Vous êtes sur la page 1sur 42

Chapter 2 nth-Order Linear Eqns

Much of the theory of rst-order linear equations can be extended in a very nice way.

2.1 Second-Order Equations


Denition. A second-order DiEq is called linear if it can be written in the form y + a1(x)y + a0(x)y = b(x) (18)

Such an equation is called homogeneous if b(x) = 0. For any equation like (18), the equation y + a1(x)y + a0(x)y = 0 (ie b replaced by 0) is called the associated homogeneous equation.

Observation 1. Note that the term homogeneous here means something completely dierent from the homogeneous equations studied in 1.6. 2. If y has a coecient a2(x), we can divide the whole equation by it and thus bring it into the form (18). Then we have to watch for points x where a2(x) = 0, because they could create discontinuities of the new coecients ai(x)/a2(x). A typical system governed by such an equation is a mass-spring-dashpot assembly. Theorem 1 (Superposition / homog) If y1, y2 are two solutions of a homogeneous linear 2nd-order equation on some interval I, then c1y1 + c2y2 is another solution of that same equation on the same interval. Example 1 It is easily checked that cos x, sin x are both solutions of y + y = 0. Therefore,

4 cos x + 5 sin x is another solution of y + y = 0. Theorem 2 (Exist/Unique for IVPs) An Initial Value Problem (IVP) consisting of an 2nd order linear DiEq with coecient functions continuous in some interval I, together with 2 initial conditions y(a) = y0, y (a) = y1 for some a in I has exactly one solution for any choice of the values y0, y1. This solution is dened in all of I. Denition. A pair of functions y1, y2 is said to be linearly independent on an interval I if none of the functions is a constant multiple of the other on that interval. If one is a multiple of the other, they are called linearly dependent. Example 0 and any other function f (x) are linearly dependent since 0 = 0 f (x). Eg

e2x, e3x are linearly independent. The interval I does not play a role. But eg the functions 2x, |x| are linearly independent on [1, 1], linearly dependent on [0, 1]. Denition. For two given dierentiable functions y1, y2, dene the Wronskian as a determinant, W (y1, . . . , yk ) := y1 y2 y1 y2 = y1y2 y2y1 (19)

Theorem 3 (Wronskian) If y1, y2 are solutions of a homogeneous 2nd-order DiEq like (18) with b(x) = 0 on some open interval I, then EITHER

y1, y2 are linearly independent, and W = 0 for all x in I

OR

y1, y2 are linearly dependent, and W = 0 for all x in I. Theorem 4 (Gal Solution, hom.) If y1, y2 are two linearly independent solutions of a homogeneous 2nd-order linear DiEq, then c1y1 + c2y2 is the general solution of that DiEq. Denition. For the a linear 2nd-order homogeneous DiEq with constant coecients a2y + a1y + a0 = 0 (20)

dene the characteristic polynomial as p(r) = a2r2 + a1r + a0 and the characteristic equation as a2r2 + a1r + a0 = 0 (21)

Theorem 5 (2 distinct real roots) If the characteristic equation has real roots r1 = r2 then Equation (20) has the linearly independent solutions y(x) = er1x, er2x and the general solution of (20) is y(x) = c1er1x + c2er2x. Theorem 6 (Double root) If the characteristic equation has only one root r1 then Equation (20) has the linearly independent solutions y(x) = er1x, xer1x and the general solution of (20) is y(x) = (c1 + xc2)er1x.

2.2 Gal Soln of nth-Order Equations


An n-th order linear dierential equation is one which can be rewritten in the form y (n) + an1y (n1) + + a1y + a0y = b (22) where all an1, . . . , a0, and b can be functions of x. Theorem 1 (Superposition/homog.) Suppose we are given functions y1, . . . , yk of a homogeneous linear n-th order DiEq like (22) with b = 0. Then for any choice of constants c1, . . . , ck , the linear combination c1y1(x) + + ck yk (x) is another solution of the same DiEq. Theorem 2 (Exist/Unique for IVPs) An Initial Value Problem (IVP) consisting of an n-th

order linear DiEq with coecient functions continuous in some interval I, together with n initial conditions y(a) = y0, y (a) = y1, . . . , y (n1)(a) = yn1 for some a in I has exactly one solution for any choice of the values y0, . . . , yn1. This solution is dened in all of I. Denition. A list of functions y1, . . . , yk is said to be linearly independent if the equality c1y1(x) + + ck yk (x) = 0 (valid for every x) can only be true for c1 = = ck = 0. Example Two functions are linearly independent in the earlier sense (not multiples of each other) exactly if they are linearly independent in the new sense. 5, 4x, 6x 15 are linearly dependent because

(6) 5 + 3 4x 2 (6x 15) = 0. 1, x, x2 are linearly independent. Denition. A matrix is a rectangular array of numbers. If A is a square matrix (size n n, say), then the determinant of A is dened recursively by det(A) = |A| = (1)n+1an1 det(An1) +(1)n+2an2 det(An2) + . . . +ann det(Ann) (23) Here, aij is the entry of A in row i, column j, and Aij is the square matrix obtained from A by deleting row i and column j. The powers of 1 produce an alternating sign pattern, ending with a + in the last column. Calculating these determinants is so tedious that we try to avoid them for matrices bigger than 3 3. Note - all major software packages have det as built-in function, even Excel.

Denition. For k given functions y1, . . . , yk , suciently often dierentiable, dene the Wronskian as a determinant, y1 y1 W (y1, . . . , yk ) := . . . y1
(k1)

. . . yk yk . . .
(k1)

(24)

. . . yk The recursive formula for the determinant is particularly nice for Wronskians: W (y1, . . . , yk ) = (1)k1y1 +(1)k y2 + yk
(k1)

W (y2, . . . , yk )(25)

(k1)

W (y1, y3, . . . , yk )

(k1)

W (y1, y2, . . . , yk1)

Theorem 3 (Wronskian) If y1, . . . , yn are solutions of a homogeneous n-th order equation like (22) with b = 0 on some open interval I, then EITHER

y1, . . . yn are linearly independent, and W = 0 for all x in I

OR

y1, . . . , yn are linearly dependent, and W = 0 for all x in I.

Recap how to calculate a Wronskian for three functions f, g, h. W (f, g, h) = f g h g g h f h +h f h f g f g (26)

Let us conrm our earlier calculations and compare the answers with the Wronskian technique.

Example If g = kf (a constant multiple) then W (f, g) = 0 linearly dependent. 5, 4x, 6x 15 has Wronskian 5 4x 6x 15 W (5, 4x, 6x 15) := 0 4 6 =0 0 0 0 1, x, x2 has Wronskian 1 x x2 W (1, x, x2) := 0 1 2x = 2 0 0 2

2.3 Homogeneous, Constant Coecients


Example How do we make up a homogeneous, 3rd-order linear equation with constant coecients which has for solutions y = 3e5x, e2x, e10x? We use again the characteristic equation r3 + a2r2 + a1r + a0 = 0.

If r is any root of that equation then erx solves y (3) + a2y + a1y + a0y = 0. So if we want e5x to be a solution, we better make sure that r + 5 is a factor of the characteristic polynomial r3 + a2r2 + a1r + a0. It should be of the form (r + 5)(r2 + ...). We also want factors r 2 and r 10, so the characteristic polynomial has to be (a multiple of) (r + 5)(r 2)(r 10) = r3 7r2 40r + 100. This corresponds to the equation y (3) 7y 40y + 100y = 0. Whenever a linear equation has a solution erx, then constant multiples of that are also a solution. This covers 3e5x and e2x. Denition. For the linear n-th order homogeneous equation any (n) +an1y (n1) + +a1y +a0y = 0, (27)

dene the characteristic equation anrn + an1rn1 + + a1r + a0 = 0 (the left-hand side of the characteristic equation is called the characteristic polynomial). Theorem 4 (n distinct real roots) If the characteristic equation has n distinct real roots r1, . . . , rn then Equation (27) has the n linearly independent solutions y(x) = er1x, . . . , ernx and the general solution of (27) is y(x) = c1er1x + + cnernx. In case there are less than n distinct real roots you still get some solutions, just not enough to nd the general solution. Example Equations with repeated roots are (r 5)2 (r 5)2(r2 + 1)

These characteristic equations belong to differential equations y 10y + 25y = 0 (28)

y (4) 10y (3) + 26y 10y + 25y = 0 as seen by expanding the characteristic polynomials above. Denition. A number r0 is said to be a root of the characteristic polynomial with multiplicity k if (r r0)k is a factor of that polynomial. Theorem 5 (Repeated real roots) If r0 is a root of the characteristic polynomial with multiplicity k then y(x) = er0x, xer0x, . . . , xk1er0x are k linearly independent solutions of the linear homogeneous equation with that characteristic polynomial.

D-operator language
This is a useful device for verifying things like Theorem 5. The letter D stands for dierentiation, seen as a linear operator. So D2 stands for dierentiate twice, D3 for ... three times and so on. And D + 5 means dierentiate and then add 5 times the original function. Similarly, we may write (D2 10D + 25)y = y 10y + 25y. Generally, you can rewrite the dierential equation having characteristic polynomial P as P (D)y = 0. Verifying Theorem 5 rests on (D r0)(xmer0x) = mxm1er0x, (29)

so D r lowers the degree of the polynomial which multiplies erx. Apply (D r)k and any

polynomial of degree less than k will be reduced to zero. Suppose the characteristic polynomial is P (r) = (r r0)k Q(r). We can apply the operator P (D) to y(x) = xmerx. P (D)y = Q(D)(D r0)k xmer0x which is zero for all m < k. That means xmerx is a solution for P (D)y = 0. Example The rst equation in (28) has two solutions e5x, xe5x. So it has the general solution y(x) = c1e5x + c2xe5x. These functions are also solutions of the other equation in (28), but not the general solution (we need to nd 2 more solutions so that we have 4 linearly independent solutions).

Complex Numbers
They are dened as a+bi where a, b are any real number. They can be added and multiplied like polynomials in i with the extra rule i2 = 1. The complex conjugate of z = a + bi is dened as z = a bi. Complex conjugation has the following nice properties. r+s = r+ s rs = r s r = r i r real And nally, we need the Euler/DeMoivre Formula exp(bi) = cos b + i sin b. Example (3 + 5i) + (6 0.4i) = 9 + 4.6i (3 + 5i)(6 0.4i) = 20 + 28.8i (6 0.4i)(3 + 5i) 6 0.4i = 3 5i 32 + 5 2 20 28.8 = + i 34 34

In the last line, notice that z = a2 + b2. z

Complex Roots
Suppose r = a+bi is a root of the characteristic polynomial P of some linear equation. If r is not real, then b = 0 and the complex conjugate r = a bi
is another root of P . So erx and erx are both solutions of the given equation.

Theorem 6 (Complex Roots) If r = a+bi is a root of the characteristic polynomial of some linear equation, then y1(x) = exp(ax) cos bx y2(x) = exp(ax) sin bx are two linearly independent, real-valued solutions of that equation.

Proof Take the linear combinations y1(x) = (exp((a + bi)x) + exp((a bi)x))/2, y2(x) = (exp((a + bi)x) exp((a bi)x))/(2i) and use the Euler/DeMoivre Formula. Example 4 a) The equation y 4y + 5y = 0 has characteristic polynomial P = r2 4r + 5 with roots r = 2 i. So it has solutions y1(x) = e2x cos x y2(x) = e2x sin x b) The equation y (4) 10y (3) + 26y 10y + 25y = 0 has characteristic polynomial P = (r 5)2(r2 + 1)

with roots r = 5, 5, i (meaning 5 is a double root). So it has solutions y1(x) = e5x y2(x) = xe5x y3(x) = cos x y4(x) = sin x and the general solution is y(x) = c1y1(x) + c2y2(x) + c3y3(x) + c4y4(x). because here, r = i

Example 6 The equation y (4) +18y +81y = 0 has characteristic polynomial r4 + 18r2 + 81 = (r2 + 9)2 with roots 3i, 3i, 3i, 3i. As with real roots, four linearly independent solutions are sin 3x, x sin 3x, cos 3x, x cos 3x.

2.4 Mechanical Vibrations Mass-Spring-Dashpot models


are just examples for behavior which happens in more complex systems - examples which we can solve. Total force acting on cart is F = FS + FR + FE where FS = kx due to spring, FR = cv due to dashpot/friction, and FE is an external force. All of these change with time. So mx = ma = F = kx cx + FE , c k 1 x + x + x = FE . (30) m m m If c = 0 we call the motion undamped, otherwise damped. If FE = 0 we call the motion free, otherwise forced.

Pendulum
Pendulum = mass suspended at the end of a string, or thin rod

Write L for the length of the pendulum, for the angle it makes with the vertical at time t. Then the height of the mass is h = L(1 cos ) and its potential energy is V = mgh, so V = mgL(1 cos ). The distance traveled at time t is s = L, so the kinetic energy is 1 ds 2 1 d 2 1 = mL2 T = mv 2 = m 2 2 dt 2 dt The law of conservation of energy says T +V = C is constant, so
2 1 2 d mL + mgL(1 cos ) = C. 2 dt

This is an implicit DiEq - cant solve it.

Dierentiate to get mL2 + mgL sin = 0. (31)

Cancel out mL to get a second - order equation for . Approximate sin by (good for small ) to get a linear second-order equation with constant coecients. Add a term c to account for (small) friction and we get an equation like (30), g + c + = 0. (32) L Eg for c = 0, we get a general solution (t) = c1 cos(0t) + c2 sin(0t) where 0 = g/L is the circular frequency of the system, measured in Radians per Time. This is a case of Free Undamped Motion, see below.

Free Undamped Motion


Here, c = 0 and FE = 0 in (30), so mx + kx = 0. Put 0 = k/m so roots of P are i0, and the general solution is x(t) = A cos 0t + B sin 0t. Where does x have max/min?? Put C = Then nd such that A cos = C B sin = . C A2 + B 2 .

Up to a multiple of , = tan1(B/A). Then we get the nicer expression for x, x(t) = C cos(0t ). So C is the maximum of |x|, called the amplitude. is called phase angle, and 0 the

circular frequency. The period of this motion is the time it takes until it repeats. It is given by 2 T = 0 the frequency is = 1/T = 0/(2). This gives the number of complete cycles per unit of time. Example 1 A body with mass 1/2 kg is attached to a spring in initial position x0 = 1, velocity v0 = 5. The spring can be stretched 2 meters by a force of 100 N . Find x(t), C, T, , and .

Solution 1 x(t) = cos 10t sin 10t 2 1 C = 5 2 1 tan = 2 = 2 tan1(1/2) 5.8 1 x(t) = 5 cos(10t 5.8) 2

Free Damped Motion

The character-

istic poly. of mx + cx + kx = 0 is mr2 + cr + k which has roots c2 4km 2 r= = p p2 0 (33) 2m where p = c/(2m). There are three cases depending on the critical damping c0 = 4km. c

1. Overdamped motion, c > c0. General solution has two real exponentials with negative exponents, both solutions tend to zero.

2. Critically damped, c = c0. A repeated root means the general solution is x(t) = ect/(2m)(A + Bt) Not that much dierent. once unless B = 0. Crosses t-axis

3. Underdamped, c < c0. Imaginary roots, solutions have exponential with negative exponent. Oscillates about t-axis while tending to zero, x(t) = ect/(2m)(A cos(1t) + B sin(1t)). Here, the number 1 = 4km c2 = 2m
2 0 p2

is called circular pseudofrequency, 2/1 the pseudoperiod, and Cept the time-varying amplitude.

Example 2 As Example 1, but now the mass is also hooked up to a dashpot which provides a resistance of 1 Newton per unit of velocity. Again, x(0) = 1 and v(0) = 5. Find ... everything! Solution Since c = 1, the equation is now 1 x + x + 50x = 0. 2 The characteristic equation has roots r = 1 i 99. So t (A cos 99t + B sin 99t). x(t) = e

Initial conditions x(0) = 1, x (0) = 5 give A = 1 C = et 4 B = 99 A2 + B 2 = et 115 99

x(t) = C cos( 99t 1) 1 = 2 tan1(B/A) 5.90 1 = 1/1 0.59


1.5

0.5

0.5

1.5

0.5

1.5

2.5

Solutions of 0.5x + cx + 50x = 0 with x(0) = 1, x (0) = 5 for c = 0, . . . , 1. The frequencies are almost the same.

1.5

0.5

0.5

1.5

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

Solutions of 0.5x + cx + 50x = 0 with x(0) = 1, x (0) = 5 for c = 0 (black), c = 0.9, 1.9, . . . , 9.9 (blue) and the critical damping c = 10 (red). The frequencies are visibly decreasing.

2.5 Non-Hom Linear Equations


We are still assuming constant coecients. Example 1 Find a particular solution of y + 3y + 4y = 4x 5. Solution Try yp = Ax + B, you nd yp = x 2. Make sure to tie this in with the general theory the general solution is y = yp + yc, where yc is the general solution of the associated homogeneous problem y + 3y + 4y = 0. Example 2 Find a particular solution of y 4y = 2e3x Solution Try yp = Ae3x, you should nd A = 2/5.

Example 3 Find a particular solution of y 2y = 2 cos x Solution Try yp = A cos x + B sin x. Substituting yp into the given DiEq leads to the system A 2B = 0 B 2A = 2 which gives B = 2/5 and A = 4/5. Suppose f is a linear combination of polynomials times exponentials eax and/or sines and/or cosines sin bx, cos bx. Write a list of all terms that appear in f, f , f , . . . - this is a nite list. Let yp be a linear combination of these terms with undetermined coecients. Substitute into the given equation Ly = f to determine these coecients.

Rule 1 - no duplication

Rule 1 is guaranteed to work unless some term in the list also occurs in the complementary function, that is the solution of Ly = 0. See example 4. Note that I wrote eax and cos bx, sin bx instead of the textbooks erx, cos kx, sin kx. This is so you can better see the connection to section 2.3. No term from the list can occur in the complementary function, unless a + bi is a root of the characteristic polynomial P dened in 2.3.

The Case of Duplication


Now what happens if some term of the list above does occur in the complementary function (ie r = a + bi is a root of P )? Example 4 Find a particular solution of y 4y = 3e2x

Solution If you try yp = Ae2x you nd Lyp = 0 regardless of A because 2 is a root of the characteristic polynomial r2 4. Similar to repeated roots situation - try y = Axe2x. You will nd that A = 3/4 works. Example - massive, intimidating! particular solution of (D r0)3y = (2x 3)er0x Solution The characteristic polynomial P (r) = (r r0)3 has a root r0 with multiplicity 3 so the complementary function is y(x) = (A + Bx + Cx2)er0x and of course we have duplication. Notice that any solution of this equation also satises the homogeneous linear equation (D r0)5y = (D r0)2[(2x 3)er0x] = 0 Find a

because the operator D r operates on a polynomial times exponential by just dierentiating the polynomial. So y can be written as y(x) = (A + Bx + Cx2 + Dx3 + Ex4)er0x. We may even choose A = B = C = 0 because these terms will just give zero when we substitute them into the given DiEq. With the remaining constants, (D r)3y = (6D + 24Ex)er0x = (2x 3)er0x is what we want so D = 1/2, E = 1/12 will indeed give a particular solution.

Rule 2 - duplication

Suppose f is a poly-

nomial of degree m times an exponential eax times cos bx or sin bx. Then try yp(x) = xsQmeax cos bx + xsRmeax sin bx with undetermined polynomials Qm, Rm of the same degree m as before. Choose s equal to

the multiplicity of the root a + bi in the characteristic polynomial of the dierential equation Ly = 0. Then substitute yp into the given Di.Eq Ly = f to determine the coecients of Rm, Qm. Alternatively, s is the smallest integer so that there is no overlap between terms in yp and terms in the complementary function. If f has several terms with duplication then Rule 2 has to be applied to each one, and s should be chosen separately for each of them. Afterwards, the particular solutions for each term in f need to be added up to give Ly = f . On the other hand, it may be possible to save time and variables by combining multiples of the same exponential and/or trig function into one term (see eg problem # 17).

Example 9 Determine the appropriate form of a particular solution of y + 6y + 13y = (1 + x)e3x sin(2x)

Solution The characteristic polynomial is r2 + 6r + 13 with roots r = 3 2i, so the complementary function is yc(x) = c1e3x cos(2x) + c2e3x sin 2x. Clearly, we have duplication, so Rule 2 applies. The roots are simple so s = 1. The degree m = 1, so the appropriate form is

yp(x) = x(Ax+B)e3x cos(2x)+x(Cx+D)e3x sin(2x). Note - nobody asked for the actual values of A, B, C, D, so stop right here!!

Variation of Parameters

Cut from exam.

2.6 Forced Oscillations


We study the special case of mx + cx + kx = F (t) where F is a simple harmonic mx + cx + kx = F0 cos t (34)

(or the same with a sine function instead of a cosine). Example A ywheel of mass m0 rotates on a cart of mass m m0. Suppose the center of the ywheel is directly above the centroid (center of mass) of the cart, both displaced by x from the equilibrium position. If is the frequency of the ywheel and a the distance of its center of mass from its center, then the center of mass of the ywheel has x-coordinate x + a cos t. The combined centroid x of cart and ywheel has x-coordinate x = (m m0)x + m0(x + a cos t) m m0 a = x+ cos t. m

Without friction, Newtons Second Law says m = kx and, writing everything in terms of x x, mx + kx = m0a 2 cos t (35)

(could also imagine the ywheel exerting a force of magnitude m0a 2 on the cart). To study Equation (35), set 0 = k/m, so

xc(t) = c1 cos 0t + c2 sin 0t Let F0 = m0a 2. Assume 0 = . To nd a particular solution xp, try xp = A cos t + B sin t Right away, we can see B = 0, and then F0 A= 2 m(0 2) (36)

As in 2.4, we could rewrite xc as multiple of a shifted cosine if we wanted to.

Example 1 Suppose m = 1, k = 9, F0 = 80, = 5, so we solve x + 9x = 80 cos 5t. Find x(t) if x(0) = x (0) = 0.

Beats
With initial conditions x(0) = x (0) = 0, we get c2 = 0 and c1 = A = Use the trig identity cos(u v) cos(u + v) = 2 sin u sin v to rewrite x: x(t) = A(cos t cos 0t) ( )t ( + )t = A sin 0 sin 0 2 2 Graph for close to 0. F0 2 m(0 2)

Resonance
This happens in the case of duplication (see Rule 2!), where = 0. No huge forces needed, just constant reinforcement of existing oscillations.

Damped Forced Oscillations


Interesting feature: the complementary function xc approaches zero for t for any initial conditions! therefore the xc-part is called a transient solution, xp a steady periodic oscillation. Example 6 Consider a mass-spring-dashpot system with m = 1, c = 2, k = 26, and external force F (t) = 82 cos 4t. Find the position x(t) if x(0) = 6, x (0) = 0 and compare to the steady periodic oscillation.

Vous aimerez peut-être aussi