Lecture 13: Delta function potential. Justifying the node theorem. Simple harmonic oscillator.

L13.1 Delta function potential I: Preliminaries (16:14)

L13.2 Delta function potential I: Solving for the bound state (15:21)

L13.3 Node Theorem (13:01)

L13.4 Harmonic oscillator: Differential equation (16:45)

L13.5 Behavior of the differential equation (10:31)

L13.1 Delta function potential I: Preliminaries (16:14)

MITOCW | watch?v=vcuY46RwoV0
PROFESSOR: Delta function potential. So it’s still a one-dimensional potential-- potential is a function of x.
We’ll write it this way-- minus alpha delta of x, where alpha is positive. So this is a delta
function in a negative direction.
So if you want to draw the potential-- there’s no way to draw really nicely a delta function. So
you just do a thick arrow with it pointing down. It’s a representation of a potential that,
somehow, is rather infinite at x equals zero-- but infinite and negative.
It can be thought of as the limit of a square well that is becoming deeper and deeper. And, in
fact, that could be a way to analytically calculate the energy levels-- by taking carefully the limit
of a potential. It is becoming thinner and thinner, but deeper and deeper, which is the way you
define or regulate the delta function.
You can imagine the delta function as a sequence of functions, in which it’s becoming more
and more narrow-- but deeper at the same time. So that the area under the curve is still the
same.
So, at any rate, the delta function potential is a potential that should be understood as 0
everywhere , else except at the delta function where it becomes infinite. And there are all kinds
of questions we can ask.
OK. Are there bound states? What are bound states in this case? They are energy eigenstates
with energy less than zero. So bound states, which means e less than zero. Do they exist?
Does this potential have bound states?
And, if it does, how many bound states? 1, 2, 3? Does It depend on the intensity of the delta
function? When you get more bound states, the deeper the potential is. Well, we’ll try to figure
out.
In fact, there’s a lot that can be figured out without calculating, too much. And it’s a good habit
to try to do those things before you-- not to be so impatient that you begin, and within a
second start writing the differential equation trying to solve it. Get a little intuition about how
any state could look like, and how could the answer for the energy eigenstates-- the energies–
what could they be?
Could you just reason your way and conclude there’s no bound states? Or one bound state?
Could you just reason your way and conclude there’s no bound states? Or one bound state?
Or two? All these things are pretty useful. So one way, as you can imagine, is to think of units.
And what are the constants in this problem? In this problem we’ll have three constants. Alpha,
the mass of the particle, and h-bar. So with alpha, the mass and the particle, and h-bar you
can ask, how do I construct the quantity with units of energy?
If there there’s only one way to construct the quantity with units of energy, then the energy of a
bound state will be proportional to that quantity-- because that’s the only quantity that can
carry the units.
And here, indeed, there’s only one way to construct that quantity with units of energy-- from
these three. That’s to be expected. With three constants that are not linearly dependent–
whatever that is supposed to mean-- you can build anything that has units of length, mass, or
time. And from that you can build something that has units of energy.
So you can now decide, well, what are the units of alpha? The units of alpha have give you
energy, but the delta function has units of one over length. This has one over length. ,
Remember if you integrate over x the delta function gives you 1. So this has units of 1 over
length. And, therefore, alpha has to have units of energy times length.
So this is not quite enough to solve the problem, because I want to write e-- think of finding
how do you get units of energy from these quantities? But l-- we still don’t have a length scale
either. So we have to do a little more work.
So from here we say that units of energy is alpha over l. There should be a way to say that this
is an equality between units. I could put units or leave it just like that. So in terms of units, it’s
this. But in terms of units, energy-- you should always remember-- is p squared over m. And p
is h-bar over a length. So that’s p squared and that’s m. So that’s also units of energy
From these two you can get what has units of length. Length. You pass the l to this side-- the l
squared to this left-hand side. Divide. So you get l is h squared over m alpha. And if I
substitute back into this l here, e would be alpha over l, which is h squared, alpha squared, m.
So that’s the quantity that has units of energy. M alpha squared over h squared has units of
energy. If this has units of energy-- the bound state energy. Now, if you have a bounce state
here, it has to decay in order to be normalizable. In order to be normalizable it has to decay,
so it has to be in the forbidden region throughout x.
So the energy as we said is negative, energy of a bound state-- if it exists. And this bound
state energy would have to be negative some number m alpha squared over h squared. And
that’s very useful information.
The whole problem has been reduced to calculating a number. It better be and the answer
cannot be any other way. There’s no other way to get the units of energy. So if a bound state
exists it has to be that. And that number could be pi, it could be 1/3, 1/4, it could be anything.
There’s a naturalness to that problem in that you don’t expect that number to be a trillion. Nor
do you expect that number to be 10 to the minus 6. Because there’s no way-- where would
those numbers appear? So this number should be a number of order one, and we’re going to
wait and see what it is.
So that’s one thing we know already about this problem. The other thing we can do is to think
of the regulated delta function. So we think of this as a potential that has this form. So here is v
of x, and here is x. And for this potential-- if you have a bound state-- how would the wave
function look?
Well, it would have to-- suppose you have a ground state-- it’s an even potential. The delta
function is even, too. It’s in the middle. It’s symmetric. There’s nothing asymmetric about the
delta function. So if it’s an even potential the ground states should be even, because the
ground state is supposed to have no nodes. And it’s supposed to be even if the potential is
even.
So how will it look? Well, it shouldn’t be decaying in this region. So, presumably, it decays
here. It decays there-- symmetrically. And in the middle it curves in the other direction. It is in
an allowed region-- and you remember that’s kind of allowed this way. So that’s probably the
way it looks.
Now, if that bound state exists, somehow, as I narrow this and go down-- as it becomes even
more narrow, very narrow now, but very deep. This region becomes smaller. And I would
pretty much expect the wave function to have a discontinuity.
You basically don’t have enough power to see the curving that is happening here. Especially
because the curving is going down. The distance is going down. So if this bound state exists,
as you approach the limit in which this becomes a delta function the energy moves a little, but
stays finite at some number. And the curvature that is created by the delta function is not
visible, and the thing looks just discontinuous in its derivative.
So this is an intuitive way to understand that the wave function we’re looking for is going to be
discontinuous on its derivative. Let’s write the differential equation, even though we’re still not
going to solve it. So what is the differential equation? Minus h squared over m, psi double
prime, is equal to E psi.
And, therefore-- and I write this, and you say, oh, what are you writing? I’m writing the
differential equation when x is different from 0. No potential when x is different from 0. So this
applies for positive x and negative x. It doesn’t apply at x equals 0. We’ll have to deal with that
later.
So then, no potential for x different from 0. And this differential equation becomes psi double
prime equals minus 2m e over h squared psi. And this is equal to kappa squared psi, where
kappa squared is minus 2me over h squared. And it’s positive. Let’s make that positive.
It’s positive because the energy is negative and we’re looking for bound states. So we’re
looking for bound states only. Kappa squared is positive. And this differential equation is just
this. I’ll copy it again here. Kappa squared psi.
And the solutions of this equation are-- solutions-- are e to the minus kappa x and e to the
kappa x. Or, if you wish, cosh kappa x and sinh kappa x-- whichever you prefer.
This is something we now have to use in order to produce a solution. But now, let’s see if I can
figure out how many bound states there are. If there is one bound state, it’s going to be even.
It’s the ground state. It has no nodes. It has to be even, because the potential is even. If I have
the first excited state after the ground state, it will have to be odd. It would have to vanish at x
equals 0, because it’s odd. There is it’s node-- it has to have one node.
For an odd bound state-- or first excited state-- you’d have to have psi equals 0 at x equals 0.
And the way to do that would be to have a sinh, because this doesn’t vanish at zero. This
doesn’t vanish at zero. And cosh doesn’t vanish at zero.
So you would need psi of x equals sinh of kappa x. But that’s not good. sinh of kappa x is like
this and blows up. Blows down. It has to go like this. It is in a forbidden region, so it has to be
convex towards the axis. And convex here. But it blows up. So there’s no such solution. No
such solution.
You cannot have an odd bound state. So since the bound states alternate-- even, odd, even,
odd, even, odd-- you’re stuck. You only will have a ground state-- if we’re lucky-- but no
excited state that is bound, while a finite square will.
You remember this quantity z0 that tells you how many bound states you can have. Probably
you’re anticipating that in the case of the delta function potential, you can only have one bound
state, if any. The first excited state would not exist. So, enough preliminaries. Let’s just solve
that now.

L13.2 Delta function potential I: Solving for the bound state (15:21)

MITOCW | watch?v=1dW_izzvfOk
PROFESSOR: So, what is the wave function that we have? We must have a wave function now that is
symmetric, and built with e to the k x, kappa x, and into the minus kappa x. This is the only
possibility. E to the minus of kappa absolute value of x. This is psi of x for x different from 0.
This is-- as you can quickly see-- this is e to get minus kappa x, when x is positive, A e to the
kappa x, when x is negative. And, both of them decay. The first exponential negative is the
standard decaying exponential to the right. The one with positive-- well, here x is negative as
you go all the way to the left. This one decays case as well.
And, this thing plotted is a decaying exponential with amplitude A, like that. And, a decaying
exponential with amplitude A, and a singularity there, which is what you would have expected.
So, this seems to be on the right track-- it’s a continuous wave function. The wave function
cannot fail to be continuous, that’s a complete disaster to show that an equation could not be
satisfied. So, this is our discontinuous wave function.
So, at this moment you really haven’t yet used the delta function-- the delta function with
intensity alpha down. I should have made a comment that it’s very nice that alpha appeared
here in the numerator. If it would have appeared in the denominator, I would be telling you that
I think this problem is not going to have a solution. Why? Because if it appears in the
numerator, it means that as the delta function potential is becoming stronger and stronger, the
bound state is getting deeper and deeper-- which is what you would expect. But, if it would be
in the numerator-- in the denominator-- as the potential gets deeper and deeper, the boundary
is going up. That makes no sense whatsoever. So, it’s good that it appeared there, it’s a sign
that things are in reasonable conditions.
So, now we really have to face the delta function. And, this is a procedure you are going to do
many times in this course. So, look at it, and do it again and again until you’re very
comfortable with it. It’s the issue of discovering what kind of discontinuity you can have with the
delta function. And, it’s a discontinuity in the derivative, so let’s quantify it. So, here it is-- we
begin with the Schrodinger equation, again. But, I will write now the potential term as well. The
potential is plus v of x psi of x equals E psi of x.
And the idea is to integrate this equation from minus epsilon to epsilon. And, epsilon is
supposed to be a small positive number. So, you integrate from minus epsilon to epsilon the
differential equation, and see what it does to you in the limit as epsilon goes to 0. That’s what
we’re going to try to do. So, what do we get? If you integrate this, you get minus h squared
over 2 m. And now, you have to integrate the second derivative with respect to x, which is the
first derivative, and therefore this is the first derivative at x equal epsilon minus the first
derivative at x equals minus epsilon. This is from the first term, because you integrate d x d
second d x squared psi is the same thing as d x d d x of d psi d x between A and B. And, the
integral of a total derivative is d psi d x at B A-- I think people write it like this-- A to B. Evaluate
it at the top, minus the evaluation at the bottom.
Now, the next term is the integral of psi times v of x. So, I’ll write it plus the integral from minus
epsilon to epsilon d x minus alpha delta of x psi of x-- that’s the potential. Now, we use the
delta function. And, on the right hand side this will be E times the integral from minus epsilon
to epsilon of psi of x d x. So, that’s the differential equation integrated.
And now, we’re going to do two things. We’re going to do some of these integrals, and take
the limit as epsilon goes to 0. So, I’ll write this minus h squared over 2 m limit as epsilon goes
to 0 of d psi d x at epsilon minus d psi d x at minus epsilon plus. Let’s think of this integral. We
can do this integral, it’s a delta function. So, it picks the value of the wave function at 0,
because 0 is inside the interval of integration. That’s why we integrate it from minus epsilon to
epsilon, to have the delta function inside. So, you get an alpha out, a psi of 0, and that’s what
this integral is. It’s independent of the value of epsilon as long as epsilon is different from 0.
So, this gives you minus alpha psi of 0.
And now, the last term is an integral from minus epsilon to epsilon of the wave function. Now,
the wave function is continuous-- it should be continuous-- that means it’s finite. And, this
integral, as of any function that is not divergent from minus epsilon to epsilon as epsilon goes
to 0, is 0. Any integral of a function that doesn’t diverge as the limits of integration go to 0, the
area under the function is 0. So, this is 0-- the limit. And this thing goes to 0, so we put a 0
here.
So, at this moment we got really what we wanted. I’ll write it this way. I’ll go here, and I’ll say
minus h squared over 2 m, and what is this? This expression says, calculate the derivative of
the function a little bit to the right of 0, and subtract the derivative of the function a little bit to
the left of 0. This is nothing but the discontinuity in psi prime. You’re evaluating for any epsilon
greater than 0-- the psi prime a little to right, a little too the left, and taking the difference. So,
this is what we should call the discontinuity delta at 0-- at x equals 0. And, this and this is for
discontinuity of psi prime at x equals 0 minus alpha psi of 0 equals 0. And from here, we
discover that delta zero psi prime is equal to minus 2 m alpha over h squared psi of 0.
This is the discontinuity condition produced by the delta function. This whole quantity is what
we call delta 0 of psi prime. And, what it says is that yes, the wave function can have a
discontinuous first derivative if the wave function doesn’t vanish there. Once the wave function
doesn’t vanish at that point, the discontinuity is in fact even proportional to the value of the
wave function at that point. And, here are the constants of proportionality. Now, I don’t think it’s
worth to memorize this equation or anything like that, because it basically can be derived in a
few lines. This may have looked like an interesting or somewhat intricate derivation, but after
you’ve done is a couple of times-- this is something you’ll do in a minute or so. And, you just
integrate and find the discontinuity in the derivative-- that’s a formula there. And, that’s a
formula for a potential, minus alpha delta of x. So, if somebody gives you a different potential,
well, you have to change the alpha accordingly.
So, let’s wrap this up. So, we go to our case. Here is our situation. So, let’s apply this. So, what
is the value? Apply this equation to our wave function. So, what is the derivative at epsilon? It’s
minus kappa A E to the minus kappa epsilon. That’s the derivative of psi on the positive side. I
differentiated the top line of this equation minus the derivative on the left side-- this one, the
derivative. So, this is kappa A E to the kappa epsilon-- no, kappa minus epsilon again. So,
that’s the left hand side. The right hand side would be minus 2 m alpha h squared psi at 0. Psi
at 0 is A, so that’s what it gives us.
And we should take the limit as epsilon goes to 0. So, this is going to 1, both of them. So, the
left hand side is minus 2 kappa A, and the right hand side is 2 m alpha over h squared A. So,
the 2-- it’s also minus, I’m sorry-- so the 2s cancel, the A cancels-- you never should have
expected to determine A unless you tried to normalize the wave function. Solving for energy
eigenstates states will never determine A. The Schrodinger equation is linear, so A drops out,
the minus 2 drops out, and kappa is equal to m alpha over h squared.
So, that said that’s great because kappa is just another name for the energy. So, I have kappa
m alpha over h squared, so that’s another name for the energy. So, let’s go to the energy. The
energy is h bar squared kappa minus h bar squared kappa squared over 2 m. So, it’s minus h
bar squared. Kappa squared would be m squared alpha squared h to the fourth, and there’s a
two m. All these constants.
So, final answer. E, the bound state energy is minus m alpha squared minus m alpha squared.
The m cancels it over h squared minus one half. So, back here the units worked out,
everything is good, and the number was determined as minus one half. That’s your bound
state energy for this problem.
So, this problem is instructive because you basically learn that in delta functions, with one
delta function you get a bound state. If you have two delta functions, you may get more bound
states-- three, four-- people study those problems, and you will investigate the two delta
function cases.

L13.3 Node Theorem (13:01)

MITOCW | watch?v=NwPOhzDPHKc
BARTON
ZWIEBACH:
The next thing I want to talk about for a few minutes is about the node theorem. Theorem. And
it’s something we’ve seen before. We’ve heard that if you have a one-dimensional potential
and you have bound states, the ground state has no nodes. The first excited state has 1 node.
Second, 2, 3, 4.
All I want to do is give you a little intuition as to why this happens. So this will be an argument
that is not mathematically very rigorous, but it’s fairly intuitive and it captures the physics of the
problem. So it begins by making two observations.
So in the node theorem, if you have psi 1, psi 2, psi 3, all energy-- energy-- eigenstates of a
one-dimensional potential-- bound states. Bound states. With energy E1 less than E2, less
than E3 and E4, psi n has n minus 1 nodes. Those are points where the wave function
vanishes inside the range of x.
So for this square well, you’ve proven this by calculating all the energy eigenstates. The first
state is the ground state. It has no nodes. The next state is the first excited state. It has one
node. And you can write all of them, and we saw that each one has one more node than the
next.
Now I want to argue that in an arbitrary potential that has bound states, this is also true. So
why would that be true for an arbitrary potential? The argument we’re going to make is based
on continuity. Suppose you have a potential like this-- V of x-- and I want to argue that this
potential will have bound states and will have no node, 1 node, 2 nodes, 3 nodes. How could I
argue that?
Well, I would do the following. Here is the argument. Identify the minimum here. Call this x
equals 0.
Oh, I want to say one more thing and remind you of another fact that I’m going to use. So this
is the first thing, that the square well realizes this theorem, and the second is that psi of x0
being equal to psi prime at x0 being equal to 0 is not possible. The wave function and its
derivative cannot vanish at the same point.
Please see the notes about this. There is an explanation in last lecture’s notes. It is fact that
for a second order differential equation, psi and psi prime tell you how to start the solution, and
if both psi and psi prime are equal to 0, the general solution of the differential equation is
always 0 everywhere.
So this kind of thing doesn’t happen to a wave function-- the point where it’s 0 and the
derivative is 0. That never happens. This happens-- 0 wave function with the derivative. But
this, no. Never happens. So those two facts.
And now let’s do the following. Let’s invent a new potential. Not this potential, but a new one
that I’ll mark the point minus a here and the point a here and invent a new potential that is
infinite here, infinite there, and has this part I’ll write there.
So this will be called the screened potential. Screened potential. Va of x in which Va of x is
equal to V of x for x less than a, and it’s infinity for x greater than a. So that’s a potential in
which you turn your potential into an infinite square well whose bottom follows the potential. It’s
not flat.
And now, we intuitively argue that as I take a to infinity, the bound states of the screened
potentials become the bound states of your original potential. Because when the screen is
very, very, very far away, up to infinity, you’ve got all your potential, and by the time you have
bound states that are decaying, so the screen is not going to do much at infinity. And anyway,
you can move it even further away. If you move it one light year away or two light years away,
shouldn’t matter.
So the idea is that the bound states-- bound states-- of Va of x as a goes to infinity are the
bound states of V of x. And moreover, as you slowly increase the width of the screen, the
bound states evolve, but they evolve continuously. At no point a bound state blows up and
reappears or does something like that. It just goes continuously.
These are physically reasonable, but a mathematician would demand a better explanation. But
that’s OK. We’ll stick to this. So let’s continue there.
So here is the idea, simply stated. If a is going to 0, if the width of the screen is extremely
narrow, you’re sitting at the bottom of the potential at x equals 0. And the screened potential is
basically a very, very narrow thing, and here, there’s the bottom of the potential. And for
sufficiently small a-- since you picked the bottom of the potential there-- it’s basically flat.
And then I can use the states of the infinite square well potential. As a goes to 0, yes, you
have a ground state with no nodes, a first excited state with one node, and all the states have
the right number of nodes because they are the states of the infinite square well, however
narrow it is.
So the only thing we have to now show is that if you have a wave function-- say, let’s begin
with one with no nodes-- as you increase the width of the screen, you cannot get more nodes.
It’s impossible to change the number of nodes continuously. So here it is. I’m going to do a
little diagram.
So for example, let’s assume the screen is this big at this moment, that you have some ground
state like this. You’ve been growing this, and then as the screen grows bigger, you somehow
have maybe a node. Could this have happened? As you increase this screen, you get a node.
Now I made it on this point. I didn’t intend to do that, so let me do it again somewhere. Do you
get a node?
Well, here was the original screen, and here the derivative psi prime is negative. Psi prime is
negative. On the other hand, psi prime here is already positive.
So as you grew this screen, this PSI prime that was here must have turned from negative to
positive, the way it looks here. But for that, there must have been a point somewhere here
when it was horizontal if it’s continuous. And therefore, there must have been some point at
which psi and psi prime were both 0 at the endpoint x equals a, whatever the value of a was,
because psi prime here is positive, and here is negative.
So at some point it was 0, but since it’s at the point where you have the infinite square well, psi
is also 0. And you would have both psi and psi prime equal 0, which is impossible. So
basically, you can’t quite flip this and produce a node because you would have to flip here, and
you can’t do it.
One could try to make a very precise, rigorous argument, but if you have another possibility
that you might think, well, you have this wave function maybe. And then suddenly it starts
doing this, and at some stage, it’s going to try to do this. But before it does that, at some point,
it will have to be just like this and cross, but at this point, psi and psi prime would be 0.
So you can intuitively convince yourself that this thing doesn’t allow you to produce a node. So
if you start with whatever wave function that has no nodes, as you increase the screen, you
just can’t produce a node. So the ground state of the whole big potential will have no nodes.
And if you start with the first excited state that has one node, as you increase the screen, you
still keep one node. So the next state of the full potential will have one node as well. And that
way, you argue that all your bound states of the complete potential will just have the right
number of nodes, which is 0, 1, 2, 3, 4. And it all came, essentially, from the infinite square
well and continued.

L13.4 Harmonic oscillator: Differential equation (16:45)

MITOCW | watch?v=sxzFpOsvfgU
PROFESSOR: Simple harmonic oscillator.
So what is there about this simple harmonic oscillator? Well, it’s a classical system that you
understand perfectly well. An oscillator, a spring with a mass oscillates and has an energy,
which is the kinetic energy plus the potential energy, and that’s p squared over 2m plus 1/2 m
omega squared x squared. And this may be a tiny bit unfamiliar, this way of writing it. But you
may recall that omega is equal to the square root of k over m, the so-called spring constant, in
which the potential in terms of k would be 1/2 k x squared. And that’s the potential energy
stored in a spring that you stretch at distance x.
That’s the total energy of a harmonic oscillator. So when physicists starting with quantum
mechanics in the '20s decided, let’s do a harmonic oscillator, a quantum harmonic oscillator,
they had to invent the Hamiltonian. And the Hamiltonian they invented was a simple one. They
looked at that and said, h is going to be p hat squared over 2m plus 1/2 m omega squared x
hat squared. And now the difference is going to be that x with p are operators, and this is our h
bar, and that’s going to be my quantum system.
So this is a quantum system that is inspired by classical mechanics in the purest and simple
way. Anyone could have invented this quantum system. It was very natural.
Still, the result of the quantization is very surprising because while this mechanical oscillator
can oscillate with any amplitude, the quantum oscillator has quantized amplitudes and
quantized energies therefore. So all kinds of interesting things happen with this oscillator.
Now the reason this is also very ubiquitous is that this potential is exactly a quadratic potential
v of x. v of x is 1/2 m omega squared x squared. We have x without the hat. This is a good
approximation to almost any system we consider in nature, any oscillating system. Because for
any potential that has a minimum, there is some parabolic approximation at the bottom. At the
bottom the derivative vanishes, so the Taylor series says that approximately at the bottom is a
quadratic potential.
And therefore this quadratic potential will govern the quantum oscillations of a diatomic
molecule, the quantum oscillations of a periodic system, all kind of quantum oscillations will be
approximately governed by a harmonic oscillator. Light has a harmonic oscillator description
for its photons.
This Hamiltonian is the most famous Hamiltonian there is. In fact, when you have electrons in
a magnetic field, somehow this shows up. And this becomes some sort of problem that you
solve very well, understand very well, and suddenly it pops up in all kinds of contexts. So we
need to understand it.
And let’s go directly to the issue of solving this problem, because it has many important
lessons. So there’s two ways to solve this problem of finding the bound states of the energy
eigenstates of the harmonic oscillator.
This is a very interesting potential because all its energy eigenstates are bound states. That’s
not the case for the delta function potential. In the delta function potential we found one bound
state, but they’re all kind of unbound states with positive energy. But this potential grows
forever, never stops growing. So whatever energy you have, it is a bound state. It will decay. It
will be localized. So you just have bound states. It’s marvelously nice because of that property.
Much simpler than anything you can imagine.
So what do we have to do? We want to find the energy eigenstates. So we’ll write h, and I will
write phis. People write sometimes phis. Phi n of x is equal to E phi n of x. And we don’t know
the energy eigenstates. You know it’s a symmetric potential. It’s a real potential. This we used
to go psi n’s, but I will write them as phi n’s as many people do, because they are the
harmonic oscillator ones that are very famous. And we don’t know-- this could be En, energy
of the n-th state. And we don’t know what are the energies nor how the wave functions look.
And we have to solve a differential equation.
As I was saying, there’s two ways of solving this differential equation. One, treating this as a
differential equation and understanding why the energy is quantized from the differential
equation. This actually gives you a lot of insight as to what’s going on, and it will relate to the
kind of things you’re doing in the homework this week.
The other way is to be very clever and invent what are called raising and lowering operators,
and sort of solve this whole system without solving the general differential equation, by solving
a first order very simple differential equation, and then doing everything else algebraically with
creation and annihilation operators. We will also do that. But we will not develop that too far.
That’s The applications of that method are mostly for 805. So we’ll introduce creation and
annihilation operators, which are very nice and very useful. But we leave some of the
applications to coherent states of harmonic oscillators, to squeezed states of harmonic
oscillators, for later.
So this is what we want to solve. So we have minus h squared over 2m d second phi n dx
squared. And I will probably forget about the labels. The labels will come later as we solve the
equation. Plus v of x 1/2 m omega squared x squared phi of x, is equal to E phi of x.
OK. This is the question we want to solve. Now we’re going to do one thing first with this
differential equation. We don’t like all these dimensionful constants. If you had to put it in the
computer, what? Are you going to put 6 times 10 to the minus 23 over m? And you won’t solve
a differential equation. We have to clean this up.
To clean this up, there is a procedure that is guaranteed to do the job for you. And the
procedure is to change the x variable into a variable that has no units. This is guaranteed to
lead you to the way to solve this differential equation. You will be using this throughout the
semester. Cleaning differential equations is a nice skill.
And the fact is that there’s a method, and the method always proceeds by first writing x equals
au, where this will be unit-free, this quantity u. So this will become a differential equation on a
unit-free constant, which is ideal for your numerical solution and is much nicer.
But then this a to have units of length. So the first thing you have to do is look up in your
problem. What are your constants? And you have a mass m, a frequency omega, an h bar.
And you need to find a constant with units of length. Wasn’t that in your test? I think so.
How do you find a constant with units of length here? Well, energy is equal to-- you can write it
in two ways. The energy can be written as p squared over m, so it will be h squared over m
times a length squared. But from the potential, it also has the units of m omega squared a
squared. These are units, equations for units. The units of energy are these, and the units of
energy from the second term in the Hamiltonian are those. From where you get that a squared
is equal to h over m omega.
So that’s the constant that you need. If you have that constant, your differential equation
becomes what? Well, it becomes the following. Let’s write it out. It’s very simple, because x is
equal to au, and therefore you get minus h squared over 2m a squared d second phi du
squared. x is equal to au, so it basically just shows up here. Plus 1/2 m omega squared a
squared u squared phi, is equal to E phi.
Now things have to work nicely. If you did the job well, they have to work nicely. Let’s think of
the units of this equation. Phi is here, phi is here, phi is here, so phi is not an issue. The units
of phi are irrelevant. Here is units of energy, but u has no units. This has no units. So this must
have units of energy. And since u has no units, this derivative has no units, and this must have
units of energy. And these two numbers must be something nice if you substitute a squared.
And indeed, if you substitute a squared here, this whole number becomes h bar omega. And
this number becomes h bar omega as well, which is very nice. So this whole equation has
become minus 1/2 times h bar omega d second phi du squared plus 1/2 h bar omega u
squared phi equal E phi.
The next step is to say, you know, I don’t even want these energy units. Even though they
don’t look that bad, this equation looks much nicer than the original equation which had all
kinds of strange units. So I will multiply this equation by 2 over h omega to cancel it. So I get
minus d second phi du squared plus u squared phi is equal to 2E over h bar omega phi.
So look at this. The equation is now almost in perfect form. And in order to make it perfect, I
would say that the right hand side-- now, see again. Phi, whatever units it has, it doesn’t
matter. It’s all over the place. But this has no units, this derivative, and this has no units, this
multiplication. So this must have no units. And indeed, you know that h omega has units of
energy, and that’s energy.
So it suggests that you define a unit-free energy, free energy, which is 2E over h omega. And
calculating curly E is the same thing as calculating the energy, because if you know this
number you know the energy. The energy is h bar omega over 2 times curly E. The advantage
is that curly E will be either 1, 2, 1/2, a nice number, while E is some 0.87 Ev, or things like
that.
So all of this equation has been reduced to this very nice equation. Minus d second phi du
squared plus u squared phi is equal to E phi. Or, d second phi du squared is equal to u
squared minus E phi.

L13.5 Behavior of the differential equation (10:31)

MITOCW | watch?v=eNf8nH1yEYc
PROFESSOR: Now, what should happen? Somehow, this equation probably has solutions for all values of the
energy, but those solutions diverge and are not normalizeable. It’s the kind of thing you will
find with a shooting method that you’re doing with your computer. Count the solution and it
suddenly diverges up or diverges down and cannot be normalized. But for some specific
values, it can be normalized. So what we need is an intuition why this differential equation has
normalizeable solutions, only if the energies, curly E, take specific values. That’s intuition that
we need.
For that, we’ll look at the equation a little closer, and try to understand what happens at the
place where it can get in trouble, which is large X. That’s where you expect it to get in trouble.
So what does this equation become as u goes to plus minus infinity? Well this equation, at that
stage, becomes like this. The second phi, the u squared is roughly equal to u squared phi.
Because e is a constant, it doesn’t blow up, so the differential equation, the terms that are
supposed to be largest in this right hand side, is the u squared.
So how does this look as a solution? And you can see that that’s definitely not like a power
solution. If you have a power of u, say phi equals u to the n, after u differentiates, the power
goes down. But here it goes up. You have a u to the end. The relative is supposed to give you
u to the n plus 2, but that doesn’t work. So that’s not a power solution. So it has to be different.
So what it is, you can try a phi. And what it should be is of the form e to the u squared. It’s kind
of like that, because this is the function that when you’re differentiate, you bring down the
derivative of this quantity, which is a u. When you differentiate again, you can bring another u
to get the u squared, or you can differentiate this one. But if you differentiate the thing that is in
the bottom, you get something that diverges less. So morally speaking, this function is about
right. So I’ll put here, for example, e to the alpha over 2. And I’ll even put here a uk.
Can this work? A u to the k times alpha u squared. Well if we take two derivatives, if we take
one derivative, if I differentiate the u to the k, I get u to the k minus 1. I lose powers of u. If I
differentiate the exponential, I can get alpha times u times the same function. See, that’s,
roughly speaking, what’s happening. You differentiate the thing that diverges the most.
So if you differentiate twice, each time you differentiate you get a factor of alpha times u
squared phi, roughly. This is plus subleading. When you differentiate a function and you’re
wanting to show the most divergent thing, then you-- because we’re looking at the most
divergent part, you will always differentiate this, and this u to the k is really a spectator, it
doesn’t do anything, because when you differentiate that, you get something much smaller,
that doesn’t matter.
So yes, with these exponentials, we get something like this. Beside double pronged should
have been here. And therefore, you see that alpha is plus minus 1. Alpha is plus minus 1, and
those are likely to be approximate solutions as x goes to infinity. So we could expect solutions
for alpha equals 1, and I will write that it, and all this, I should say is always as u goes to
infinity. So always as u go to infinity, all of this in this blackboard.
So also as u go to infinity, we would expect maybe solutions of the form A u to the k, e to the
minus u squared over 2 plus B u to the k e to the u squared over 2. The two values of alpha
equal plus minus 1 are the possibilities for these two equations to match. So you would expect
things like this to be solutions.
And here you are seeing the beginning of the danger. Well a minus u squared over 2 times a
power sounds pretty good, but a u squared over 2 times a power sounds pretty bad. So
maybe this is what we want to happen. This is not an exact solution of anything yet. We’re just
looking at u going to plus and minus infinity, and maybe we’ll have such a behavior or such
behavior. But we want this one, otherwise we will never be able to normalize it.
So here it is. Without any loss of generality and inspired by this, this analysis is absolutely
crucial, you see, we’re following a very logical procedure. Cleaning the equation then looking
where the divergence would happen and learning something about the form of the solution.
Now without any loss of generality, I can write, I will write 5x is going to be a-- not of x
anymore. U, h of u times e to the minus u squared over 2.
So it’s an an sat, but it’s without any loss of generality, because you can always write any
function as another function times e to the minus u over 2, because you take the function, you
multiply by e to the plus u over 2 and e to the minus u over 2 and it’s written like that. But this
should be nice, because it has sort of the right behavior already. And here is the hope. The
hope is that this function now is a proxy for phi. If you know h, you know phi, which is what you
want. And this function, hopefully, won’t diverge. This will be into [INAUDIBLE].
So if this function doesn’t diverge, it will be a great thing. In particular, we could hope that h of
u is a polynomial. You see, if somebody with have come and said, look at that equation. Could
that be a polynomial? A polynomial is something that ends up to some power x to the 20 or x
to the 30, but it doesn’t go up forever. An exponential has all powers. This equation doesn’t
have a polynomial solution. No polynomial will ever solve this. But now that you’ve isolated the
divergence, there is a hope that a polynomial will work.
So for doing that, exploring that hope, I now have to substitute-- this is no assumption-- the
differential equation for phi implies a differential equation for h, you just substitute this and look
at it. That’s a one line computation or two line computation. I’ll give the answer. So what is the
differential equation for h? So back in star, you’ll get the second h, the u squared minus 2u dh
du plus e minus 1h equals zero.
So you substitute that into this equation and you get this differential equation. And now this is a
differential equation for h. We hope it has a polynomial solution. You will see that it wants to
have a polynomial solution, but it doesn’t quite make it. And then you will discover quantization
helps, and suddenly you get the other normal solution and everything works out.