Sunteți pe pagina 1din 16

Senior Seminar II Project Analysis of the Trace Determinant Plane of Systems of Linear Dierential Equations

Joanna Sutton Fall 2012

Abstract This essay will analyze six dierent phase portraits of systems of linear dierential equations and how they are connected with the eigenvalues of that system. The dierent cases of eigenvalues for each systems and how these cases can be determined from the trace and determinant of the system will be demonstrated. Then, the link between the trace, determinant, and phase plane for the linear system will be analyzed.

Introduction

Trace-Determinant Planes are graphical representations of solutions of planar systems. These systems demonstrate dierent behaviors in the plane depending on the form of their eigenvalues. There are six dierent cases of eigenvalues that each result in dierent behavior. These cases are as follows: both eigenvalues are real and positive, both are real and negative, both are real but one is positive and one negative, the eigenvalues are complex with a real part of zero, the eigenvalues are complex with a positive real part, and the eigenvalues are complex with a negative real part. The eigenvalue cases can be linked to six dierent phase potraits. These portriats are nodal source, nodal sink, saddle point, center, spiral source, and spiral sink respectively. The trace and determinant of the system can be used to not only determine the case of the eigenvalues, but the case of the phase potrait as well.

Linear Systems

Trace-Determinant Planes are graphical representations of solutions of planar systems. A planar system is a linear system of dierential equations of dimension 2. Dierential Equations states that a system of linear dierential equations is any system of dierential equations that has the following form: x1 (t) = a11 (t)x1 (t) + + a1n (t)xn (t) + f1 (t) x2 (t) = a21 (t)x1 (t) + + a2n (t)xn (t) + f2 (t) xn (t) = an1 (t)x1 (t) + + ann (t)xn (t) + fn (t) in which x1 xn are unknown functions, and the coecients aij (t) and fi (t) are known functions of an independent variable,t, all dened for t (a, b), (a, b) is an interval in R. (Polking, Boggess, Arnold pg 425). In this paper we will be analyzing systems of the form y = Ay,where A= a11 a21 a12 a22 y1 (t) y2 (t)

and y(t) = (Polking, Boggess, Arnold pg 452)

This system can be solved through rst nding the characteristic polynomial for the matrix A. This is done through taking the determinant of A I , where I is the identity matrix. det(A I ) = det A = det a11 a21 a12 a22

= (a11 )(a22 ) a12 a21 = 2 (a11 + a22 ) + a11 a22 a12 a21 The quantity of (a11 + a22 ) is known as the trace of the matrix, abbreviated as T . The quantity a11 a22 a12 a21 is the determinant of the matrix A, abbreviated as D. This results in the equation 2 T + D (1)

To obtain the equation for the eigenvalues of A, we only need to set equation (1) to zero and solve. When applying the quadratic equation, we get T T 2 4D . (2) = 2

This equation shows that if T 2 4D < 0, then the eigenvalues are complex. If T 2 4D = 0, then the eigenvalues are repeated, and if T 2 4D > 0, then the eigenvalues are real and not repeated. Now that there is an equation for the eigenvalues, the trace-determinant plane can be drawn and analyzed. The trace-determinant plane is key in analyzing the equilibrium points. The trace is set as the horizontal axis, and the determinant is set as the vertical axis. The curve of T 2 4D = 0 is drawn on the graph separate the graph into two dierent regions (as shown below). The region above the curve represents systems with complex eigenvalues, and the region below the curve represents systems with real eigenvalues. (Polking, Boggess, Arnold pg 452).

Figure 1: Trace-Determinant Plane (Polking, Boggess, Arnold p477)

The Six Dierent Phase Portraits

Before discussing how to solve for various cases of eigenvalues, it is important to know what kinds of behavior the dierent combinations will show in the trace-determinant curve. If 1 < 0 < 2 (if there is

one positive and one negative) the origin will be a saddle. If 1 < 2 < 0 (both negative eigenvalues) then the origin will be a sink. If 0 < 1 < 2 (both positive eigenvalues) then the origin is a source. The following cases are for complex eigenvalues of the form = a bi. If a < 0 and b = 0, then the origin is a spiral sink. If a > 0 and b = 0, then the origin is a spiral source. Lastly, is a = 0 and b = 0, then the origin is the center. These behaviors are all shown on the table below.

Figure 2: Table of Phase Portraits in the Trace-Determinant Plane (Blanchard, Devaney,Hall p341) Now, to analyze the dierent equilibrium points, it is easier to rst look at the eigenvectors of the system to nd a solution. Theorem 1 Suppose A is a 2x2 matrix with real eigenvalues 1 = 2 . Suppose v1 and v2 are eigenvectors associated with these eigenvalues. Then, the general solution of the system y = Ay is y(t) = C1 e1 t v1 + C2 e2 t v2 (Polking,Boggess,Arnold p454) With this theorem and equation 3, we can now start to explain why each system has the its specic type of equilibrium point. (3)

3.1

Saddle Point

A system has a saddle point if it has one positive and one negative eigenvalue. In other words, 1 < 0 < 2 . To derive the saddle point,

we must rst let C2 = 0. This would cause y1 (t) = C1 e1 t v1 to be the exponential solution. Next, pick x to be any real number, and draw a solution for C1 = x and C1 = x. The line must go through the origin.When 1 < 0 the line decreases from innity to zero as t goes from negative innity to innity. Next, take C1 = 0. In this case we are now using y2 (t) = C2 e2 t v1 as our exponential solution. We solve for it in a similar manner to how we solved for y1 . We must draw the line for C2 = x, C2 = x where x is a real number. The line must go through the origin. (Polking,Boggess,Arnold p469) The saddle point is formed by taking the two half line solutions and adding them. Curved lines are then draw along the ends of the sum vectors. The solutions for y1 and y2 are asymptotes of each other. As t , t these solutions will never cross. These lines separate the plane into four dierent regions. The lines are called separatrices. The two solutions that approach the origin as t are called the stable solutions. The two solutions which approach the origin as t are called the unstable solutions. The equilibrium point is called a saddle point because the topographic map looks like a horses saddle. (Polking,Boggess,Arnold p469)

Figure 3: Saddle Point1

3.2

Nodal Sink

If a system has two negative eigenvalues, 1 < 2 < 0, then the plane displays a nodal sink. Again, we must begin by analyzing equation (3). Both of the values for are negative, so y(t) is innity large at negative innity. y(t) then converges to 0 on a half-line as t goes to innity. Now, we must split this into two dierent cases, one case where t and the other where t . In the rst case, we must rst rewrite equation (3) as y(t) = e2 t (C1 e(1 2 )t v1 + C2 v2 ) In this case, as t , e2 t 0. Since we know 1 < 2 < 0, we also know 1 2 < 0 which means(C1 e(1 2 )t ) 0 . This results in y(t) 0, but while converging the direction of y(t) gets closer to C2 v2 , so as y(t) 0 the solution curve on the plane will be tangent to the half-line we get from C2 v2 . (Polking,Boggess,Arnold p471) Now, for the second case we rewrite equation (3) as y(t) = e1 t (C1 v1 + C2 e(2 1 )t v2 ) As t ,e1 t since 1 < 2 < 0 implies 1 2 > 0. This also results in C2 e(2 1 )t 0. From this we know that (C1 v1 + C2 e(2 1 )t v2 ) C1 v1 and y(t) 0, but as y(t) 0, the direction of y(t) becomes that of C1 v1 .(Polking,Boggess,Arnold p471) All solution curves in a system with two negative eigenvalues tend to the origin as t . If all solutions for an equilibrium point of a system tend to the equilibrium point as t and the eigenvalues are real, then the equilibrium point is said to be a nodal sink. (Polking,Boggess,Arnold p471)

Figure 4: Nodal Sink2

3.3

Nodal Source

A nodal source is the last phase portrait a system can have if it has real eigenvalues. In this case, 0 < 1 < 2 . The properties for a nodal source are similar to that of a nodal sink but reversed. As stated above, in a nodal sink the solution curves tend to the origin as t , while in a nodal source the solution curves tend to the origin as t . (Polking,Boggess,Arnold p472) This is explained below. We must again split this into two dierent cases, one case where t and the other where t . In the rst case, we must rst rewrite equation (3) as y(t) = e2 t (C1 e(1 2 )t v1 + C2 v2 ) In this case, as t , e2 t 0. Since we know 0 < 1 < 2 , we also know 1 2 > 0 which means (C1 e(1 2 )t ) 0 . This results in y(t) 0, but while converging the direction of y(t) gets closer to C2 v2 , so as y(t) 0 the solution curve on the plane will be tangent to the half-line we get from C2 v2 .

Now, for the second case we rewrite equation (3) as y(t) = e1 t (C1 v1 + C2 e(2 1 )t v2 ) As t , e1 t since 0 < 1 < 2 implies 1 2 < 0. This also results in C2 e(2 1 )t 0. From this we know that (C1 v1 + C2 e(2 1 )t v2 ) C1 v1 and y(t) 0, but as y(t) 0, the direction of y(t) becomes that of C1 v1 .

Figure 5: Nodal Source3

3.4

Center

There are now three cases of phase portraits a system can have if the system has complex eigenvalues of the form = a + ib. For all of these cases, we rewrite equation 3 to be

y(t) = C1 eat (cos(bt)v1 sin(bt)v2 ) + C2 eat (sin(bt)v1 + cos(bt)v2 ) (4) (Polking,Boggess,Arnold p473) Where b = 0 The rst case for complex eigenvalues we let a = 0. This cases equation (4) to become

y(t) = C1 (cos(bt)v1 sin(bt)v2 ) + C2 (sin(bt)v1 + cos(bt)v2 ) (Polking,Boggess,Arnold p473) We already know that cos(bt) and sin(bt) are periodic functions that have a period of 2 |b| and a frequency of |b|. We also know that y(t) must have the same period and frequency. This means that the solution is a closed curve around the origin. When the equilibrium point of a system is surrounded by closed curve solutions it is referred to as a center. A linear system with purely imaginary eigenvalues has a center around the origin. These curves to not need to be circles though, they can be ellipses as well. (Polking,Boggess,Arnold p474)

Figure 6: Center4

3.5

Spiral Sink

We must now consider the cases where a = 0. In the rst case, we will suppose that a < 0. We rewrite equation (4) to be. y(t) = eat (C1 (cos(bt)v1 sin(bt)v2 ) + C2 (sin(bt)v1 + cos(bt)v2 )) (5)

(Polking,Boggess,Arnold p474). Here, y(t) is just the equation for a center multiplied by eat . We already know from above that this part of y(t) generates a phase portrait of ellipses centered around the origin. Since a < 0, we know that eat 0 as t . This means that as the solution curves are circling the origin, they are being drawn in to the origin as well. This results in a spiral phase portrait. (Polking,Boggess,Arnold p474) The solutions are spiraling around the origin while also approaching the origin. This is referred to as a spiral sink. A spiral sink is present anytime a system has a complex eigenvalue with a negative real part.

Figure 7: Spiral Sink5

3.6

Spiral Source

The last case we have is if a > 0. The general solution formula is again given by equation (5) above. This time however, since a > 0, eat as t . Thus, the solutions again circle the origin as with the spiral sink, but grow away from the origin as opposed to approaching it. This is called a spiral source. Any system that has a complex eigenvalue with a positive real part has a spiral source as a phase portrait.

10

We must remember that the trace-determinant plane is a parameter plane. This means that changing the parameter of the system changes the values of a11 , a12 , a21 , and a22 of the matrix A, and therefore changes the point (T, D). Small changes to the parameter dont make much of a dierence. A small parameter change will usually not change the form of the equilibrium point of the system. A spiral sink will stay a spiral sink, a center will stay a center, etc. (Blanchard,Devaney,Hall p345)

Figure 8: Spiral Source6

4
4.1

The Trace-Determinant Plane


Real Eigenvalues

If T 2 4D > 0 this signies that the eigenvalues are real, so when analyzing these systems we will always use T 2 4D > 0. There are six dierent combinations of real eigenvalues for a system. The eigenvalues can be both positive, one positive and one zero, both negative, one negative and one positive, one negative and one zero, and both zero. All of these cases can be solved for using equation 2.

11

First we will analyze the cases with a positive eigenvalue. If T > 0 in equation 2, then there must be one positive eigenvalue as the equation would be the sum of two positive numbers (Blanchard,Devaney,Hall p344), but we still need to solve for the other eigenvalue of the system. If D = 0, the equation becomes T T2 = =0 2 The other eigenvalue in the case is zero. Next consider the case of D > 0. In this case, T 2 4D < T 2 We are still only evaluating the system for when T > 0, so this equation can be solved further as follows. T 2 4D < T If this is combined with equation 2, we see that T T 2 4D >0 2 This shows that both eigenvalues will be positive. (Blanchard,Devaney,Hall p344) Now that we have shown what happens when D > 0 and D = 0, it is time to show what happens when D < 0. When D < 0, T 2 4D > T 2 Again, we are only evaluating the system for when T > 0, so this equation can be solved further as follows: T 2 4D > T T 2 4D <0 (6) 2 Since the system comes out as less than 0, there is one positive eigenvalue and one negative eigenvalue. (Blanchard,Devaney,Hall p344) T The above covers all the cases of positive eigenvalues, so we must now evaluate those with negative eigenvalues. This time, we will be analyzing the equation 2 with T < 0. If T < 0, we know that Equation (6) is the sum of two negative numbers, and therefore negative itself. Now, we only need to consider the same when T + T 2 4D 2 Similar as above, if D = 0, then there is one zero eigenvalue. If T < 0 and D = 0, the system has one negative and one zero eigenvalue.

12

If D < 0, T 2 4D > T 2 This can be simplied as follows : T 2 4D > T and T 2 4D > T . Since we know T 2 4D > 0, we only use the case T 2 4D > T . So, T + T 2 4D > 0 which results in T + T 2 4D >0 2 This results in one positive and one negative eigenvalue. We saw above that the system has a positive and negative eigenvalue if T > 0 and D < 0. This shows that a system having one positive and one negative eigenvalue is solely dependent on D. As long as D < 0 this will be the case. In the last case, take D > 0. This means that T 2 4D < T 2 So, T 2 4D < T and T 2 4D < T We know that T 2 4D must be greater than 0, so we only consider the case 0 < T 2 4D < T This results in T+ T 2 4D <0 2

This system would have two negative eigenvalues. The last thing we have not considered is if T = 0. If T = 0, D must also be zero to keep T 2 4D > 0, and both eigenvalues are zero.

4.2

Complex Eigenvalues

If T 2 4D < 0, then the eigenvalues are both complex numbers. Even further, both eigenvalues are complex conjugates of each other. Only the real part of the number has an impact on the equilibrium point when the eigenvalues are complex. All complex numbers are of the form a + bi. If we look again at equation 2, (Polking,Boggess,Arnold p476) T T 2 4D = . 2 The real part of the eigenvalue, a, is equal to T /2. If T < 0, then we know a must also be less than zero. This results in a complex

13

eigenvalue with a negative real part. If T > 0, then a > 0 as well, and this gives us a complex eigenvalue with a positive real part. Lastly, if T = 0, then the eigenvalue is purely imaginary. (Polking, Boggess, Arnold p477) All of the above cases of eigenvalues are summarized in the graph below

Figure 9: Eigenvalues on the Trace-Determinant Plane

Phase Portraits and Eigenvalues

If we look at both the table in Figure 2 and Figure 9, it is very clear to see how they are related. This relation between eigenvalues and phase potraits results in a relation between phase potraits, trace and determinant. Real eigenvalues are present when T 2 4D > 0. When T > 0 there must be at least one positive eigenvalue. We know that when T > 0 and D > 0 there must be two positive eigenvalues. We also know that two positive eigenvalues result in a nodal source. We can now relate the trace and determinant with this phase portrait. When the trace and determinant are both greater than zero, we know we will have a nodal source. Then same follows for the rest of the phase portraits. If T < 0 and D > 0, then both eigenvalues are negative which results in a nodal sink. When D < 0, then there is one positive eigenvalue and one negative eigenvalue and this results in a saddle point. This is independent of the value of T . Complex eigenvalues and phase portraits follow similarily. When T = 0, a = 0, since a = T /2 and the system would have purely imaginary eigenvalues.

14

Purely imaginary eigenvalues result in the system displaying a center. If T < 0, then then a < 0 as well and the systems phase portrait is a spiral sink. Similarly, if T > 0, a is greater than zero as well, and we have showed this results in a spiral source. This above relationship is shown through the graph below, where the complex region is shaded in blue and the real region is shaded in tan.

Figure 10: Phase Portraits and Eigenvalues on the Trace-Determinant Plane

Conclusion

The Trace-Determinant Plane is a graphical representation of solutions of planar systems. There are six important dierent types of behavior that a system can demonstrate in the trace-determinant plane. There are other types of behavior, but these are the most important. These six plane portraits are as follows: saddle, nodal sink, nodal source, center, spiral sink, and spiral source. The portrait a system has is decided by the form of the eigenvalues for that system. If a system has two positive, real eigenvalues its portrait is a nodal source, two real, negative results is a nodal source, both real with one positive and one negative results in a saddle, complex eigenvalues with a = 0 results in a center, complex with a < 0 results in a spiral sink, and nally

15

complex with a > 0 results in a spiral source.

16

S-ar putea să vă placă și