Lecture 18: Scattering in 1D (cont.). Example. Levinson’s theorem.

L18.1 Incident packet and delay for reflection (18:52)

L18.2 Phase shift for a potential well (09:13)

L18.3 Excursion of the phase shift (15:16)

L18.4 Levinson’s theorem, part 1 (14:46)

L18.5 Levinson’s theorem, part 2 (09:30)

L18.1 Incident packet and delay for reflection (18:52)

MITOCW | watch?v=EJWG9-etPFw
PROFESSOR: I’ll begin by reviewing quickly what we did last time. We considered what are called finite range
potentials, in which over a distance R, in the x-axis, there’s a non-zero potential. So the
potential is some v of x for x between capital R and 0, is equal to 0 for x larger than capital R,
and it’s infinity for x negative. So there’s a wall at x equals 0. And there can be some potential,
but this is called a finite range potential, because nothing happens after distance R.
As usual, we considered scattering solutions, solutions that are unnormalizable with energies,
h squared k squared over 2m, for a particle with mass m. And if we had no potential, we wrote
the solution phi of x, the wave function, which was sine of kx. And we also wrote it as a
superposition of an incoming wave. Now, an incoming wave in this set up is a wave that
propagates from plus infinity towards 0. And a reflected wave is a wave that bounces back and
propagates towards more positive x.
So here we’ll write this as minus e to the minus ikx over 2i, plus e to the ikx over 2i. This is the
sine function rewritten in terms of exponential in such a way that here is the incoming wave.
Remember the time dependence is minus iet over h bar. So this wave combined with a time is
a wave that is moving towards the origin. This wave is moving outwards.
Then we said that there would be, in general, with potential. With a potential, you would have a
solution psi effects, which we wrote after some tinkering in the farm i delta sine of kx plus
delta. And if you look at the part of the phase that has the minus ikx would have a minus delta
and a delta here. So they would cancel. So this solution has the same incoming wave as the
no potential solution.
On the other hand here, you would have e to the 2i delta, e to the ikx over 2i, and this solution
is only valid for x greater than R. You see, this is just a plane wave after all. There’s nothing
more than a plane wave and a phase shift. The phase shift, of course, doesn’t make the
solution any more complicated or subtle, but what it does is, by depending on the energy, this
phase shift delta depends on the energy, and we’re on k. Then, it produces interesting
phenomena when you send in wave packets.
So if we write psi, we usually write psi is equal to the phi plus psi s, where psi s is called the
scattered wave. You see, the full wave that you get, for x greater than R, we would have to
solve and work very carefully to figure out what is the wave function in the region 0 to R. But
for x greater than R is simple, and for x greater than R the wave function psi is the free wave
function, in the case of no potential, plus the scattered wave.
Quick calculation with this, things [? give to ?] you the scatter wave is e to the i delta sine delta
e to the ikx is an outgoing wave. And this coefficient is called the scattering amplitude. It’s the
amplitude of the scattered wave. This is a wave that is going out, and this is its amplitude. So it
has something to do with the strength of the scattering, because if there was no scattering, the
wave function would just behave like the no potential wave function.
But due to the potential, there is an extra piece, and that represents an outgoing wave beyond
what you get outgoing with a free no potential wave function. So it’s the scattering amplitude,
and therefore sometimes we are interested in as squared, which is just sine squared delta.
Anyway, those are the things we did last time. And we can connect to some ideas that we
were talking about in the past, having to do with time delays, by constructing a wave packet.
That’s what’s usually done.
Consider the process of time delay, which is a phenomenon that we’ve observed happens in
several circumstances. If you have an incident wave, how do you construct an incident wave?
Well, it has to be a superposition of e to the minus ikx, for sure. So we’ll put the function in
front, we’ll integrate over k, and we’ll go from 0 to infinity. I will actually add the time
dependence as well.
So let’s do phi of x and t. Then, we would have e to the minus i, e of k, t over h bar, and this
would be valid for x greater than R. Again, as a solution of the Schrodinger equation. You see,
it’s a free wave. There’s nothing extra from what you know from the de Broglie waves we
started a long time ago.
So if this is your incident wave, you have to now realize that you have this equation over here
telling you about the general solution of the Schrodinger equation. The general solution of the
Schrodinger equation, in this simple region, the outside region, is of this form, and it depends
on this delta that must be calculated. This is the incoming wave, this is the reflected wave, and
this is a solution. So by superposition, I construct the reflected wave of x and t.
So for each e to the minus ikx wave, I must put down one e to the ikx, but I must also put an e
to the 2i delta of the energy, or delta of k. And I must put an extra minus sign, because these
two have opposite signs, so I should put a minus 0 to infinity dk f of k. And we’ll have the e to
the minus i, e of k, e over h bar. And just for reference, f of k is some real function that picks at
some value k naught.
So you see, just like what we did in the case of the step potential, in which we had an incident
wave, a reflected wave packet, a transmitted wave packet, the wave packets go along with the
basic solution. The basic solution had coefficients A, B, and C, and you knew what B was in
terms of A and C. Therefore, you constructed the incoming wave with A e to ikx, and then the
reflected wave with B e to the minus ikx. The same thing we’re doing here inspired by this
solution, the psi affects we superpose many of those, and that’s what we’ve done here.
Now of course, we can do the stationary phase calculations that we’ve done several times to
figure out how the peak of the wave packet moves. So a stationary phase at k equal k naught.
As you remember, the only contribution can really come when k is near k naught, and at that
point, you want the phase to be stationary as a function of k. I will not do here the computation
again for psi incident. You’ve done this computation a few times already.
For psi incident, you find the relation between x and t, and I will just write it. It’s simple. You
find that x is equal to minus h bar k naught over mt, or minus some v velocity, group velocity,
times t. That is the condition for a peak to exist. The peak satisfies that equation, and this
makes sense when t is negative. This solution for psi incident only makes sense for x positive
if in fact x greater than R. So this solution needs x positive. So it needs t negative [? indeed. ?]
This is a wave that is coming from plus infinity, x equal plus infinity, at time minus infinity, and
it’s going in with this velocity.
For psi reflected, the derivative now has to take the derivative of delta, with respect to e, and
then the derivative of e with respect to the energy. And the answer, in this case-- you’ve done
this before-- it’s v group times t minus 2 h bar delta prime of E. So yes, in the reflected wave, x
grows as t grows and it’s positive. t must now be positive, but in fact, if you would have a just x
equal v group t, this would correspond to a particle that seems to start at the origin at time
equals 0 and goes out. But this actually there is an extra term subtracted. So only for t greater
than this number the particle begins to appear.
So this is a delay, t minus some t naught, the packet gets delayed by this potential. Now, this
delay can really get delayed. Sometimes it might even accelerated, but in general, the delay is
given by this quantity So I’ll write it here. The delay, delta t, is 2 h bar delta prime of E. And
let’s write it in a way that you can see maybe the units better and get a little intuition about
what this computation gives. For that, let’s differentiate this with respect to k, and then k with
respect to energy.
So v delta with respect to k, and dk with respect to energy. This is 2 over 1 over h bar dE with
respect to k. I do a little rearrangement of this derivative is one function of one variable k and
neither is a single relation. So you can just invert it. This is more dangerous when you have
partial derivatives. This is not necessarily true but for this ordinary derivatives is true, and then
you have this 2 to the left here.
The h bar went all the way down, and I have d delta dk. And here, we recognize that this is 2,
and this is nothing else than the group velocity we were talking before. The E, the energy, is h
squared k squared over 2m. You differentiate, divide by h bar, and it gives you the group
velocity hk naught over m. Because these derivatives all have to be evaluated at k naught. So
this derivative is really evaluated at k naught. This is also evaluated at k naught.
So this is the group velocity, d delta, dk, and finally, let me rewrite it in a slightly different way. I
multiply by 1 over R. Why? Because d delta dk, k has units of 1 over length. So if I multiply by
1 over R, this will have no units. So I claim that one over R d delta dk is equal to delta t, and
you’ll have 2 over vg and R. So I did a few steps. I moved the 2 over vg down to the left, and I
multiplied by 1 over R, and now we have a nice expression.
This is the delay. Delta t is the delay, but you now have divided it by 2R divided by the velocity,
which is the time it takes the particle with the group velocity to travel back and forth in the finite
range potential. So that gives you an idea. So if you compute the time delay, again, it will have
units of microseconds, and you may not know if that’s little or much. But here, by computing
this quantity, not exactly delta prime of v but this quantity. You get an [? insight, ?] because
this is the delay divided by the free transit time. It’s kind of a nice quantity. You’re dividing your
delay and comparing it with the time that it takes a particle, with a velocity that is coming in, to
do the bouncing across the finite range potential.

L18.2 Phase shift for a potential well (09:13)

MITOCW | watch?v=sWmY5KME7oo
PROFESSOR: So let’s do an example where we can calculate from the beginning to the end everything. Now,
you have to get accustomed to the idea of even though you can calculate everything, your
formulas that you get sometimes are a little big. And you look at them and they may not tell
you too much unless you plot them with a computer.
So we push the calculations to some degree, and then at some point, we decide to plot things
using a computer and get some insight on what’s happening. So here’s the example. We have
a potential up to distance, a, to 0. The wall is always there, and this number is minus v naught.
So it’s a well, a potential well. And we are producing energy. Eigenstates are coming here. And
the question now is to really calculate the solution so that we can really calculate the phase
shift. We know how the solutions should read, but unless you do a real calculation, you cannot
get the phase shift. So that’s what we want to do.
So for that, we have to solve the Schrodinger equation. Psi of x is equal to what? Well, there’s
a discontinuity. So we probably have to write an answer in which we’ll have a solution in one
piece and a solution in the other piece.
But then we say, oh, we wrote the solution in the outside piece already. It is known. It’s always
the same. It’s universal. I don’t have to think. I just write this. E to the i delta. I don’t know what
delta is, but that’s the answer, E to the i delta sine kx plus delta should be the solution for x
greater than a.
You know if you were not using that answer, it has all the relevant information for the problem,
time delays, everything, you would simply write some superposition of E to the i kx and E to
the minus i kx with two coefficients. On the other hand, here, we will have, again, a wave.
Now, it could be maybe an E to the i kx or E to the minus i kx. Neither one is very good
because the wave function must vanish at x equals 0. And in fact, the k that represents the
kinetic energy here, k is always related to E by the standard quantity, k squared equal to mE
over h squared or E equal the famous formula.
On the other hand, there is a different k here because you have different kinetic energy. There
must be a k prime here, which is 2m E plus v naught. That’s a total kinetic energy over h
squared. And yes, the solutions could be E to the ik prime x equal to minus ik prime x minus ik
prime x, but since they must vanish at 0, should be a sine function.
So the only thing we can have here is a sine of k prime x for x less than a and a coefficient.
We didn’t put the additional normalization here. We don’t want to put that, but then we must
put the number here, so I’ll put it here. That’s the answer, and that’s k and k prime.
Now we have boundary conditions that x equals a. So psi continues at x equals a. What does it
give you? It gives you a sine of ka is equal to E to the i delta sine of ka plus delta. And psi
prime continues at x equals a will give me ka cosine ka equal-- I have primes missing; I’m
sorry, primes-- equals k E to the i delta cosine ka plus delta.
What do we care for? Basically we care for delta. That’s what we want to find out because
delta tells us all about the physics of the scattering. It tells us about the scattering amplitude,
sine squared delta. It tells us about the time delay, and let’s calculate it.
Well, one way to calculate it is to take a ratio of these two equations so that you get rid of the a
constant. So from that side of the equation, you get k cotangent of ka plus delta is equal to k
prime cotangent of k prime a. Or cotangent of ka plus delta is k prime over k. We’ll erase this.
And now you can do two things. You can display some trigonometric wizardry, or you say, OK,
delta is arc cotangent of this minus ka. That is OK, but it’s not ideal. It’s better to do a little bit
of trigonometric identities. And the identity that is relevant is the identity for cotangent of a plus
b is cot a cot b minus 1 over cot a plus cot b.
So from here, you have that this expression is cot ka cot delta minus 1 over cot ka plus cot
delta. And now, equating left-hand side to this right-hand side, you can solve for cotangent of
delta. So cotangent of delta can be solved for-- and here is the answer. Cot delta is equal to
tan ka plus k prime over k cot k prime a over 1 minus k prime over k cot k prime a tan ka.
Now, who would box such a complicated equation? Well, it can’t be simplified any more. Sorry.
That’s the best we can do.

L18.3 Excursion of the phase shift (15:16)

MITOCW | watch?v=Cb_3sOYLjUI
That’s a solution. It’s an accomplishment to have such a solution. If somebody gives you a value of the energy,
you can calculate what is the phase shift, but we probably want to do more with it.
So you decide to plot this on a computer. Again, there’s lots of variables going on here, so you would want to
figure out what are the right variables to plot this.
And the right variables suggest themselves. From k squared equal 2 me over h squared, unit less constant are
things like ka, k prime a, and that’s it.
Well, so ka is a proxy for the energies. OK, a squared is really 2me, a squared over h bar squared. And so this we
could call anything.
Well, let’s call it u. On the other hand, k prime squared then-- if you have k prime a squared that it’s also unit free
would be 2me a squared over h squared plus 2mv0 a squared over h squared.
You probably recognize them. The first one is just u squared. I should call this u squared, sorry. U squared, and
this is our friend z0 squared. It’s that number that tells you the main thing you always want to know about a square
well.
That ratio between the energy v0 to the demand to the energy that you can build with h bar m and a. So here we
go. We have k prime a given by this quantity, and therefore let me manipulate this equation.
Might as well do it. It probably easier to consider just tan delta, which is the inverse of this. You would have 1
minus the inverse of this would be k prime a over ka, put the a’s always, so cot k prime a tan ka over tan ka plus k
prime aka cot k prime a.
So in terms of our variables, see k prime a is the square root of this, so k prime a square root of u squared plus z0
squared, and k prime a over ka, you divide now by u. So it’s square root of 1 plus z0 squared over u squared.
That’s this quantity. So how big, how much space do I need to write it? Probably, I should write it here.
1 minus square root of 1 plus z0 squared over u squared cot k prime a is the square root of z0 squared plus u
squared and tan of k a, which is u over tan u plus square root of 1 plus z0 over u squared cotangent of square
root of z0 squared plus u squared. OK, it’s not terrible. That’s tan delta.
So if somebody gives you a potential, you calculate what z0 is for this potential, you put z0 there, and you plot as
a function of u with Mathematica. And plotting as a function of u is plotting as a function of ka. And that’s perfectly
nice thing to do. And it can be done with this expression.
In this expression, you can also see what goes on when u goes to 0. Not immediately, it takes a little bit of
thinking, but look at it. As u goes to 0, well, these numbers are 1, that’s perfectly OK. That seems to diverge, goes
like 1 over u, but u going to 0. This goes to 0. So the product goes to a number.
So the whole-- the numerator goes to a number, some finite number. On the other hand, when u goes to 0, the
denominator will go to infinity, because while this term goes to 0 the tan u, this number is finite. And here you have
a 1/u. So the denominator goes to infinity. And the numerator remains finite. So as u goes to 0, tangent of delta
goes to zero.
So you can choose delta to be 0 for 0 energy. So as u goes to 0, you get finite divided by infinity, and goes to
zero. So tan delta goes to 0. And we can take delta of ka equals 0, which is u to be 0. The phase shift is 0 for 0
energy.
Let me go here. So here is an example. z0 squared equal 3.4. That actually correspond to 0.59pi for z0. z0 equal
0.59pi. You may wonder why we do that, but let me tell you in a second.
So here are a couple of plots that occur. So here is u equals ka. And here’s the phase shift, delta of u. You have
the tangent of delta, but the phase shift can be calculated. And what you find is that, yes, it starts at 0, as we
mentioned. And then it starts going down, but it stabilizes at minus pi, which is a neat number. That’s what the
phase shift does.
The so-called scattering amplitude, well you could say, when is this scattering strongest? When you get an extra
wave of this propagating more strongly? So you must plot sine squared delta and sine squared is highest for
minus pi over 2. So this goes like this, up, and decays as a function of u.
Third thing, the delay, is 1/a. The delay is 1/a d delta dk, as a function of u. So that, you can imagine, that takes a
bit of time, because you would have to find the derivative of delta with respect to u, and do all kinds of operations.
Don’t worry, you will have a bit of exercises on this to do it yourselves.
But here the delay turns out to be negative. And this is unit-free. And here, comes to be equal minus 4 for equals
0, and goes down to 0.
So in this case, the delay is negative. So the reflected packet comes earlier than you would expected, which is
possible, because the reflected packet is going slowly here. Finally, at this point, reaches more kinetic energy,
just-- and then back.
So that’s the delay. And you can plot another thing. Actually it’s kind of interesting, is the quantity a, this coefficient
here. That gives you an idea of how big the wave function is in the well. How much does it stick near the well?
So it peaks to 1. And it actually goes like this, and that’s the behavior of this form. Basically, it does those things.
So, so far so good. We got some information.
And then you do a little experiment, and try, for example, z0 equals 5. And you have delta as a function of u, and
here is minus pi, minus 2pi. And actually, you find that it just goes down, and approaches now minus 2pi.
So actually, if you increase this z0 a bit, it still goes to pi, a pi excursion of the phase. But suddenly, at some value,
it jumps. And it now goes to 2pi. And if you do with a larger value, at some point it goes to 3pi and 4pi. And it goes
on like that.
Well if z0 would have been smaller, like half of this, the phase would go down and would go back up, wouldn’t go
to minus pi. It does funny things. So what’s really happening is that there is a relation between how much the
phase moves, and how many bound states this potential has.
And you say, why in the world? This calculation had nothing to do with bound states. Why would the phase shift
know about the bound states? Well actually, it does. And here is the thing. If you remember, you’ve actually solved
this problem in homework, the half square well, in which you put an infinite wall here.
And if you had the full square well, from minus a to a, this problem has all the old solutions of the full square well.
All the old solutions exist. And if you remember the plots that you would do in order to find solutions, you have
pi/2, pi, 3pi/2, 2pi. And here is the even solution. Here is the odd solution. I’ll do it like that. Here is an even
solution. Here is an odd solution.
And I marked the odd solutions, because we care about the odd ones, because that’s what this potential has. So
z0 equals 0.59pi is a little more than pi/2. So it corresponds to one solution. So there is one bound state for this
z0. z0 equals 5 is about here. it’s in between 3pi/2 and this. And there’s two nodes, two intersections. Therefore,
two solutions in the square well. And here we have that the phase has an excursion of, not just pi for one, but 2pi.
And if you did this experiment for awhile, you would convince yourself there’s a magic relation between how much
the phase shift moves, and how many bound states you have in this potential. This relation is called Levinson’s
theorem. And that’s what we’re going to prove in the last half an hour of this lecture.

L18.4 Levinson’s theorem, part 1 (14:46)

MITOCW | watch?v=GyukKStk6Ls
PROFESSOR: Levinson’s theorem, in terms of derivations, that we do in this course, this is probably the most
subtle derivation of the semester. It’s not difficult, but it’s kind of interesting and a little subtle.
And it’s curious, because it relates to things that seem to be fairly unrelated. But the key thing
that one has to do is you have to use something. How all of the sudden are you going to relate
phase shifts to bound states?
The one thing you have to imagine is that if you have a potential and you have states of a
potential, if the potential changes, the states change. But here comes something very nice.
They never appear or disappear. And this is something probably you haven’t thought about
this all that much. Because you had, for example, a square, finite square well.
If you made this more deep, you’ve got more bound states. If you made it less deep, the
bound states disappear. What does it mean the bound states disappear? Nothing, really can
disappear.
What really is happening, if you have the bounds, the square well. You have a couple of bound
states, say. But then you have an infinity of scatterings states. And as you make this potential
less shallow, the last bound state is approaching here. And at one state, it changes identity
and becomes a scatterings state, but it never gets lost.
And how, when you make this deeper and deeper you get more state, is the scattering state
suddenly borrowing, lending you a state that goes down? The states never get lost or
disappear. And you will, say, how could you demonstrate that? That sounds like science
fiction, because, well, there’s infinitely many states here. How do you know it borrowed one?
Well, you can do it by putting it in a very large box. And then the states here are going to be
finitely countable and discrete. And then you can track and see indeed how the states become
bound. So you’d never lose or gain states. And that’s a very, very powerful statement about
quantum states in a system.
So this is what we’re going to need to prove Levinson’s theorem. So Levinson theorem
theorem-- so it relates a number of bound states of the potential to the excursion of the phase
shift. So let’s state it completely. It relates the number N of bound states of the potential to the
excursion, excursion of the phase from E equals 0 to E equals infinity.
So in other words, it says that N is 1 over pi delta of 0 minus delta of infinity, a number of
bound states of your potential is predicted by the behavior of the phase shift of scattering. So
how do we do this? This is what we want to prove.
So consider, again, our potential of range R and 0 here and here is x. And I want you to be
able to count states, but discovering states are a continuous set of states. So in order to count
states, we’re going to put a wall here, as well, at some big distance L much bigger than A than
R.
And we’re going, therefore-- now the states are going to be quantized. They’re never going be
quite scatterings states. They’re going to look like scatterings states. But they’re precisely in
the way that they vanished at this point.
Now you would say, OK, that’s already a little dangerous to be. Oh, you’ve changed the
problem a little, yes. But we’re going to do the analysis and see if the result depends on L. If it
doesn’t depend on L and L is very large, we’ll take the limit this L goes to infinity. And we
claim, we have answer.
So we argue that L is a regulator, regulator to avoid a continuum, continuum of states to avoid
that continuum of states. All right, so let’s begin to count. To count this thing, we start with the
case where there is no potential again. And why is that? Because we’re going to try to
compare the situation with no potential to the situation with potential.
So imagine let V identical is 0, no potential and consider positive energy states. These are the
only states that exist. There are no bound states, because the potential is 0.
Well, the solutions were found before, we mentioned that these are what we call phi effects or
sine of kx. But now we require that phi of L is equal to 0, because we do have the wall there.
And therefore, si of kl must be 0 and kl must be n pi and n is 1, 2 to infinity.
You know we manage with the wall to discretize this state, because the whole world is now an
infinite, a very big box, not infinite, but very big. So you’ve discrete the state. The separations
are microelectron volts, but they are discrete. You can count them.
And then with this state over here, we think of counting them. And you say, well, I can count
them with n. So if I imagine the k line from 0 to 50, the other states are over, all the values of
k. And they could say, well, I even want to see how many states there are in a little element dk.
And for the that you would have that dk taken a differential here is dn times pi. So of the
number of states that there are in dk, dn-- let me right here-- dn equals l over pi dk is the
number of positive energy states in dk. In the range dk, in the range of momentum, dk that
little interval, there are that many positive energy eigenstates. So far, so good.
So now let’s consider the real case. Repeat for the case there is some potential. Well, you
would say, well, I don’t know how to count. I have to solve the problem of when the potential
makes a difference. But no, you do know how to count. So repeat for V different from 0.
This time we have a solution for x greater than R. We know the solution. This is our universal
solution with the phase. So there you have the si effect is e to i delta sine kx plus delta of k.
That’s a solution.
Yes, you have the solution always. You just don’t know what delta is. But you don’t know what
delta-- you don’t need to know what delta is to prove the theorem, You just need to know it’s
there.
So here it goes. And this time the wall will also do the same thing. We’ll demand the si of x
vanishes for x equal L. So this time we get that kx-- no kL plus delta of k is equal to some
other number n prime times pi multiple of pi. This phase-- this total phase has to be a multiple
of pi.
And what is n prime? I don’t know what is n prime? It is some integers. I don’t know whether it
starts from 1, 2, 3 or from 100 or whatever.
The only thing we care is that, again, taking a little differential, you have dk times L plus d
delta, dk, times dk is equal to the dn prime times pi. We take an infinitesimal version of this
equation, which again tells me how many positive-- all these states are positive energy states.
They have k.
So all these are positive energy states. So from this equation, we get that dn prime is equal to
L over pi dk plus 1 over pi d delta dk times dk, which gives me if I know the dk, again, the
number of states that you have in that range of k, You see it’s like momentum is now
quantized.
So for any little range of momentum, you can tell how many states there are. And here it is
how many states there are, positive energy states, positive. This is the number of positive
energy states in dk with V different from 0, here is with V equal 0.

L18.5 Levinson’s theorem, part 2 (09:30)

MITOCW | watch?v=yhI3jTX4dY4
PROFESSOR: We have two equations now relating this number of states. And now you can say, oh, OK, so I
look at the k line. And I look at the little piece of the k and say, oh, how many states were there
with 0 potential, some number, first blackboard. How many states are there now with some
potential, some other number? It has changed.
For every-- because these two equations, the n for equal dk, the n is not equal to the dn
prime. In one case, the energy levels or the momentum levels are more compressed or more
separated, but whatever it is, whatever the sign is, there is a little discrepancy. So both of them
are giving me the total number of positive energy states in the little dk. Case So if I take the
difference, I will get some information.
So I would say the following, if I want to calculate the number of positive energy solutions and
now I think the following, I take the potential V equals 0 and slowly but surely deepen it, push
it, and do things and create the potential V of x slowly, slow the formation. In this process, I
can look at a little interval dk and tell how many states are positive energy states I lost. So if,
for example, dn is bigger than dn prime, dn equal 5 and dn prime is equal to 3, I started with 5
positive energy states in this little interval and by the time I change the potential I ended up
with 3. So I lost 2.
So let me write here the number of positive energy solutions lost in the interval dk as the
potential is turned on is dn the original number minus the dn prime. If that’s positive, I’ve lost
state. If it’s negative, I gained state, positive energy states. In this number, we can calculate
the difference. This is minus 1 over pi d delta dk dk. I’ll put it here.
We’ll we’re not far. We’ll this is what you lost. The number of positive energy eigenstates that
you lost in little dk. To see how many positive energy states you lost over all, you must
integrate over all the dk’s and see how much you lost in every little piece.
So the number of positive energy solutions lost, not in the dk, but lost as the potential is turned
on is equal to the integral over k from 0 to infinity of minus 1 over pi d delta dk is in the way of
that expression of that right coincide. But this is a total derivative. So this is minus 1 over pi
delta of k evaluated between infinity and 0. And therefore, the number of states lost is 1 over
pi, because of the sign down to 0 minus delta of infinity.
So we’re almost there. This is the number of positive energy solutions lost. Now I want to
emphasize that the situation is quite interesting. Let me make a little drawing here.
So suppose here is the case where you have the potential equal to 0 and here is energy equal
to 0. Then you have all these states. Now even though we’ve put the wall, the wall allows us to
count the states, but there are still going to be an infinite number of states. The infinite square
wall has an infinite number of states.
So that thing really continues, but what happens by the time v is deferred from 0? Here is that
the E equals 0 line and here is the E equals 0 line. As we’ve discussed, as you change the
potential slowly, this are going to shift a little and some are going to go down here, are going
to become bound states. They’re going to be a number of bound states, N bound states,
number of bound states equal N. And then there’s going to be still sub states here that’s also
go to infinity.
So you cannot quite say so easily, well, the number of states here minus the number of states
here is the number lost. That’s not true, because that’s infinite, that’s infinite, and subtracting
infinity is bad. But you know that you’ve lost a number of finite number of positive energy
solutions.
So as you track here, the number of states must-- the states must go into each other. And
therefore, if these four states are now here, before they were here, and those were the
positive energy solutions that were lost, in going from here to here, you lost positive energy
solutions. You lost a finite number of positive energy solutions. Even though there’s infinite
here and infinite here, you lost some. And you did that by keeping track at any place how
much you lost.
And therefore the states lost are never really lost. They are the ones that became the band
states here. So the positive energy states that got lost are the bound states. So the number
bound states is equal to the number of positive energy solutions, because there are no lost
states. So this is equal to a number of bound states, because there are overall no lost states.