + ce". The point of the problem is not to explain a (complicated) solution method for a simple problem, but to explain the relation between systems and single equations and their solutions. In the present case the formulas (9) and (10) give y, = y.y2 = y’ and Yay = (because the given equation can be written y” = 9y, hence y = ys, but yf = ys). In matrix form (asin Example 3 of the text this is ‘The characteristic equation is awa] = B-960. ‘The eigenvalues are 2, = 3 and Ay 2= Ay. thatis, 3. For 2, you obtain an eigenvector from (13) in Sec. 3.0 with wim [25 2} [ee] From the first equation ~3) +.x2 = O you have x = 31y. An eigenvector is determined only up toa nonzero constant. Hence, inthe present case. a convenient choice isx, = 1.x, = 3. The second equation gives the same result and is not needed. For the second eigenvalue, by = ~3, the procedure isthe same, namely, 31 fa ][ sue anme (5S -ee] You now have 3x, +4; = 0, hence x; = ~3x,, and can choose x) = 1, x2 obtained are 3. The eigenvectors x= () 3 and x@ = (1-3) Multiplying these by e* and e™™, respectively, and taking a linear combination involving two arbitrary constants ¢, and ca gives a general solution of the present system in the form yredl 3Te+eg{1 -3]"e* In components, this is ye cre tere ya = Beye" Bene. Here you sce that y, = y is a general solution ofthe given equation, and y2 = y} = y’ isthe derivative of this solution, as had to be expected because of the definition of ya atthe beginning of the process. Incidentally, you can use y» = y} for checking your result. Sec, 3.3 Homogeneous Systems with Constant Coefficients. Phase Plane, Critical Points Example 2. The characteristic equation is det (A ~ Al) ‘Thus a = 1 isan eigenvalue. Any nonzero vector with two components is an eigenvector because AX for any x: indeed, A is the 2 x 2 unit matrix! Hence you can take x‘ = (1 OJ? and x® = (0 1]F or any xChap. 3 ‘Systems of Differential Equations, Phase Plane, Qualitative Methods 25 other two linearly independent vectors with two components. This gives the solution on p. 165. Example 3. (1 — 4)(-1 4) = (2~ 1)(2 + 1) = 0. and so on [33] = (1-AC1-A)- 13 = 2-420 Problem Set 3.3. Page 169 3. General solution. The matrix ofthe system is ‘The characteristic equation is det (A a1) a Hence the eigenvalues are h, = 2 and Ay = ~2. An eigenvector corresponding A, is obtained by solving (A=AiDx = (A= 2Dx = 0, in components, (2x + x0 3x, + C12) = 0. Each of the equations gives x, = x; (and you need only one of them). Hence you can take x" = (117 as an eigenvector corresponding to A, = 2. For hy = ~2 those component equations are (1 +2)r +; = 0 | (both of the same form) and you can take x, = 1, x2 = ~3, so that an eigenvector corresponding 0 Ay = ~2isx® = [1 -3]". This gives as a general solution of the system seal fool Te yes cre beet in components, ya = ce —3ene™ ‘This agrees with the answer in Appendix 2, with c, and cs interchanged. (Of course, the notation for arbitrary constants is up tous.) 5. General solution. In this problem you will see the typical calculations in the case of complex eigenvalues. ‘The matrix of the system is 1-1 A= to This gives the characteristic equation ! Ina ol det(A =a) = : y 1ood-a ‘The eigenvalues are complex conjugates, 2; = 1 +i and hy = 1 ~ i. Bigenvectors x are obtained from | (A=ADx = 0, as before. This is a vector equation, and you need only the first ofthe two corresponding scalar equations. For A= Ay = 1+ iitis Q-(4))x =x thus sizer say, So the new aspect is that this eigenvector x" = (1 2 = hy = 1 =i you get an eigenvector from = iJ" is no longer real but is complex. Similarly, for26. 15. Ordinary Differential Equations Part A (-(-))x a = thus ixpext say, melo and xe “This gives the eigenvector x® = [1 (J7. Using the Euler formula et = cosrtisint, e = cost—isint (see Sec. 2.3) you can write a complex general solution ] e'(cos t= isin 1). rel i Jecessisnoeef ; ‘Writing this in terms of components and collecting cosine and sine terms, you obtain yr = cy (cos r+ isin) + c2e"(cos 1~i sin’) = e'(Acos 1+ Bsin 1), Azci ten, Jcqel(cos 1+ isin) + ieye'(Cos~ isin.) vn *(Ceos t+ Dsin 2) Ce-icj tic, D=eter ‘You see that C = ~B and D = A. You can now write « real general solution in vector form, namely, ETE) Initial value problem. The matrix ofthe given system is 14 10 A . 5 1 From A you obtain the characteristic equation -14-A2 10 et (A= aD aaNet sea (-14-a)( = 2) +50 = 0. Simplification gives a? + 132+ 36 = 0. Hence the eigenvalues are 2, = ~9 and Ay = ~4. Corresponding eigenvectors are obtained from [E14 - (9), + 10x2 = 0, thus ~5x, + 10x, = and (14-4) +102, = 0, thus ~ 10x; + 103, Hence the corresponding general solution is 2 len Mele yee +e, et 1 1 From this and the initial conditions y,(0) 1, written in vector form, you obtain 2 1 ite 0) =e ey tics) aeheceaiay 1 1 cyte “The solution is ¢, = -2,c = 3,30 that you obtain the particular solution 2 eee eel ee lee eee leat as scene 1 1 yaa De 43" ‘The figure shows that both y, (the lower curve) and y have a maximum and then approach zero in a monotone fashion.Chap. 3 ‘Systems of Differential Equations, Phase Plane, Qualitative Methods 2 os Section 3.3, Problem 15. Particular solutions y, (lower curve) and y Sec.3.4 Criteria for Critical Points. Stability Problem Set 3.4. Page 174 3. Saddle, The type ofa critical point is determined by quantities closely related to the eigenvalues of the matrix of a system, namely, the trace p, which isthe sum of the eigenvalues, the determinant g, which is the product of the eigenvalues, and the discriminant 4, which equals p? — 4g; see (9) in Sec. 3.8, In Prob. 3 the matrix is 12 Ae . [ oo | Hence p = 141 = 2,g=1-4 = -3, and A = 4~44(-3) = 16, Since q < 0, the system has a saddle point at 0, which is aiways unstable, as follows from (10) in Sec. 3.4, and is plausible from Fig. 80 in Sec. 3.3. To solve the system, you need the eigenvalues, which you obtain as solutions ofthe characteristic equation 1-2 2 det (A - Al) A en =H-22-3= (4-941) Hence the eigenvalues are -1 and 3. Their signs differ, which makes g negative and causes a saddle point. An eigenvector x" for -1 is obtained from (= CI)K +2 = 0, thus ean, say, el oe Hence x! = [1 - 1)". Similarly, for the eigenvalue 3 you obtain an eigenvector from (13m 424-0, thus a, sy RL eh This gives the eigenvector x = [1 1]¥. Hence a general solution is. 17. Perturbation of center. Example 4 in Sec. 3.3, to which the problem refers, shows two methods of solution, a systematic method and the shortcut. The first of them is similar to the procedure explained in this Manual in Prob. 5 of Problem Set 3.3 and can be completed following that method, The purpose of this Prob. 17 is to become aware of the fact that inaccuracies in the coefficients of a system (errors caused by rounding or in the process of physical measurements, etc.) can change the type of a critical point. In this problem itis suggested to go from A to B = A +0.11, but it will be obvious from the analysis that smaller deviations would have a similar effect. Given28. Ordinary Differential Equations, Part A o4 A= hence B=A+011= a y . 40 4 on For A you have p = 0+0.= 0,q = —1-(-4) = 4. This confirms thatthe system in Example 4 has center 28 it critical point; see (9) in Sec. 3.4. For B you have p = 0.1+0.1 = 0.2 #0,q= 0.0144 = 401, and A = p?~ 4q = 0.04 16,04 = ~16 < 0, which gives a spiral point by (9d) in Sec. 34. The eigenvalues of A are pore imaginary, 2i and ~2i (see Example 4 in Sec. 3.3), and itis interesting that the eigenvalues of B are 0.1 + 2i and 0.1 — 2, thas, they were changed by the same amount by which the main diagonal entries were changed (this reflects a general "shifting property"). Indeed, the characteristic equation of B is O1-az 1 det (B= a1) = okt y 4 Ola (01a? 44 = 2?7- 0224401 = 0. The roots (the eigenvalues of B) are 0.1 + 2i and 0.1 ~ 2 Sec. 3.5 Qualitative Methods for Nonlinear Systems Example 1. The critical point at (0, 0) turns out to be a center. This follows from the general criteria in Sec. 3.4, This is the first result. The next result follows from this and the periodicity of sin @ = sin yy with 2n, Namely, the points 42m, 44x, .. must also be centers, (Keep in mind that y, is just another notation for 8, introduced to fit the notation of our general discussions in this chapter.) The third result concerns the critical point (1,0) at 8 = x of the @-axis. The trick now is to move the origin to this point because our criteria were derived under the assumption that the critical point to be discussed is at the origin. This is the idea of the transformation (a translation) O-n=y, thus O=n+y, a“ You see that @ = now corresponds to our new yy = 0; we are at the new origin. Think about this before going on, From (A), sin x = 0, and cos = ~1 it follows that va sin = sin (x +y,) = sin wcos y, +08 4 siny, = -siny) = -y, +2 4. as indicated in the example. Problem Set 3.5. Page 183 §, Linearization begins with the determination of the positions of the critical points. As a system the given equation y" ~ y+y* = O becomes (see Sec. 3.1 for the general formula) n @ Wa yaI. ‘The critical points are obtained from y2 = O (then yj = 0). yy? = yy(1 ~ yi) = 0 (then y4 = 0). Hence they are at O and I on the yy-axis. Linearization is then done for each critical point separately. and in each ‘case the point is first shifted to the origin by a suitable change of coordinates, as explained in somewhat more detail just above (in connection with Example 1)- Accordingly, begin with the critical point (0,0). No transformation of y, or ya is necessary because the point already has the required position. From (a) you obtain the system linearized atthe origin simply by dropping the quadratic term, This linearized system is Its matrix is,Chap. 3 ‘Systems of Differential Equations, Phase Plane, Qualitative Methods 29 Calculate p = 0+0= 0,9 = -1- (You do not need p and A.) The second critical poin = 145.92 = Fae Then y\ = Fi, y= WC) = WH Since q < 0, this isa saddle by (9b) in aty; = 1,¥2 = 0. Hence make a shift by and (a) takes the form Heda Jae iH, Linearize this by dropping the nonlinear term (the last term in the second differential equation). This gives the linearized system whosemauixis Ap=| 9 | q -10 .q = 1, and conclude from (9c) in Sec. 3.4 thatthe critical point at (1,0) isa center. Calculate p = 13. Trajectories. yy" + y? = (yy')' = 0. By integration, yy’ = const or y,y2 = const. These are the familiar hhyperbolas with the coordinate axes as asymptotes. Sec. 3.6 Nonhomogeneous Linear Systems Example 1. The solution of the homogeneous system (not shown in the text) proceeds as before. That is, the characteristic equation of the matrix A is Q-A3-2)-C4)-1 = P4A-2= A- 42) 20, Hence the eigenvalues are 1 and 2. Bigenvectors are obtained for 2 = 1 from Q-Dr-45=0, sy, x84 =] and for 2. = ~2 from Q-(2))x)-4y = 0, say, el, om This gives the solution of the homogeneous equation shown in the answer on p. 185. Problem Set 3.6. Page 189 3. General solution, ¢ and ~3e” are such that you can apply the method of undetermined coefficients for determining a particular solution of the nonhomogeneous system. For this purpose you must first determine a general solution of the homogeneous system. The matrix of the latter is ela 10 Ithas the characteristic equation 2? ~ 1 = 0. Hence the eigenvalues of A are Ay = -I and Ay = 1 Eigenvectors x = x” and x are obtained from (A ~ AI)x = Owith A = 2) = -Land d=) = 1 respectively. For 4, = -1 you obtain atx 20, thus xem, say, pec Similarly, for 4g = 1 you obtain cate <0, this mex, say ml eh Hence eigenvectors are x = [1 = 1]Tandx® = [1 1]¥. This gives the general solution of the homogeneous system30 Ordinary Differential Equations Part A. Now determine a particular solution of the nonhomogeneous system. Using the notation in the text (Sec 3.6) you have on the right g = [1 ~3}’e* This suggests the choice yO =ue= [uy us)Te™ @) Here u is a constant vector to be determined. The Modification Rule is not needed because 3 is not an eigenvalue of A. Substitution of (a) into the given system y' = Ay +g yields on lf 1 (0) = 3ue"= Ay +g = Oe Ga 7 7 e(olie] [3] ‘Omitting the common factor e™, you obtain in terms of components Buy apt! ordered = 3uy- = I Sup = uy 3 muy +3uy = -3, Solution by elimination or by Cramer's rule (Sec. 6.6) gives uy = Oand up = ~1. Hence the answer is 17. Network, First derive the model. For the left loop of the electrical network you obtain from Kirchhoff's voltage law Lh +R -h) = E (a) because both currents flow through Ry, but in opposite dictions, so that you have to take ther difference. For the right loop you similarly obtain Ri) +Rah+ & [lade = 0. ) Insert the given numerical values in (a). Do the same in (b) and differentiate (b) in order to get rid of the integral. This gives h4Uh-h 2h, —) + 8h +2h = 0. Write the terms in the first ofthese two equations inthe usual order, obtaining Uy = -2h, + 2h, + 200. @l) ‘Do the same in the second equation as follows. Collecting terms and then dividing by 10, you first have 10/4 - 2F, +22 or f-0.21, +021, = 0. To obtain the usual form, you have to get rid of the term in J, which you replace by using (a1). This gives Ty = 0.2(-2f, + 2p + 200) + 0.212 = 0. Collecting terms and ordering them as usual, you obtain Ty = 0.41, +0.2l2 + 40. (b1) (at) and (b1) are the two equations ofthe sytem that you use in your futher work. The matrix of the corresponding homogeneous system is a [ 22 ] -04 02 Its characteristic equation is (Lis the unit matrix) det (A = AI) = (-2~2)(0.2~ a) ~(-Od).2 = 27 + 1.82404 =Chap. 3 ‘Systems of Differential Equations, Phase Plane, Qualitative Methods 31 This gives the eigenvalues A, = 09+ JOAT = ~0.259688 and da = -0.9- JOAT = -1,540312. Eigenvectors are obtained from (A Al)x = O with A = 2, and A= Aa. For, this gives C2- Ae 2m =O say. xp = 2 and xy = 242, Similarly, fry you obtain (2A), +2) ‘The eigenvectors thus obtained are LA Lom | ‘This gives as a general solution of the homogeneous system TO = 04 x0 el ey xO Ot, say, x = 2 and xp = 24a, and You finally need a particular solution 1©) of the given nonhomogeneous system J’ = AJ +g, where = [200 40)" isconstant, and J = [/, 43)" is the vector of the currents. The method of undetermined coefficients applies. Since g is constant, you can choose a constant 1? [u, ua] = const and substitute i into the the system, obtaining, since u’ = 0, pois, 200 Duy +2 +200 Autg= . . 04 02 || w 40 Ou, + 0.22 + 40 Hence you can determine w and up from the system = 2uy $+ 2uy = -200 = 0.44, +0.2u, = ~40. The solution is uw, = 100, uz = 0. The answer is 1. = 1410.