L19.1 Time delay and resonances (18:18)
L19.2 Effects of resonance on phase shifts, wave amplitude and time delay (14:53)
L19.3 Modelling a resonance (15:37)
L19.4 Half-width and time delay (08:17)
L19.5 Resonances in the complex k plane (15:14)
L19.1 Time delay and resonances (18:18)
MITOCW | watch?v=mnvYIEbJXlM
PROFESSOR: Let me begin by introducing the subject. The subject is resonances. And we have seen,
actually, a little bit of this in the resonant transmission of the Ramsauer-Townsend effect.
Because of a resonance phenomenon within the square well obstacle, somehow, for some
particular frequencies, for some particular energies, the particles were able to zoom by without
experiencing any reflection, whatsoever.
So let’s begin the subject of resonances by asking a question. If you have the usual potential,
the short range potential, which means, that for some distance, R, greater than 0, the potential
is 0. Here we put a barrier, and over there could be anything, some potential. We’ve computed
some-- this concept of time delay, there’s a formula for the time delay. In fact, it was given by
2 h bar d delta dE, the time delay, 2 h bar d delta dE.
And we discussed that this time delay can be positive or it can be negative. If it’s positive, it
really means a time delay. You send in a wave packet. And it takes time to come back, more
time than it would have taken if there had been no potential.
You see, the time delay, you have a packet coming in from time minus infinity. And then it
bounces back a time equal infinity. But nevertheless, you compare that with a situation in
which there’s no potential. And you see that there is some time delay. If you time the wave
packet to reach at time equals 0, here, it will not reach back to where you were by time-- by
whatever time-- suppose you have the wave packet here at t equal minus 10, then it goes
here, and it delays, and at t equal 10, the packet has not reached, there is a time delay, a
positive time delay.
A negative time delay is the opposite. The packet arrives a little earlier. And the question I
want to ask you, if you have a negative time delay, can it be arbitrarily large. Well, if you send
in a wave packet, it may find an infinite wall here, and then may bounce, and then yes, it
comes back earlier than you expected, because the free packet would have gone here and
back.
But you wouldn’t expect it to be able to come earlier than if there was an infinite wall here,
because there is no infinite wall here, nor an infinite wall here. So it’s just not going to bounce
before it reaches here. The best it can do is bounce when it reaches here. So you should not
expect, and this, sometimes, will [INAUDIBLE], there is nothing that can make it bounce until
you reach here. So you cannot expect that the time advance is larger as if it would have
bounced before reaching the obstacle, whatsoever.
So this is important. We cannot have a negative time delay that this infinitely large. So, in fact,
the time delay as, we’re right in here, should be greater than the total travel distance that you
may save. If you bounce here, you would save 2R over v. And you must be greater than that
negative number, which is the total travel time that it would take to go back and forth, here.
So we can do a little arithmetic, here. This is equal to 2 h bar d delta dk, and here, dE dk. This
is still greater than or equal than minus 2R over v. And I want to put a sim, because our
argument is not completely rigorous as to what’s happening when it reaches here. It seems
very plausible classically, but there’s a bit of a correction if you do it exactly. So it’s not an
exact inequality we’re deriving.
And what is the E dk is h bar times the velocity. Remember, dE dk, you are differentiating h
squared k squared over 2m. And you get h bar times hk over m. So therefore, this is h bar and
the velocity. And the h bars cancel. The velocities cancel. Between these two sides, the 2s
cancel. And you’ll get that d delta dk must be greater than or equal, approximately, to R.
And that’s sometimes called Wigner’s condition on scattering. And it basically is the idea that
the time delay, the time advance cannot be too large. OK, so now we can ask the second
question. How about the time delay, a true time delay, can it be very large? Can it be arbitrarily
large?
Suppose we have a barrier of this form. And now you send a particle with a little bit higher
energy here. Now, this particle is going to have very little kinetic energy. So it’s going to travel
quite slowly here, and go back. And this time, it’s going to delay quite a bit, probably. But the
problem is, if you create-- there’s nothing very peculiar about this, if you go a little lower, than
you’re advanced, and then suddenly, it gets delayed. It’s not that evident, but the phenomenon
of resonance is precisely what we get when we, sort of, trap the particle. And then we make it
be, as far as it seems, arbitrarily large, if you design a well properly.
But the thing that we have to design, the example of what we’re going to design, is different
from all the things I’ve drawn so far. It’s the following way, this is just an example. I have this
zero line of the energy. This is v of x. This is x. And then I put an attractive potential here. And
here is minus v0. And then I put a barrier here with a v1.
So what I’m going to aim at is, you see, if v1 will be extremely large, there will be-- well, if v0 is
extremely large, then begin there would be bound states here, but these are not scattering
states. On the other hand, if v1 will also be infinite, you would have bound states here, but
they could not escape. So certainly, if I combine these two, I put a v0 and maybe a larger v1, I
can almost create bound states here.
But they’re not really bound states, because they can leak out and produce scattering states.
But these are going to be resonances. This part and this, this being a attractive, trying to keep
the particle in, and this being a barrier, can combine to produce a state that gets trapped here,
and stays a very long, time, will have a very long time delay.
And that’s the phenomenon of resonances. We need to trap that particle, somehow. And we’re
going to see now the details of how this works, and what the properties are. Now, it’s very
interesting that actually, these resonances occur at some particular energies. And they have
different properties.
But we can identify energies of resonances. And these are not bound states. They’re just
resonances. They eventually escape. And they’re not normalizable, really, but in some ways
they behave as bound states for awhile. They stay there for a while and do nice things.
So let’s set this off. Now we’re going to spare you a little bit of these calculations, because the
important thing is that you know how to set it up, and if you get an answer, you know how to
plot it, how to get the units out, how to try to understand it. So that’s what we’re going to do.
I’m going to put an energy here, an energy, E. And I’m going to receive E to be less than v1
and greater than 0. I don’t expect true resonances beyond, because the particle just bounces
out. It doesn’t get trapped. The phenomenon of resonance is a little more intricate than just
having a long time delay. There’s more that has to happen.
Another thing that will happen, is if the particle spends a lot of time here, you would find, in this
spirit of resonance, that the amplitude of the wave function here is going to be very big. So you
will scan the energy and the amplitude. It will be normal, normal, normal. And suddenly for
some energy it becomes very big. And we’re going to do that.
The way I’m going to develop that, we’re going to calculate this, plot these things. And then we
are going to ask whether there is a mathematical condition that picks resonances. Well, how
do I, if I want to explain to somebody in 30 seconds where are the resonances, how do you
calculate them, you cannot tell that somebody, OK, calculate it for all energies, do all the plots,
and see some peak in some thing, and this is a resonance. This is what we’re going to do to
begin with, but then we’ll get more sophisticated.
So let’s put k in. So let’s call this k prime, the wave number in this area. Kappa here, because
it’s a forbidden region, and k over here, as usual. So k squared is 2mE over h bar squared. K
prime squared is equal to 2m, the total kinetic energy is E plus v0, over h squared. And kappa
squared is again, similar formula, but this time is the energy differential between v1 and E, so
2m v1 minus E over h squared.
All of these three numbers are positive. And they are the relevant constants to write wave
functions. So we have to write a wave function. And I’m going to write a wave function because
it takes a little tinkering to do it in an efficient way.
There is one that you don’t have to think, you just have to remember. It’s the one outside. It’s
the universal formula, e to the i delta sine kx plus delta is valid for for x greater than R. This
one we derived at the beginning of our analysis of scattering.
How about the other region. Oops, I should have put letters here. These are a and 2a they are
positions. And therefore, it’s not R in here. Well, it’s R, it’s the range of the potential, but here
is 2a.
How about the other one? In this region, it’s kind of simple again. The wave function has to
vanish here, has to be sines or cosines of k prime. So it has to be a sine function of k prime.
And since we don’t put an extra constant in here, we kind of put an extra constant in here,
there must be a constant here, A sine of k prime x. And that must be for x between 0 and a.
We used k prime, the wave from over there. And there is A. And what we were saying about
resonances, is that, well, A may depend on k. And when you have a resonance, A is going to
[INAUDIBLE], presumably because the particle spends a long time inside the well.
And now I have to write this one in here. And this is the one that, you can do it, do a little bit
more work, or do it kind of, efficiently. In that region we have exponentials, like we have e to
the kappa x and e to the minus kappa x. Or I may want to have sinh of kappa x and cosh of
kappa x to write my solutions.
But I actually don’t want either of them too much, because I would like to write an answer that
almost imposes continuity in a nice way. So I could use sinh of kappa x minus a and cosh of
kappa x minus a. These are all solutions. You can choose whichever pair you want.
So for example, if I want to implement continuity with this thing, this wave function, I want to
write something that I don’t have to write another equation for continuity. So I will write A sine
of k prime a-- so far, this wave function, if x equal a, coincides with this one. But this is no
wave function yet, not with an x dependence, so I have to put more. But then, I know that cosh
is 1 for x equals zero. So I put here a cosh kappa x minus a.
And now this is a solution that matches that one at x equals a. At x equal a, the cosh becomes
1 and matches. But this kind of need a complete solution. It’s not general enough. So you
have to put a B sinh of kappa x minus a.
And this won’t ruin the matching, because at x equal a, that second term vanishes. So we’re
still matching well there. And Well matching here is non-trivial when I impose some conditions.
So you still have to match derivatives and do a little bit of work but not too much work.
L19.2 Effects of resonance on phase shifts, wave amplitude and time delay (14:53)
MITOCW | watch?v=VY-_xLxHQbA
PROFESSOR: This is the answer. Tan of 2 k a plus delta. It’s a little messy, no wonder. Just plotted this
functions. [INAUDIBLE] Sin k plus a cosh kappa a [INAUDIBLE] plus a prime over kappa cosh
a prime a sinh kappa a. You have sin a prime a sinh. The sinh causes outer [INAUDIBLE] the
same but this one changes. Centered here, k prime over kappa cos k prime a cosh kappa a.
So it’s a well-defined expression, if you know the number of kappa a, you can calculate
everything. Let me just make sure you can see that. If you know-- if you said, for example, ka,
this is a natural variable. ka is just-- the k is related to the energy direction. So ka has no limits,
call it u.
Then you have two parameters of the square well [INAUDIBLE], a depth we don’t and a height
we want. So there’s natural to define just [INAUDIBLE] would have z 0 squared which is 2m v 0
a squared over h squared. You also define z 1 squared [INAUDIBLE] as 2m v1 a squared over
h squared.
And then the energy-- we can do another thing, we’re going to find the energy divided say, by
v 1 to be little e. And that’s reasonable because these energies is compared always with v 1,
and we’re solving-- we’ve solved this problem in the domain when the energy is less than v 1,
and that’s why we’ll have kappas here, and if energy was bigger than v 1, you would have
trigonometric functions everywhere.
Now you can switch from trigonometric to hyperbolic by analytic continuation, letting and angle
become imaginary, a trigonometric function becomes a hyperbolic function. Most of us are
rather comfortable doing this at the beginning, because of the matter of sine if you mess up a
sine, it’s a big problem.
But at the end of the day, it actually saves a lot of work. So a little of that in the homework you
will see. But at any rate, this is valued for e in this range. And therefore this ratio is h squared k
squared. You can put an a squared over 2m v 1 a squared. And you can see a u squared
here. And-- I’m sorry, not there. A u squared here, and the h squared divided here gives you a
z 1. So this is u squared over z 1 squared.
A couple of extra things-- I’m just putting it here because if you ever have the curiosity of doing
this plots, this may help. k prime squared is related to the energy, therefore u squared and v 0
z 1 squared. This is-- c 0 squared, [INAUDIBLE] 0. And for the other one, kappa a squared is
equal to z 1 squared minus u squared.
So everything becomes a function of u Wherever you have a kappa prime a you have a
square root of u. Now, somebody has to give you the values of z 0 and z 1. Those are the
data of the potential-- z 0 and z 1 are numbers you know then. Therefore you know kappa a is
a function of u, kappa-- k prime a is a function of u, and ka, which is u. Therefore this equation
becomes a function of u [INAUDIBLE] this arc tangent here, and so for delta.
It’s messy. There’s no-- remember I mentioned the other day the fact that you could do
trigonometric identities, and so for tan delta here. But then this I don’t think simplifies when you
solve that. It becomes just a bigger mess. You did solve for tan delta, it’s very messy.
So anyway, let’s leave it there. Are there any questions in solving this problem? So in principle,
we solved it. I didn’t plot anything, so you still don’t have any insight as to what’s happening.
But you’ve learned in principle how to solve it. Any questions?
OK, so let’s plot this then. Maybe I should start here. OK. I’ll start here. Oh, if I’m going to plot,
I have to choose things. So we’ll choose z 0 squared equals to 1, or z 0 equals to 1. And z 1
squared equals to 5. So we put a big barrier there.
And now let’s go delta as a function of ka or u. Now, from this equation, we see that u and the
little e must be at most 1 for our formulas to be correct. So u must be up to square-- up to z 1,
because you must keep this ratio less than 1. So we can plot u up to square root of 5. 5,
here’s 2. And that’s it.
And here we’re going to plot delta of u. And here is minus pi over 2 minus pi. And so what
does it do, the phase? There no way you can guess, I think, from this formula. I could not
guess from this formula. So we could try to imagine-- it’s possible to guess, actually, after
you’ve solved this week’s homework.
You probably will have a good guess that the phase shift begins with minus 2ka, as if with very
little energy you reflect back here. So there will be a shift that is calculable without doing any
work. So it begins linearly. And it represents a time advance. So it goes linearly and negative.
That’s how it begins.
Because for very little energy, it’s going to bounce back. And you know that the delay is
proportional to this derivative. So must be negative like that. So then what does it do? It
crosses this point, which is almost pi, and here is what something quite remarkable happens.
That some value you star, which is about 1.8523, I think. That’s what I calculated. Let’s see
[INAUDIBLE].
The face that is almost minus pi suddenly jumps very, very fast, crosses pi over 2, and then
about [INAUDIBLE] something like this. I don’t know what else it does, because we haven’t
calculated it. But it jumps very fast. Almost a value of pi. Now, let me-- this is quite interesting.
If you think of what we used to call the scattering amplitude as squared was sin squared delta.
The amplitude of the scattered wave, sin squared delta, the amplitude of the scattered wave is
going to be quite large here, it’s going to be 1, which is the maximum, as squared is here. So I
want to keep these two blackboards aligned.
So here it goes like this. It’s going to go up like this, and do this, and broadly go down, and
then very sharply go up. More sharply, at least. Go up and then something like that. So here it
is. This point over here has a strong scattering amplitude, but there’s nothing too dramatic
happening here.
This sine corresponds to time advanced, the derivative is negative. And time advanced cannot
be too large, as you know. On the other hand, here is time delayed, and apparently they can’t
be very large. So we think this must be the resonance, and this is not a resonance, even
though the scattering amplitude is bigger.
So we continue here, and plot the time-- interestingly, the amplitude inside the well. So here it
is. How does the wave function become, through this constant, inside the well? And indeed,
the this confirms that nothing very special is happening here. What happens now is some sort
of behavior like this, and a big jump here, in which the amplitude apparently-- I have not quite
confirmed this number, it’s at least a value 3. [INAUDIBLE] very large and short.
I should have room for one more plot [INAUDIBLE] the star plot of this thing is-- I’ll do it
compressed here. The total delay, 1 over a d delta dk. Well, it begins negative, and remember
that when we did the 1 over a, d delta dk, this is a pure number. It expresses the delay in
terms of the time that it would take to travel the inside region.
So how many-- if you would get a 1, or a minus 1, it’s just a delay of the size of the time
needed to travel back and forth. So actually this goes a little negative at the beginning-- we
know the derivative is like that-- and when you plot this, you see that it’s very sharp, and it’s a
value of about fourteen. Fourteen times gets delayed from what you would expect naturally of
the time that it should have spent travelling back and forth A gigantic time delay. A peak in the
time delay. Peak in time delay. Peak in the amplitude inside the well.
L19.3 Modelling a resonance (15:37)
MITOCW | watch?v=8Dxo4LPK_9w
PROFESSOR: That’s how it looks, a resonance. You can see it basically in the phase shift. And great
increase of the phase of almost minutely pi over a very small change of energy. And it should
[INAUDIBLE] with a very big [INAUDIBLE].
So this is how it looks. And I want to now proceed, after if there are some questions, of how do
we search for residences a little more mathematically rather than plotting them. How could I
write an equation for a resonance. Cannot say, oh, the phase changes fast. Well, that’s not a
very nice way of saying it. It’s good. It’s intuitive. But we should be able to do better.
So how do I find resonances? So let’s model resonances a little bit. How do we find
resonances? So let’s model this behavior. By that is writing a formula that is simple enough
that seems to capture what’s happening. And that formula’s going to inspire us to think of
resonances perhaps a little more clearly.
So suppose you have a resonance near k equal alpha. I claim the following formula would be
a good way to represent the resonance? We would say that tan delta is equal to beta over
alpha minus k. Or-- yeah, we would say that. [INAUDIBLE] Or if you wish, delta is tan minus 1
of beta over alpha minus k.
Why is that reasonable? It’s a little surprising, but not that surprising. You see that-- delta is
equal to minus pi over 2. The tangent of delta goes to infinity. So there’s something going on
here in which you have this property. So let’s plot this. So let’s plot beta over alpha minus k.
You need a clock to understand this.
So this is k, and we’re plotting this quantity. Well, it’s going to go crazy at k equal alpha. That
we know. When k is less than alpha, I’m going to assume that alpha and beta are positive.
They both have units of k. And when k is less than alpha-- we begin here-- then this
denominator is positive, the numerator is positive, ratio is positive. It’s small, maybe.
And then suddenly, when k reaches alpha it goes to infinity. So it’s going to be like that. Now, it
actually is true that when k differs by alpha by beta, it reaches value 1. So here is alpha minus
beta. That point it reaches value 1.
So if I want this thing to be very sharp, I need beta to be small so that it’s little until it reaches
beta within-- distance beta within alpha, and then it shoots up. So I want beta to be small for
sharp behavior. On the other hand here, it goes the other way. It goes from minus infinity back
to 0, and has value minus 1 at alpha plus beta.
So within minus beta, and beta off of the center alpha, most of the things happen. If we plot
now the tangent of this, or the arctangent of this, tan minus 1 of beta, alpha minus k, well, if
the tangent of an angle is very little, the angle can be taken to be very little. At this point, it will
reach pi over 2, so the angle is little, will go to pi over 2, and then quickly becomes larger than
pi over 2, you’re thinking tangents.
So the tangent is going up, is blowing up at pi over 2. Then continuously, it goes to minus pi
over 2, and then continuously goes to 0 so it reaches pi. So this is the behavior of delta. Delta
is this tan minus 1 of beta over that. And delta is doing the right thing. It’s doing this kind of
behavior.
There is a shift. I could add a constant here to produce this shift, but it’s not important at this
moment. The resonance is doing this thing, up to a total shift of pi that doesn’t change the
tangent of an angle. So this is one way of modeling what’s happening to the phase shift near
our resonance. So let’s explore it a little more.
I can do a couple of calculations. For example, I can compute what is d delta dk at k equals
alpha. That’s should be a nice quantity. Is a derivative of the face. At the resonance, at the
position alpha of the resonance. So here’s delta, here is k, and there’s the derivative of this k.
And how should it be? Well, basically, the phase changes by amount pi over a distance beta or
2 beta. So this must be a number divided by beta. You can calculate this derivative from this
equation. It’s a nice exercise. It’s actually just 1 over beta. That’s a result. 1 over beta.
The other quantity that is nice to understand is how does this scattering amplitude behave
near the resonance. So what is the value of as squared? Oh, that’s the absolute value of psi s
squared, which is sine squared delta. That’s the same thing as As squared. Well, you know
what is the tangent of delta? A little trigonometric play should be able to do it, and can give you
the sin squared delta.
And here is the answer. It’s beta squared over beta squared plus alpha minus k squared. Kind
of a nice, almost bell shape. Of course, it’s polynomial, but it looks a little like just a nice
symmetric shape around alpha equal k.
Now, this division is so famous it has been given a name. It’s called the Breit-Wigner
distribution. Breit-Wigner. But it’s described as the Breit-Wigner distribution, and it’s usually
referring to terms of energy. Of energy, not momentum. So-- and it’s-- what should happen to
scattering amplitude in general when you have a resonance.
So the way to do this calculation now is to say, well, what is alpha minus k? Let’s try to relate it
to the energy minus the energy at k equal alpha. Well, this is h squared k squared over 2 m
minus h squared alpha squared over 2 m, which is h squared over 2 m, k squared minus
alpha squared. On the other hand, I have here alpha minus k squared. I don’t have k squared
minus alpha squared.
So-- approximations. if the resonance is narrow enough, if beta is small, let’s do an
approximation. We do h squared over-- everybody knows this approximation, shouldn’t be
afraid of doing it. It’s alpha-- how could I write it-- k minus alpha times k plus alpha. And the
approximation is that all the interesting thing comes from the difference between k and alpha,
how close k is to alpha.
So when k is close to alpha, all the dependence is going to be here. This is going to be about
2 alpha when k is near alpha. And if it’s a little more than that, it doesn’t matter, because it’s
[INAUDIBLE]. So this could be approximated to 2 alpha, and therefore this becomes h squared
alpha over m times k minus alpha. So that’s a little help.
Then size of s squared, doing a little more of algebra with the constants there. Probably you
want to do it with two lines. It’s tricky. It’s really simple. It’s always written in this form-- 1 over 4
gamma squared over e minus e alpha squared plus 1 over four gamma squared, and that’s
the so-called Breit-Wigner distribution.
And gamma is a funny constant here. We’ll try to understand it better. 2 alpha beta h squared
over m. It has to be something that depends on alpha and beta, because after all, we weren’t
modeling the resonance with alpha beta. So this curve is very famous. That’s the distribution
of the scattering amplitude over energies whenever you have a resonance. So we should plot
it.
You have an e alpha. You have an e alpha plus gamma over 2 and e alpha minus gamma
over 2. But the energy minus e alpha is equal to the gamma over 2, you get the gamma
squared over 4 so the total amplitude goes down to 1/2 of the usual amplitudes. When the
energy is equal to e alpha, you get 1. 1 for psi s squared.
But when the energy differs from e alpha by gamma over 2, you get half. So-- actually, I’m not
sure of the deflection point, where it is. Probably not there. Or is it there? I don’t know. I drew it
as if it is there. So that’s the distribution, and the width over here is gamma. So gamma is
called the width at half power, or at half intensity. Yeah. Width-- the half width of the
distribution.
L19.4 Half-width and time delay (08:17)
MITOCW | watch?v=mnvYIEbJXlM
PROFESSOR: Let me begin by introducing the subject. The subject is resonances. And we have seen,
actually, a little bit of this in the resonant transmission of the Ramsauer-Townsend effect.
Because of a resonance phenomenon within the square well obstacle, somehow, for some
particular frequencies, for some particular energies, the particles were able to zoom by without
experiencing any reflection, whatsoever.
So let’s begin the subject of resonances by asking a question. If you have the usual potential,
the short range potential, which means, that for some distance, R, greater than 0, the potential
is 0. Here we put a barrier, and over there could be anything, some potential. We’ve computed
some-- this concept of time delay, there’s a formula for the time delay. In fact, it was given by
2 h bar d delta dE, the time delay, 2 h bar d delta dE.
And we discussed that this time delay can be positive or it can be negative. If it’s positive, it
really means a time delay. You send in a wave packet. And it takes time to come back, more
time than it would have taken if there had been no potential.
You see, the time delay, you have a packet coming in from time minus infinity. And then it
bounces back a time equal infinity. But nevertheless, you compare that with a situation in
which there’s no potential. And you see that there is some time delay. If you time the wave
packet to reach at time equals 0, here, it will not reach back to where you were by time-- by
whatever time-- suppose you have the wave packet here at t equal minus 10, then it goes
here, and it delays, and at t equal 10, the packet has not reached, there is a time delay, a
positive time delay.
A negative time delay is the opposite. The packet arrives a little earlier. And the question I
want to ask you, if you have a negative time delay, can it be arbitrarily large. Well, if you send
in a wave packet, it may find an infinite wall here, and then may bounce, and then yes, it
comes back earlier than you expected, because the free packet would have gone here and
back.
But you wouldn’t expect it to be able to come earlier than if there was an infinite wall here,
because there is no infinite wall here, nor an infinite wall here. So it’s just not going to bounce
before it reaches here. The best it can do is bounce when it reaches here. So you should not
expect, and this, sometimes, will [INAUDIBLE], there is nothing that can make it bounce until
you reach here. So you cannot expect that the time advance is larger as if it would have
bounced before reaching the obstacle, whatsoever.
So this is important. We cannot have a negative time delay that this infinitely large. So, in fact,
the time delay as, we’re right in here, should be greater than the total travel distance that you
may save. If you bounce here, you would save 2R over v. And you must be greater than that
negative number, which is the total travel time that it would take to go back and forth, here.
So we can do a little arithmetic, here. This is equal to 2 h bar d delta dk, and here, dE dk. This
is still greater than or equal than minus 2R over v. And I want to put a sim, because our
argument is not completely rigorous as to what’s happening when it reaches here. It seems
very plausible classically, but there’s a bit of a correction if you do it exactly. So it’s not an
exact inequality we’re deriving.
And what is the E dk is h bar times the velocity. Remember, dE dk, you are differentiating h
squared k squared over 2m. And you get h bar times hk over m. So therefore, this is h bar and
the velocity. And the h bars cancel. The velocities cancel. Between these two sides, the 2s
cancel. And you’ll get that d delta dk must be greater than or equal, approximately, to R.
And that’s sometimes called Wigner’s condition on scattering. And it basically is the idea that
the time delay, the time advance cannot be too large. OK, so now we can ask the second
question. How about the time delay, a true time delay, can it be very large? Can it be arbitrarily
large?
Suppose we have a barrier of this form. And now you send a particle with a little bit higher
energy here. Now, this particle is going to have very little kinetic energy. So it’s going to travel
quite slowly here, and go back. And this time, it’s going to delay quite a bit, probably. But the
problem is, if you create-- there’s nothing very peculiar about this, if you go a little lower, than
you’re advanced, and then suddenly, it gets delayed. It’s not that evident, but the phenomenon
of resonance is precisely what we get when we, sort of, trap the particle. And then we make it
be, as far as it seems, arbitrarily large, if you design a well properly.
But the thing that we have to design, the example of what we’re going to design, is different
from all the things I’ve drawn so far. It’s the following way, this is just an example. I have this
zero line of the energy. This is v of x. This is x. And then I put an attractive potential here. And
here is minus v0. And then I put a barrier here with a v1.
So what I’m going to aim at is, you see, if v1 will be extremely large, there will be-- well, if v0 is
extremely large, then begin there would be bound states here, but these are not scattering
states. On the other hand, if v1 will also be infinite, you would have bound states here, but
they could not escape. So certainly, if I combine these two, I put a v0 and maybe a larger v1, I
can almost create bound states here.
But they’re not really bound states, because they can leak out and produce scattering states.
But these are going to be resonances. This part and this, this being a attractive, trying to keep
the particle in, and this being a barrier, can combine to produce a state that gets trapped here,
and stays a very long, time, will have a very long time delay.
And that’s the phenomenon of resonances. We need to trap that particle, somehow. And we’re
going to see now the details of how this works, and what the properties are. Now, it’s very
interesting that actually, these resonances occur at some particular energies. And they have
different properties.
But we can identify energies of resonances. And these are not bound states. They’re just
resonances. They eventually escape. And they’re not normalizable, really, but in some ways
they behave as bound states for awhile. They stay there for a while and do nice things.
So let’s set this off. Now we’re going to spare you a little bit of these calculations, because the
important thing is that you know how to set it up, and if you get an answer, you know how to
plot it, how to get the units out, how to try to understand it. So that’s what we’re going to do.
I’m going to put an energy here, an energy, E. And I’m going to receive E to be less than v1
and greater than 0. I don’t expect true resonances beyond, because the particle just bounces
out. It doesn’t get trapped. The phenomenon of resonance is a little more intricate than just
having a long time delay. There’s more that has to happen.
Another thing that will happen, is if the particle spends a lot of time here, you would find, in this
spirit of resonance, that the amplitude of the wave function here is going to be very big. So you
will scan the energy and the amplitude. It will be normal, normal, normal. And suddenly for
some energy it becomes very big. And we’re going to do that.
The way I’m going to develop that, we’re going to calculate this, plot these things. And then we
are going to ask whether there is a mathematical condition that picks resonances. Well, how
do I, if I want to explain to somebody in 30 seconds where are the resonances, how do you
calculate them, you cannot tell that somebody, OK, calculate it for all energies, do all the plots,
and see some peak in some thing, and this is a resonance. This is what we’re going to do to
begin with, but then we’ll get more sophisticated.
So let’s put k in. So let’s call this k prime, the wave number in this area. Kappa here, because
it’s a forbidden region, and k over here, as usual. So k squared is 2mE over h bar squared. K
prime squared is equal to 2m, the total kinetic energy is E plus v0, over h squared. And kappa
squared is again, similar formula, but this time is the energy differential between v1 and E, so
2m v1 minus E over h squared.
All of these three numbers are positive. And they are the relevant constants to write wave
functions. So we have to write a wave function. And I’m going to write a wave function because
it takes a little tinkering to do it in an efficient way.
There is one that you don’t have to think, you just have to remember. It’s the one outside. It’s
the universal formula, e to the i delta sine kx plus delta is valid for for x greater than R. This
one we derived at the beginning of our analysis of scattering.
How about the other region. Oops, I should have put letters here. These are a and 2a they are
positions. And therefore, it’s not R in here. Well, it’s R, it’s the range of the potential, but here
is 2a.
How about the other one? In this region, it’s kind of simple again. The wave function has to
vanish here, has to be sines or cosines of k prime. So it has to be a sine function of k prime.
And since we don’t put an extra constant in here, we kind of put an extra constant in here,
there must be a constant here, A sine of k prime x. And that must be for x between 0 and a.
We used k prime, the wave from over there. And there is A. And what we were saying about
resonances, is that, well, A may depend on k. And when you have a resonance, A is going to
[INAUDIBLE], presumably because the particle spends a long time inside the well.
And now I have to write this one in here. And this is the one that, you can do it, do a little bit
more work, or do it kind of, efficiently. In that region we have exponentials, like we have e to
the kappa x and e to the minus kappa x. Or I may want to have sinh of kappa x and cosh of
kappa x to write my solutions.
But I actually don’t want either of them too much, because I would like to write an answer that
almost imposes continuity in a nice way. So I could use sinh of kappa x minus a and cosh of
kappa x minus a. These are all solutions. You can choose whichever pair you want.
So for example, if I want to implement continuity with this thing, this wave function, I want to
write something that I don’t have to write another equation for continuity. So I will write A sine
of k prime a-- so far, this wave function, if x equal a, coincides with this one. But this is no
wave function yet, not with an x dependence, so I have to put more. But then, I know that cosh
is 1 for x equals zero. So I put here a cosh kappa x minus a.
And now this is a solution that matches that one at x equals a. At x equal a, the cosh becomes
1 and matches. But this kind of need a complete solution. It’s not general enough. So you
have to put a B sinh of kappa x minus a.
And this won’t ruin the matching, because at x equal a, that second term vanishes. So we’re
still matching well there. And Well matching here is non-trivial when I impose some conditions.
So you still have to match derivatives and do a little bit of work but not too much work.
L19.5 Resonances in the complex k plane (15:14)
MITOCW | watch?v=0T83-47Vi-M
PROFESSOR: So I want to go a little further to try to put resonances in a more intriguing footing. That you
can play with and if you-- at some point interested.
So let’s think of discovering [INAUDIBLE] that we have. We had A s-- remember the scattered
wave was A s e to the ikx [INAUDIBLE] that divided 2. And what was A s? Well, A s squared–
the sine square delta. So if you remember this was sine delta e to the i delta. So let’s stick to
that and try to write it in a funny way. Certainly, A s is becoming large near resonance, so let’s
think when A s becomes large. Well, in another way let’s be a little creative about things, It’s
good sometimes not to be logical.
So let’s write this as sine delta-- I’ll do it here-- sine delta over e to the minus i of delta . And
that’s sine delta over close delta minus i sine delta. That’s all good. A s-- let me divide by sine
delta both sides-- both numerator and denominator. So-- no divide it by cosine delta, so I’ll
have tan delta over 1 minus i tan delta. I divide it by cosine.
You want A s large? You really want it large, choose tan delta-- equals to minus i. Sounds
crazy, but it’s not really crazy. The reason it sounds crazy and it’s somewhat strange and not
very logical is tan delta is a phase and the tangent of any phase is never an imaginary
number.
So then I would have think of delta itself as a complex number. And what would that mean. So
things are weird. But it’s certainly the fact that A s will become infinite-- not just large-- but
infinite. A s will become infinite. And you say, wow, this doesn’t make any sense.
But maybe it makes sense in the following way. This is the line of real phase shifts.
[INAUDIBLE] are real. And here is the world of complex phase shifts. These are the real phase
shifts and there are the complex phase shifts. Maybe if the phase shift becomes infinite-- off
the real axis-- it’s just large on the real axis. So actually, if you wanted it to be very large you
would have to get off the real axis. If this sounds vague, it is still vague. But in a minute we’ll
make it precise.
So I suggest that we take this idea seriously-- that maybe this means something. And we can
try to argue that by looking back at what resonances do. So what I will do is look with
[INAUDIBLE] a resonance here-- tangent delta. So let’s look at what A s does. We have it
there. A s is tan delta-- well, tan delta-- we had it in the middle of blackboard is beta over alpha
minus k, 1 minus i beta over alpha minus k, again.
So that’s how A s behaves in general. That’s fine, there’s no – at this moment there’s nothing
crazy about this. Because this is something you all agreed, nobody complained about this
formula. So A s is given by that formula-- that’s also legal math, so far. So we’ll have this. And
then let’s simplify it a little bit which is beta over alpha minus k minus i beta. So this still beta
over alpha minus i beta minus k.
So we usually would plot A s as a function of k. That’s what we’re trying to do, it’s a function of
k. And now here is the formula for A s as the function of k. And here is k. But let’s be daring
now and not say this is k, this is the complex k-plane. And yes, you work with real k, but that’s
because that has a direct physical interpretation. But maybe the complex plane has a more
subtle physical interpretation and that’s what they claim is happening here.
This quantity becomes infinite near the resonance. Here was the resonance, what you call the
resonance. But this becomes really infinite not at alpha-- for when k is equal to alpha, but
when k is equal to alpha minus i beta. Beta was supposed to be small for a resonance. So
here is minus i beta and here is this very unusual point. Where the scattering amplitude blows
up. It has what is in complex variables-- if you’ve taken 1806 it’s called a pole.
In a complex variable when you have a denominator that vanishes linearly we call it a pole.
Things blow up. So this carrying amplitude has a pole off the real axis. And interpretation is
correct. At this point, this function becomes infinite. And what is happening on the real line that
A s is becoming large is just the remnant of that infinity over here that is affecting the value of
this point. So in the complex plane you understand the function a little better. You see why it’s
becoming big and you can see also with a little [INAUDIBLE] why the phases shifting very fast
because you have this point. And that’s called the resonance. And this is the mathematically
precise way of searching for resonances.
If you want to search for resonances what you should do is you have your formula for delta as
a function of k. I mean, it’s a complicated formula, but now try to solve the equation tan delta
of this is equal to minus i because that’s what guarantees that you have a pole that indeed it
blows up at some value. That’s where A s blows up which we see directly here-- it’s this value.
Alpha minus i beta, so alpha minus i beta is a pole of A s.
And therefore, you must be happening when tangent of delta is equal to minus i. So you have
a very complicated formula maybe for tangent of delta. But set it equal to minus i and asked
mathematically to solve it. And a number will come-- k a equal 2.73 minus 0.003. And you will
know-- oh, that’s a resonance, it’s off the axis. And the real part is the value of alpha. And
since this is beta the closer to the axis – if you find more-- the more resonant it is. And by the
time it’s far from the axis, some people call it the resonance-- some people say, no that not the
resonance. It’s a matter of taste. But there are important things which are these poles. So I will
not give you exercises on that, but you may want to try it if you want to have some
entertainment with these things.
I want to say one more thing about this. And it’s the reason why this viewpoint is interesting, as
well. We already found that if we want to think of resonances more precisely. We can think of
them as just an equation. You solve for the equation, so that it gives you the resonance. And
this is the equation you must solve and you must admit complex k. But now you can say, look
actually you have e is equal to h squared, k squared, over 2m. And we have real k’s-- this is
the physical scattering solutions, complex k’s, also resonances.
How about imaginary k’s? If k is equal to i kappa-- kappa belonging to the real numbers-- then
the energy becomes minus h squared, kappa squared, over 2m and its less than zero and it
could represent bound states. So you’ll be then discovering solutions of real k representing
your waves. Now mathematically, you are led to resonances understood as poles in the
scattering amplitutde we did here. We see that k’s in the imaginary axis would represent
bound states. So the complex k-plane is very rich. It has room for your scattering solutions, it
has room for your resonance, it even has room for your bound states. They’re all there. That’s
why it’s a valuable extension. I have now proven for you that bound states correspond to
poles. It’s a simple calculation, and that I would assign it to you with a little bit of guidance. And
you will see that also for the case of bound states, you get a pole in the scattering amplitude,
and that will complete the interpretation of that.
Now people go a little further, actually, and they invent poles in this part and they’re called
anti-bound states. And you’ll say, what’s that? If you have a bound state you match a solution
to a pure decaying exponential for the [INAUDIBLE] region. In an anti-bound bound state you
match your solution to a pure increasing exponential. A pure one. Does that have an
interpretation? It actually does have interpretation. Some nuclear states are associated with
anti-bound states.
So the mathematical description-- the rich complex plane is ready for you if you just do
scattering amplitude k, resonances-- complex k. Normal bound states, imaginary k-- positive.
Anti-bounds is negative k. It’s a nice start.