MAT 239 (Dierential Equations, Prof. Swift) 3.7 and 3.

8: Oscillators and Complex Numbers

Topics: Oscillators without external force. The R, description of oscillator solutions. The polar form of complex numbers. Driven oscillators. Oscillators: For most of you, the reason you are required to take this class is so that you can solve the DEs describing oscillators. Suppose that m is the mass of an oscillator, is the damping constant, and k is the spring constant. If there is no external force, then the DE for the position of the mass, u(t), is 2 mu + u + ku = 0, or u + u + 0 u = 0, m where 0 = k/m is the natural frequency of the oscillator. It turns out that 0 is the frequency of the undamped oscillations when = 0, but remember that 0 is dened even if the oscillator is damped (i.e. > 0). Recall that the weight is not the mass. The weight is W = mg . In English units, g = 32ft/sec2 , so the mass m = W/g has units [lb sec2 /ft]. Furthermore, has units [lb sec/ft], and k has units [lb/ft]. Things are more straightforward in metric units. Assuming that m and k are positive, and 0, the roots of the characteristic equation, and the resulting motion of the oscillators, fall into four classes: = 0, Undamped: roots i0 . 2 0 < < 2m0 , Underdamped: roots 2 0 ( 2 i , where = )2 < 0 m m = 0 , with multiplicity 2 = 2m0 , Critically Damped: roots 2 m 2 > 2m0 , Overdamped: two distinct real, negative roots 2 ( 2 )2 0 . m m One goal of this handout is to get a better understanding of the motion in the undamped and underdamped cases. The R, description of oscillator solutions. The general solution of the undamped oscillator can be written in two equivalent ways: u(t) = C1 cos(0 t) + C2 sin(0 t) = R cos(0 t ), where the constants C1 , C2 , R and satisfy the equations C1 = R cos , C2 = R sin . (2) (1)

The constants C1 and C2 are usually better for initial value problems, but the second form of the solution in equation (1) is easier to understand. The amplitude of the oscillation is R, and gives a phase shift. Every solution is periodic with period T = 2/0 . In the underdamped case, the general solution is u(t) = e 2m t (C1 cos t + C2 sin t) = R e 2m t cos(t ).

The envelope of the decaying oscillation is u(t) = R e 2m t . (Recall that 1 cos(t ) 1.) We call the quasi frequency, and Td = 2/ is the quasi period, since the motion sort of repeats after time Td . The Polar Form of Complex Numbers: A complex number z = C1 + i C2 can be written in polar form z = Rei , where R is the radius and is the phase of z . Since C1 + i C2 = R ei = R(cos + i sin ), the relationship between the rectangular form and the polar form is given by equation (2): C1 = R cos and C2 = R sin . I will now prove equations (1) and (2). Recall from 3.4 that when the roots of the characteristic equation are r = i 0 the general solution to can be written as a linear combination of the real and imaginary parts of the complex solution u(t) = ei0 t : ( ) ( ) u(t) = C1 Re ei0 t + C2 Im ei0 t = C1 cos 0 t + C2 sin 0 t = Re [(C1 i C2 )(cos 0 t + i sin 0 t)] Note the minus sign! ( ) = Re (C1 i C2 )ei0 t Now, if we choose R and so that C1 + i C2 = R ei then C1 i C2 = R ei , and continuing from the last equation we get ( ) u(t) = Re R ei ei0 t ( ) = R Re ei(0 t) = R cos(0 t ) This completes the proof of equations (1) and (2). Equations (2) and (3) say the same thing. They relate rectangular coordinates (C1 , C2 ) to polar coordinates (R, ). Given R and , it is easy to nd C1 and C2 . The reverse procedure is not so easy. If we know C1 and C2 , and want R and , we have the formulas C2 2 2 R = C1 + C2 , tan( ) = . C1 The formula for R is wonderful. But, the formula for tan( ) does not determine . It is not always true that = arctan(C2 /C1 ). If is in the second or third quadrant, then = arctan(C2 /C1 ) + . If C1 = 0, then tan( ) is undened. Draw a picture of the point (C1 , C2 ) in the plane to help you determine . Example: Write u = cos 3t + sin 3t in the form u = R cos(0 t ). First of all, we must have 0 = 3. Then observe that C1 = 1 and C2 = 1. Now draw a picture of the point (1, 1) in the (C1 , C2 ) plane. The polar coordinates of 2 and = 3/4. (We have also this point are easily obtained from geometry: R = i3/4 .) Therefore, shown that 1 + i = 2e u(t) = cos(3t) + sin(3t) = 2 cos(3t 3/4). The amplitude of the oscillation is R = 3. This identity can, and should, be checked with a calculator. Plot the two expressions as y 1 and y 2, with either y 1 dotted or y 2 thick. You should see the graph of y 2 drawn on top of the graph of y 1. (3)

Driven Oscillators: The most important problems involving oscillators have a forcing (or driving) term on the right hand side: We will focus on the problem mu + u + ku = F0 cos(t), (4)

where m > 0, 0, and k > 0. This problem can be solved by the method of undetermined coecients. This is quite easy when = 0, but its a mess when > 0. We separate the problem into three cases: Case I: = 0, = 0 = k/m. Case II: = 0, = 0 . Case III: > 0 Examples of Case I and II: Solve the IVPs u + ku = cos(3t), u(0) = 0, u (0) = 0, with k = 10 (Case I) and k = 9 (Case II). The solutions are u(t) = cos( 10t) + cos(3t), and u(t) = 1/6 t sin(3t), respectively. Here are some graphs of solutions.
u 0.6 0.4 0.2 0.2 0.4 0.6 k 10 0.6 0.4 0.2 t 1 2 3 4 0.2 0.4 0.6 1 2 3 4 u k 9

Figure 1: The solutions to the IVPs for a short time interval. The solutions look almost identical.

u 2 k 1 t 10 1 2 20 30 40 2 4 6 10 6 4 2

u k 9 t 10 20 30 40

Figure 2: The solutions to the IVPs for a longer time interval. The envelope of the solution with k = 10 can be obtained from the trig identities cos(A B ) = cos(A) cos(B ) sin(A) sin(B ).

We can combine these together, anticipating that we want an identity for u(t) = cos( 10t) + cos(3t), where 10 > 3: cos(A + B ) + cos(A B ) = 2 sin(A) sin(B ) Applying this to our problem with A + B = 10 and A B = 3, so that A = ( 10 + 3)/2 3.08 and B = ( 10 3)/2 0.0811: u(t) = cos( 10t) + cos(3t) = 2 sin(A t) sin(B t) = [2 sin(B t)] sin(A t) The carrier wave is sin(A t) which has period 2/A 2.04 and the envelope is u(t) = 2 sin(B t). Now, 2 sin(B t) has period 2/B 77.4. The envelope is plotted as a dotted line in Figures 1 and 2. Figure 2 (k = 10) shows the phenomenon of beats, wherein the amplitude of the oscillation varies with a period of half of 77.4, which is 38.7. Case III: This is the most important case. Any oscillators you will run into will have some damping. The general solution to equation (4) is u(t) = uh (t) + Up (t) where uh (t) 0 as t , since the oscillator is damped ( > 0). The steady state solution, also called the forced response, has the same frequency as the driving term: Up (t) = A cos(t) + B sin(t) = R cos(t ) If the A, B form of the particular solution is substituted into equation (4), we get a system of two linear equations for A and B which have the solution
2 F0 m(0 F0 2) A= 2 2 , B= 2 2 . 2 2 2 2 m (0 ) + m (0 2 )2 + 2 2 Note that these expressions use 0 = k/m, even though the oscillator is damped. The rectangular coordinates A and B can be converted to polar coordinates R and , as described in in 3.7. Note that B > 0, so is in quadrant I or II. The inverse cosine function gives an angle in quadrant I or II, so the best expressions are ( ) 2 F0 m(0 2) R= , = arccos 2 2 m2 (0 2 )2 + 2 2 m2 (0 2 )2 + 2 2

These expressions for the forced response are quite complicated. Nonetheless, we can get a feeling for them by considering limits and drawing gures. Note that as 0, the response has A F0 /k and B 0. This makes sense, because the right-hand side of (4) is F0 when = 0, and a particular solution in this case is Up (t) = F0 /k . Furthermore, R 0 as . All of the constants m, , etc. make the expressions for A and B look complicated. However, things are much simpler if we dene two dimensionless quantities: x= m0 , and Q = 0

Here, x is the ratio of the driving frequency ( ) to the natural frequency (0 ). The amount of damping is measured by the so-called oscillator Q, or quality factor. A high Q oscillator has very little damping. With these two variables x and Q, the expression for R is more understandable: R= F0 1 2 k (1 x )2 + (x/Q)2

If we treat Q as a constant, then R as a function of x is a maximum at xm , with a maximum value of Rm , given by ( ) F0 1 Q F0 1 2 , Rm = xm = 1 Q+ 2Q2 k k 8Q 1 1/(2Q)2 provided that Q2 1/2. The approximate expression for Rm is good for Q large. Note that x2 m is half way between 1 (when = 0 ) and the square of the quasifrequency , since it can be shown that (/0 )2 = 1 1/Q2 .
Amp 10

2 1 x 0.25 0.5 0.75 1 1.25 1.5 1.75 2

Figure 3: The scaled amplitude (Amp = Rk/F0 ) of the steady state response as a 1 1 . The dotted line goes through the function of x = /0 for Q = 10, 5, 2, , and 3 2 maxima of the curves, (xm , Rm ).

3 4 2 4

x 0.25 0.5 0.75 1 1.25 1.5 1.75 2

Figure 4: The phase of the steady state response as a function of x = /0 for Q = 10, 5, 2, 1/ 2, and 1/3. As Q , the phase approaches the step function = 0 if x < 1 and = if x > 1.

Finding the Steady State Oscillator Response Using Complex Numbers: The standard method of calculating of A and B in the particular solution is truly gruesome. However, it is quite easy using (what else?) complex numbers. The ODE (4) has the form ( ) L[u(t)] = mu + u + ku = Re F0 eit where L is a linear operator. We look for a solution of the form ( ) Up (t) = A cos(t) + B sin(t) = Re (A i B )eit When we plug this into the ODE to nd Up (t), we just have to solve L[(A i B )eit ] = F0 eit . Example: Find a particular solution to L[u] = u + u + 4u = cos(t) To nd the particular solution, we need to solve L[(A i B )eit ] = eit for A and B . The operator L[(A i B )eit ] is easily computed, since A and B are constants. The result is (A i B )( 2 + i + 4)eit = eit . We can divide both sides by eit to get (A i B )( 2 + i + 4) = 1, or (A i B ) = 1 1 4 2 i = = . 2 + i + 4 4 2 + i (4 2 )2 + 2

4 2 Therefore A = and B = . The steady state solution is (4 2 )2 + 2 (4 2 )2 + 2 4 2 Up (t) = cos( t ) + sin(t) (4 2 )2 + 2 (4 2 )2 + 2 be a complex Alternative Method: The way to really do these problems is to let A amplitude and write it ) = Re(A ) cos(t) Im(A ) sin(t) Up (t) = Re(Ae Then the previous problem becomes: ( 2 + i + 4)eit = eit . A or = A 1 1 4 2 i = = . 2 + i + 4 4 2 + i (4 2 )2 + 2

This gives the same solution as before. (These two methods are almost the same.) Extra Credit: Use either of these complex number methods to justify the general formulas (10), (11), and (12) in 3.8 of the book. This is worth 5 class points, and its not really very hard. Due in class on the review day for Exam 2.