Second order linear differential equations and systems of lin ear

advertisement

Text 6

Johann K. Brunner

Mathematics, October 2010

Second-order linear di¤erential equations and systems of linear di¤erential equations

1. A second-order linear di¤erential equation with constant coe¢ cients (for _ and x ) has the form

• + a x + bx = c; (1) with a; b; c 2 R ; b = 0 :

1

It is useful to consider the corresponding homogeneous di¤erential equation (where we leave out c ) :

• + a x + bx = 0 ; (2)

Its solution depends on the roots of the characteristic equation r

2

+ ar + b = 0 : (3)

Theorem 6.3.1 of Sydsaeter et. al. (2008) shows us the general solutions of (2) for three possible cases, (I) two di¤erent real roots of (3), (II) one double real root, (III) two complex roots. Note that a general solution has two integration constants (denoted by A and B ).

Returing to the nonhomogeneous di¤erential equation (1), one observes easily that the constant function u ( t ) = c=b solves (1). (Use u instead of x in (1) and note that

_ = 0 ; u = 0 ; so b c=b = c: ) u = c=b is called a particular solution of the nonhomogeneous di¤erential equation (1).)

As a consequence of Theorem 6.2.1 of Sydsaeter et. al. 2008, it is generally true that any solution of the nonhomogeneous di¤erential equation (1) can be written as the sum of a solution of the corresponding homogeneous di¤ erential equation (2) (given by Theorem

1 Note that in case of b = 0 we can substitute u = _ _ = • in (1) and we arrive at a …rst-order linear di¤erential equation for u. After having solved it, we …nd x by integrating u.

1

6.3.1) plus a particular solution of the nonhomogeneous di¤ erential equation (1), which is c=b . To …nd precise values for the constants A and B, two conditions, say an initial condition x ( t

0

) = x

0 and a further condition x ( t

1

) = x

1

; must be given.

Clearly, the particular solution u ( t ) = c=b is an equilibrium of the di¤erential equation

(1): if the solution assumes the value c=b from the beginning (that is, the side conditions state x

0

( t

0

) = x

1

( t

1

) = c=b ), it remains there and the di¤erential equation is always ful…lled. We can again ask what happens if we start with a value = c=b (that is, the side conditions specify a path out of equilibrium): if t ! 1 ; will the solution converge to c=b or not? If yes, we call (the equilibrum of) the di¤erential equation globally asymptotically stable . The answer depends on the behaviour of the general solution of the associated homogeneous di¤erential equation (2), whether it goes to zero for t ! 1 (then any solution converges to the particular solution c=b ).

Considering case I of Theorem 6.3.1, it is clear that the function x converges to zero if both roots of the characteristic equation are negative. In case II, the double real root must be negative. In case III, with complex roots i ; its real part must be negative.

We conclude generally that for asymptotic stability the real parts of all eigenvalues must be negative (in case of a real eigenvalue the real part is the eigenvalue).

2. A system of two linear di¤erential equations with constant coe¢ cients is written as:

_ = a

11 x + a

12 y + b

1

; y _ = a

21 x + a

22 y + b

2

;

(4)

(5) or in vector-form, with A as the matrix of the coe¢ cients a ij

: y _

= A x y

+ b

1 b

2

: (6)

An equilibrium of (4), (5) is found by setting _ = 0 ; y _ = 0 ; which means (see (6)) to solve

A x y

= b

1 b

2

: (7)

By Cramer’s rule (chapter 1 of Sydsaeter et. al. 2008), the solution is x = ( a

22 b

1

+

2

a

12 b

2

) = j A j ; y = ( a

11 b

2

+ a

21 b

1

) = j A j :

To see the relation of the system (4), (5) to a second-order linear di¤erential equation, we use (4) to get

1 y = a

12 a

11 a

12 x b

1 a

12

: (8)

Di¤erentiating both sides with respect to t gives

1 y _ = a

12 a

11 a

12

(9)

Using (8) and (9) in (5) gives us

1 a

12 a

11 a

12

_ = a

21 x + a

22 a

12 a

22 a

11 x a

12 a

22 b

1 a

12

+ b

2 or

• ( a

11

+ a

22 x + ( a

11 a

22 a

21 a

12

) x = a

22 b

1

+ a

12 b

2

: (10)

So we are back to a second-order linear di¤erential equation (1) with constant coe¢ cients a = ( a

11

+ a

22

) = tr ( A ) ; b = a

11 a

22 a

21 a

12

= j A j ; c = a

22 b

1

+ a

12 b

2

:

Observe …rst that its equilibrium c=b = ( a

22 b

1

+ a

12 b

2

) = j A j is indeed the same as we found above for x as the equilibrium of the system of two equations (4) and (5).

Next, we have to consider the corresponding homogeneous equation

• ( a

11

+ a

22

) x + ( a

11 a

22 a

12 a

21

) = 0 with characteristic equation r

2

( a

11

+ a

22

) r + ( a

11 a

22 a

12 a

21

) = 0 : (11)

Inspection shows immediately that this is just the characteristic equation of the matrix A

(see chapter 1 of Sydsaeter et.al. 2008). Hence the roots of (11) are the eigenvalues of A .

We know from above that (10) is globally asymptotically stable, if the roots of (11), i.e. the eigenvalues of A are negative (or, if complex numbers, have negative real parts).

3

One can show that this is equivalent to the two properties: tr ( A ) < 0 ; j A j > 0 :

2

Note …nally, that an alternative procedure is to use (5) to …nd an expression for x , di¤erentiate and substitute in (4) to get the second-order di¤erential equation in y :

• ( a

11

+ a

22

) _ + ( a

11 a

22 a

12 a

21

) y = a

11 b

2

+ a

21 b

1

; whose equilibrium ( a

11 b

2

+ a

21 b

1

) = j A j is the same as found for y from (7) above.

Moreover, by comparison one observes that the characteristic equation of this secondorder di¤erential equation is the same as (11) above. Thus, the conditions for stability tr ( A ) < 0 ; j A j > 0 - are the same.

2

Remember that tr ( A ) is the sum of the two eigenvalues, while j A j is their product.

4

Download