Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: [Phys-l] question about coupled oscillators



It took me a while but I think I see where the problem is. Following Carl's initial wording I used the word "phase" incorrectly in connection
with the ratio of components of each eigenvector. For the simplest case where the 2 oscillators have the same mass and the same
natural frequency ( omega squared = k/m) one finds two modes of oscillation: One where the two masses move "in phase" with each other
because the ratio of the amplitudes of the components of the eigenvector is 1 ( omega squared = k/m for this one), and one where
the two masses move in "opposite" phase because the ratio of the amplitudes is -1 (omega squared = k/m + kc/m where kc is the spring constant of
the spring coupling the two masses) . My understanding was that Carl was referring to the ratio of 1 as phase =0 and -1 as phase = pi.
The actual motion of the masses is of course a linear superposition of the two modes with magnitudes and relative phases that depend on the initial conditions.

Now, unless I goofed in my calculation, if the masses are not equal and or the springs tying each mass to a wall (in my example, but in general this would mean that the free oscillations of each
oscillator have different frequencies) are different, the ratios of the amplitudes of the components for each eigenvector will not be the usual +1 or -1. The spring + mass case is rather messy but
the case of two pendulums of same lengths and different masses is cleaner. One finds that in the lower frequency normal mode (omega squared = g/l) the ratio of amplitudes (angles in this case) is 1. In the higher frequency mode ( omega square = g/l + k/mu where k = coupling spring constant and mu = reduced mass) the ratio of the amplitudes is - m1/m2
(negative of the ratio of the masses which goes to -1 if m1 = m2) .

Karim Diff


Jack Uretsky wrote:
Hi all-
The coupled oscillator problem is discussed at length in Olenick, et al. <The Mechanical Universe> (NOT the advanced edition), Ch 22.
The problem of 2 coupled oscillators can be worked out easily, TKarim's example is OK (though the conclusion is erroneous). The paradigm equation is (' stands for time derivative, f^2 means f squared)
x'' + f1^2 x +k(x-y) =0
y'' + f2^2 y +k(y-x) =0

The eigenvalue equation is a quadratic in the square of the eigenfrequencies; f1, f2, k are real constants. It is easy to show that
both squared eigenfrequencies are real and positive.
The eigenvectors are two dimensional column vectors, one for each (squared) eigenfrequency. The eigenvalue condition gives 1 relation between the two components of each eigenvector. The normalization condition requires the squared magnitudes of the two components to sum to unity. Since only the magnitudes are involved, there is an undetermined, arbitrary phase for each eigenvector. The relative phases may be determined from the starting conditions on x and y. Different phases may refer to different starting conditions, but 0 and/or pi are certainly not mandated in general.
Regards,
Jack


On Wed, 8 Nov 2006, Karim Diff wrote:

I did not work it out in detail yet but what about this:
2 unequal masses & 3 horizontal springs (different k's) between 2 walls.

In crude ASCII

Wall |/\/\/\/\[ m 1 ]/\/\/\[m 2 ]/\/\/\/| Wall

The phases are 0 & pi if the two masses and the springs are the same
(basic coupled oscillators)
but I believe the phase relationship becomes more complicated in the
general case (2 different masses and 3 different k's).

Karim Diff

Carl Mungan wrote:
So if the eigenvectors can be complex even when one component is
chosen to be real, then that implies that the oscillators in a normal
mode need NOT pass through the equilibrium position at the same
instant (ie. relative phases other than 0 or pi are possible).

In that case, can someone rig up a *simple soluble* example of such
an eigenvalue problem?

To make my request clear for those who have forgotten the beginning
of this thread. I want an example of a system of coupled oscillators,
as simple as possible, where the relative phases between oscillators
in a normal mode are neither 0 nor pi. All the usual textbook
examples I can think only have in-phase or exactly-out-of-phase
relative motions of the oscillators.


The correct statement, that you seem to be striving for, is that the
eigenvalues of a Hermitean matrix (includes real, symmetric
matrices) are real. This says nothing about the scaling of the
eigenvectors, which are usually scaled by normalizing them to a
(squared) norm of unity - which leaves them with an undetermined
phase. That phase is just the subject of the present discussion.
Regards,
Jack


Is that true even if I insist that one component arbitrarily have the
real value 1, to remove the indeterminacy in the overall scaling of
an eigenvector? This way d_i is the relative phase, which is what I
meant.


The argument is erroneous. The reality condition is on the
eigenvalues, not the eigenvectors.
Jack


Having thought about it a little more, this is what I come up with:

Newton's third law says that oscillators i and j exert equal
magnitude forces on each other. This in turn will lead to elements
K_ij and K_ji in the force-constant matrix K to be equal. In turn,
this means K is symmetric (and of course it's real); and so is the
mass matrix M. But this means the eigenvectors must be real, even if
I write the oscillator displacement vector x in complex form where
the i-th component is A_i exp (i d_i). But this means d_i can only be
0 or pi.

What do you think of this argument? -Carl


_______________________________________________
Forum for Physics Educators
Phys-l@carnot.physics.buffalo.edu
https://carnot.physics.buffalo.edu/mailman/listinfo/phys-l