WEBVTT

00:00:01.430 --> 00:00:03.030
PROFESSOR: Good morning.

00:00:03.030 --> 00:00:07.920
We're going to continue doing
perturbation theory today.

00:00:07.920 --> 00:00:11.990
We've finished doing
non-degenerate perturbation

00:00:11.990 --> 00:00:15.650
theory, and then turned to the
degenerate perturbation theory

00:00:15.650 --> 00:00:17.390
as well.

00:00:17.390 --> 00:00:23.660
We had computed already
the first order corrections

00:00:23.660 --> 00:00:30.080
to the state, first order
corrections to the energy,

00:00:30.080 --> 00:00:32.710
and second order
correction to the energy.

00:00:32.710 --> 00:00:36.470
So here in this formulas
that I've written

00:00:36.470 --> 00:00:40.330
are all the results
we have so far.

00:00:40.330 --> 00:00:43.700
And a useful thing
is to ask yourself

00:00:43.700 --> 00:00:47.030
whether you know what
every symbol here means.

00:00:47.030 --> 00:00:51.740
This is the state who
change the deformation

00:00:51.740 --> 00:00:56.520
of the original state n0
of the Hamiltonian H0.

00:00:56.520 --> 00:00:59.550
We have the Hamiltonian
H of lambda,

00:00:59.550 --> 00:01:05.840
which was equal to H0
plus lambda delta H.

00:01:05.840 --> 00:01:12.560
And n0 was an eigenstate
of H0 with energy En0.

00:01:12.560 --> 00:01:15.380
And we found that as
you turn on lambda,

00:01:15.380 --> 00:01:17.690
lambda becomes non-zero.

00:01:17.690 --> 00:01:19.190
The energies change.

00:01:19.190 --> 00:01:22.810
The eigenstate changes.

00:01:22.810 --> 00:01:27.560
And here is lambda, first
order in lambda, second order

00:01:27.560 --> 00:01:28.310
in lambda.

00:01:28.310 --> 00:01:31.325
We didn't compute the state
to second order in lambda.

00:01:33.920 --> 00:01:37.720
Moreover, we have
the symbol delta Hkn

00:01:37.720 --> 00:01:42.590
that was defined to
be the matrix element

00:01:42.590 --> 00:01:51.680
on the unperturbed eigenstate
of the operator delta H. Well,

00:01:51.680 --> 00:01:54.470
that's another symbol in here.

00:01:54.470 --> 00:01:58.680
We are summing over all
states in the spectrum,

00:01:58.680 --> 00:02:02.960
except for the state n here
because the denominator here

00:02:02.960 --> 00:02:04.590
would give you a 0.

00:02:04.590 --> 00:02:07.400
And this would be
problematic in general.

00:02:07.400 --> 00:02:10.100
So that was what we derived.

00:02:10.100 --> 00:02:14.220
And these are the
formulas we have.

00:02:14.220 --> 00:02:18.120
It's worth making a
few remarks as to what

00:02:18.120 --> 00:02:20.190
we see from these
formulas because they're

00:02:20.190 --> 00:02:22.110
a little complicated.

00:02:22.110 --> 00:02:27.570
And you may not have
immediate intuition about it.

00:02:27.570 --> 00:02:30.260
So a few remarks.

00:02:37.940 --> 00:02:40.050
And the first one--

00:02:40.050 --> 00:02:42.170
in fact, both remarks
that I'm going

00:02:42.170 --> 00:02:47.810
to make concern the behavior
of the energy as represented

00:02:47.810 --> 00:02:50.390
by this formula.

00:02:50.390 --> 00:02:55.760
First remark is that if you
were to look at the ground state

00:02:55.760 --> 00:03:04.040
energy of the system, the
first two terms here, so for n

00:03:04.040 --> 00:03:09.800
equals 0, if we call the
ground state by the label n

00:03:09.800 --> 00:03:14.270
equals 0 for this
state n equals 0,

00:03:14.270 --> 00:03:19.630
these first two terms on
the energy overestimate

00:03:19.630 --> 00:03:21.730
the ground state energy.

00:03:21.730 --> 00:03:25.690
They always give you more than
what the true ground state

00:03:25.690 --> 00:03:29.560
energy is for any
value of lambda.

00:03:29.560 --> 00:03:35.350
That is kind of plausible,
given that you see here

00:03:35.350 --> 00:03:40.310
that for the ground state
energy, when n is equal to 0,

00:03:40.310 --> 00:03:43.270
and it's the lowest
energy state,

00:03:43.270 --> 00:03:45.940
these differences
are all positive.

00:03:45.940 --> 00:03:48.370
All these states
have more energy.

00:03:48.370 --> 00:03:50.320
Therefore, this
term is negative.

00:03:50.320 --> 00:03:53.650
And the next correction
to order lambda square

00:03:53.650 --> 00:03:58.510
tends to lower the energy
of the ground state.

00:03:58.510 --> 00:04:11.890
So the claim is that the
first v lambda ground state

00:04:11.890 --> 00:04:30.005
energy overestimate the
true ground state energy.

00:04:35.930 --> 00:04:38.050
And so how do we see that?

00:04:38.050 --> 00:04:44.590
So we consider E0, because
we're doing n equals 0.

00:04:44.590 --> 00:04:46.210
This is the ground state.

00:04:46.210 --> 00:04:49.090
We start labeling with
n equals 0 sometimes.

00:04:51.810 --> 00:04:54.475
Plus lambda H00.

00:04:57.300 --> 00:05:02.610
That is the order lambda
estimate for the ground state

00:05:02.610 --> 00:05:04.990
energy.

00:05:04.990 --> 00:05:12.990
The first term can be
viewed as the expectation

00:05:12.990 --> 00:05:17.400
value of the original
Hamiltonian on the ground

00:05:17.400 --> 00:05:17.940
state.

00:05:17.940 --> 00:05:24.050
That is E0 because
this H0 and 0 is E00.

00:05:24.050 --> 00:05:26.910
So this is E00.

00:05:30.180 --> 00:05:31.585
Plus lambda.

00:05:34.630 --> 00:05:36.990
I'm sorry here.

00:05:36.990 --> 00:05:38.950
I made a little mistake.

00:05:38.950 --> 00:05:51.190
That should be
delta H. Delta H00

00:05:51.190 --> 00:06:00.690
plus lambda, the
expectation value of delta H

00:06:00.690 --> 00:06:03.630
on the same state.

00:06:03.630 --> 00:06:06.620
So these two things
together are nothing

00:06:06.620 --> 00:06:14.450
but the expectation value on the
ground state of H0 plus lambda

00:06:14.450 --> 00:06:19.690
delta H.

00:06:19.690 --> 00:06:20.950
That's kind of nice.

00:06:20.950 --> 00:06:22.750
Isn't it?

00:06:22.750 --> 00:06:25.960
That these two terms
really are nothing else

00:06:25.960 --> 00:06:30.550
but the expectation value on
the ground state of this thing

00:06:30.550 --> 00:06:33.115
that we called H of lambda.

00:06:46.240 --> 00:06:52.520
And now comes the
variational principal

00:06:52.520 --> 00:07:00.470
that says that if you evaluate
the Hamiltonian on an arbitrary

00:07:00.470 --> 00:07:04.040
state, the expectation value
of the Hamilton in an arbitrary

00:07:04.040 --> 00:07:08.370
state, you get more energy
than the ground state

00:07:08.370 --> 00:07:10.760
energy of the system.

00:07:10.760 --> 00:07:11.795
You always get more.

00:07:11.795 --> 00:07:13.400
And when you hit
the ground state,

00:07:13.400 --> 00:07:15.180
you get the lowest value.

00:07:15.180 --> 00:07:16.100
So look at that.

00:07:16.100 --> 00:07:18.260
This is the Hamiltonian.

00:07:18.260 --> 00:07:21.630
And this is the
unperturbed ground state.

00:07:21.630 --> 00:07:24.170
This is not the
real ground state.

00:07:24.170 --> 00:07:26.510
The real ground state is
what we're trying to find.

00:07:26.510 --> 00:07:28.950
So this is like
saying, OK, you're

00:07:28.950 --> 00:07:34.800
evaluating the exact Hamiltonian
on some arbitrary state.

00:07:34.800 --> 00:07:40.460
Therefore, this is greater than
or equal to the true ground

00:07:40.460 --> 00:07:46.260
state energy that we
would call E0 lambda.

00:07:46.260 --> 00:07:50.255
True ground state energy.

00:07:54.810 --> 00:07:59.550
Which in that notation,
indeed, is E0 lambda.

00:07:59.550 --> 00:08:04.800
So that's a nice result.
And it matches with the idea

00:08:04.800 --> 00:08:16.910
that for the ground state
the order lambda squared

00:08:16.910 --> 00:08:27.760
correction is minus lambda
squared, the sum over k

00:08:27.760 --> 00:08:41.590
different from 0 delta Hk0
squared over Ek minus E00.

00:08:41.590 --> 00:08:45.440
And this thing, as we mentioned,
the numerator is positive.

00:08:45.440 --> 00:08:48.010
The denominator is
positive because this

00:08:48.010 --> 00:08:49.780
was the ground state.

00:08:49.780 --> 00:08:53.950
And this is a
non-degenerate ground state.

00:08:53.950 --> 00:08:59.560
And K, therefore, has
more energy here, Ek0.

00:08:59.560 --> 00:09:01.280
So everything is positive.

00:09:01.280 --> 00:09:03.120
So this is negative.

00:09:15.620 --> 00:09:19.190
So actually here, there
is a generalization

00:09:19.190 --> 00:09:20.670
of this observation.

00:09:20.670 --> 00:09:25.830
And you can imagine that you
have now a particular state,

00:09:25.830 --> 00:09:28.580
n here.

00:09:28.580 --> 00:09:32.210
Then you have all these
states that are above

00:09:32.210 --> 00:09:34.160
and all these states
that are below.

00:09:37.500 --> 00:09:41.660
So let's look at the
second order correction

00:09:41.660 --> 00:09:47.060
to the energy of the state En,
the second order correction.

00:09:47.060 --> 00:09:49.490
The first order
correction shifts.

00:09:49.490 --> 00:09:53.300
The energy is
proportional to delta Hmn.

00:09:53.300 --> 00:09:55.930
But the second
order correction is

00:09:55.930 --> 00:10:02.180
minus lambda squared, the
sum over k different from n.

00:10:02.180 --> 00:10:04.310
So let's split it.

00:10:04.310 --> 00:10:09.070
k greater than n.

00:10:09.070 --> 00:10:19.780
And we'll write the same thing
Hkn squared over Ek0 minus En0.

00:10:19.780 --> 00:10:25.500
And then I want to write the
other states, the states where

00:10:25.500 --> 00:10:27.840
k is less than n.

00:10:33.250 --> 00:10:35.170
But let me do one thing here.

00:10:35.170 --> 00:10:40.930
On those states, I'll
switch the order of the sum.

00:10:40.930 --> 00:10:49.300
I'll put En0, the order of the
sign in the denominator, Ek0,

00:10:49.300 --> 00:10:51.650
and change the sign here.

00:10:51.650 --> 00:10:55.570
So these two changes
of sign are correlated.

00:10:55.570 --> 00:10:58.070
I changed the sign
in the denominator.

00:10:58.070 --> 00:11:01.480
And I changed the sign in front.

00:11:01.480 --> 00:11:04.210
And now, we see the following.

00:11:07.110 --> 00:11:14.690
All these states that are
above n, k is greater than n.

00:11:14.690 --> 00:11:16.310
This difference is positive.

00:11:16.310 --> 00:11:17.810
This is manifestly positive.

00:11:17.810 --> 00:11:19.940
This is negative.

00:11:19.940 --> 00:11:30.570
So these states are kind
of pushing your state down.

00:11:30.570 --> 00:11:34.680
The fact of this states on
top is a negative correction

00:11:34.680 --> 00:11:36.850
are pushing the states down.

00:11:36.850 --> 00:11:39.420
On the other hand,
the lower states,

00:11:39.420 --> 00:11:42.810
the states with k less
than n, again, I've

00:11:42.810 --> 00:11:44.290
ordered now convenient.

00:11:44.290 --> 00:11:47.710
The other states k
now have less energy.

00:11:47.710 --> 00:11:49.330
So this is positive.

00:11:49.330 --> 00:11:50.290
This is positive.

00:11:50.290 --> 00:11:51.090
This is positive.

00:11:51.090 --> 00:11:59.770
So these states over here are
actually pushing that state up.

00:12:03.370 --> 00:12:05.620
So the upper states
are pushing it down.

00:12:05.620 --> 00:12:07.780
The lower states
are pushing it up.

00:12:07.780 --> 00:12:12.560
This is usually referred
to as level repulsion.

00:12:12.560 --> 00:12:14.390
The levels repel each other.

00:12:14.390 --> 00:12:17.770
The upper states don't want the
state n to go close to them.

00:12:17.770 --> 00:12:22.580
The lower states don't want
them to go close to them either.

00:12:22.580 --> 00:12:26.410
So it's a nice dynamic
that helps you understand

00:12:26.410 --> 00:12:29.610
what's going on here.

00:12:29.610 --> 00:12:31.490
Another point we
want to make has

00:12:31.490 --> 00:12:34.160
to do with the validity
of this expansion.

00:12:34.160 --> 00:12:39.380
So in general, when you
have series expansions

00:12:39.380 --> 00:12:42.410
the issues of
convergence are delicate.

00:12:42.410 --> 00:12:48.440
So we can get a lot of insight
just by doing an example.

00:12:48.440 --> 00:12:58.255
So let me talk about the
validity of the perturbation

00:12:58.255 --> 00:12:58.755
series.

00:13:01.410 --> 00:13:05.650
This is supposed to
give us some insight.

00:13:05.650 --> 00:13:13.630
One thing we've said is
that we think delta H is

00:13:13.630 --> 00:13:17.710
supposed to be smaller than h0.

00:13:17.710 --> 00:13:19.150
Now, these are operators.

00:13:19.150 --> 00:13:21.520
So the statement
that they're smaller

00:13:21.520 --> 00:13:23.350
has to be made more precise.

00:13:23.350 --> 00:13:25.450
What is the size of an operator?

00:13:25.450 --> 00:13:27.700
And you could think,
well, we could

00:13:27.700 --> 00:13:30.370
say the entries of
the matrix elements

00:13:30.370 --> 00:13:32.620
should be much smaller.

00:13:32.620 --> 00:13:36.070
And that is true,
but it's not enough.

00:13:36.070 --> 00:13:39.190
For that, let's
consider an example.

00:13:39.190 --> 00:13:44.320
So this will be a two by two
matrix Hamiltonian, in which

00:13:44.320 --> 00:13:51.720
H of lambda will be given
by some H0 plus lambda v

00:13:51.720 --> 00:14:07.870
hat, which will be E1
0 E2 0 plus lambda.

00:14:07.870 --> 00:14:12.970
That v hat matrix will
be of the form v v star.

00:14:12.970 --> 00:14:14.230
And must be her mission.

00:14:14.230 --> 00:14:15.915
So v is a number.

00:14:18.800 --> 00:14:20.830
v star is the complex conjugate.

00:14:20.830 --> 00:14:24.770
These are the two energies,
because the Hamiltonian

00:14:24.770 --> 00:14:25.820
is known.

00:14:25.820 --> 00:14:29.190
Each time we talk about the
Hamiltonian that is known,

00:14:29.190 --> 00:14:32.540
we write that there's
a diagonal matrix.

00:14:32.540 --> 00:14:35.960
We're saying, yes, that
is the matrix element

00:14:35.960 --> 00:14:40.520
of the Hamiltonian in the basis
of eigenstates, which we know.

00:14:40.520 --> 00:14:47.670
So a known Hamiltonian you can
represent by a diagonal matrix.

00:14:47.670 --> 00:14:50.220
So here is our whole
matrix H of lambda.

00:15:08.400 --> 00:15:14.590
And you could say, all right,
our perturbation theory,

00:15:14.590 --> 00:15:18.880
practically speaking,
is these formulas that

00:15:18.880 --> 00:15:26.140
allow us to compute the
eigenvalues of this matrices,

00:15:26.140 --> 00:15:29.020
which are of this
matrix, which is

00:15:29.020 --> 00:15:32.890
the energies of the
system of the eigenstate

00:15:32.890 --> 00:15:35.290
and compute the eigenstates.

00:15:35.290 --> 00:15:38.320
For H0, these are
the eigenvalues.

00:15:38.320 --> 00:15:43.050
And the eigenstates
are 1 0 and 0 1.

00:15:46.130 --> 00:15:48.710
So perturbation
theory is really,

00:15:48.710 --> 00:15:50.900
you can think in
terms of Hamiltonians,

00:15:50.900 --> 00:15:53.880
but also you can think of
it in terms of matrices.

00:15:53.880 --> 00:15:58.100
It's allowing you to find the
eigenvalues and eigenvectors

00:15:58.100 --> 00:15:59.760
of this matrices.

00:15:59.760 --> 00:16:04.040
So this is a two by two
matrix with numbers.

00:16:04.040 --> 00:16:06.360
You know how to compute
the eigenvalues.

00:16:06.360 --> 00:16:10.930
So I'll give you the answer.

00:16:10.930 --> 00:16:22.960
The eigenvalues E plus
minus are E1 plus E2/2

00:16:22.960 --> 00:16:32.560
plus minus E1 minus E2/2
times the square root of 1

00:16:32.560 --> 00:16:40.330
plus lambda squared absolute
value of V2 squared E1

00:16:40.330 --> 00:16:47.480
minus E2/2 squared.

00:16:47.480 --> 00:16:48.180
Long formula.

00:16:51.380 --> 00:16:51.980
That's it.

00:16:51.980 --> 00:16:53.525
Those are the exact things.

00:16:58.040 --> 00:17:04.190
If you wanted to see this as a
result in perturbation theory,

00:17:04.190 --> 00:17:08.720
you would say, well,
lambda squared v squared

00:17:08.720 --> 00:17:11.089
or lambda v is small.

00:17:11.089 --> 00:17:14.290
And I'm going to think
of this term as small.

00:17:14.290 --> 00:17:17.869
And I'm going to
expand the square root.

00:17:17.869 --> 00:17:19.890
And if I expand
this square root,

00:17:19.890 --> 00:17:21.950
I'm going to get
all kinds of terms

00:17:21.950 --> 00:17:24.500
with different powers of lambda.

00:17:24.500 --> 00:17:29.840
You can also see there that
there's no energy correction

00:17:29.840 --> 00:17:32.540
linear in lambda.

00:17:32.540 --> 00:17:39.270
Because when you expand 1
plus epsilon square root,

00:17:39.270 --> 00:17:44.370
it's roughly 1 plus epsilon
over 2 plus dot, dot.

00:17:44.370 --> 00:17:46.410
And therefore, when you
span the square root,

00:17:46.410 --> 00:17:48.000
the first correction
is going to be

00:17:48.000 --> 00:17:50.850
proportional to lambda squared.

00:17:54.380 --> 00:17:56.890
And that actually
conforms to that

00:17:56.890 --> 00:18:00.610
because the matrix elements
of the Hamiltonian delta

00:18:00.610 --> 00:18:07.810
Hnn along the diagonals
for the perturbation are 0.

00:18:07.810 --> 00:18:12.460
So there is no order
lambda correction.

00:18:12.460 --> 00:18:15.580
So what do you do here then?

00:18:15.580 --> 00:18:18.730
You must do this expansion.

00:18:18.730 --> 00:18:22.540
And here, I write
the relevant series.

00:18:22.540 --> 00:18:27.400
f of z, if you define
it, as 1 plus z squared

00:18:27.400 --> 00:18:34.660
is equal 1 plus z squared over
2 minus z 4th over 8 plus z6/6.

00:18:37.360 --> 00:18:39.370
And it's not that
simple after that.

00:18:39.370 --> 00:18:47.080
Minus 5/128z to the 8
plus order z to the 10.

00:18:47.080 --> 00:18:50.770
So that's this perturbative
series expansion

00:18:50.770 --> 00:18:53.770
that you would use here.

00:18:53.770 --> 00:18:58.900
And how good is the
convergence of this series?

00:18:58.900 --> 00:19:00.195
Well, it's OK.

00:19:03.130 --> 00:19:07.690
It's something that when
you study complex analysis

00:19:07.690 --> 00:19:13.090
you see this function 1 plus
square root of z squared

00:19:13.090 --> 00:19:23.490
has branch cuts at i and at
minus i in the complex plane,

00:19:23.490 --> 00:19:25.230
in the z complex plane.

00:19:25.230 --> 00:19:27.630
Those are the branch cuts.

00:19:27.630 --> 00:19:31.770
These are the places where
this square root becomes 0.

00:19:31.770 --> 00:19:35.510
And you have to deal with them.

00:19:35.510 --> 00:19:42.270
And the function is
convergent only up to here.

00:19:42.270 --> 00:19:44.570
There is a radius
of convergence.

00:19:44.570 --> 00:19:46.910
A radius of convergence is 1.

00:19:46.910 --> 00:19:51.140
You can try it with
Mathematica or with the program

00:19:51.140 --> 00:19:54.710
and take 30 terms, 40 terms.

00:19:54.710 --> 00:19:59.450
And you will see that as long
as you take a point z here,

00:19:59.450 --> 00:20:00.380
it converges.

00:20:00.380 --> 00:20:03.920
You take a point a little bit
out, it blows up the series.

00:20:03.920 --> 00:20:06.650
So it has a radius
of convergence.

00:20:06.650 --> 00:20:08.900
That's not so bad.

00:20:08.900 --> 00:20:11.330
Radius of convergence is OK.

00:20:11.330 --> 00:20:18.020
But basically, we need z to be
small, which corresponds here

00:20:18.020 --> 00:20:27.320
to the statement that lambda v
z small for a fast convergence

00:20:27.320 --> 00:20:32.150
corresponds to lambda
v absolute value

00:20:32.150 --> 00:20:43.840
being smaller than E1 0 minus
E2 0 absolute value over 2.

00:20:43.840 --> 00:20:47.040
Roughly smaller.

00:20:47.040 --> 00:20:54.910
And now, we see that the thing
that matters for a perturbation

00:20:54.910 --> 00:21:01.200
series to be good is also
that the perturbation be

00:21:01.200 --> 00:21:05.720
small compared with
the energy differences.

00:21:05.720 --> 00:21:08.700
Not just it should be
small, the perturbation.

00:21:08.700 --> 00:21:11.350
It should be small compared
with the energy difference.

00:21:11.350 --> 00:21:15.480
So if you have a Hamiltonian
with a state of energy 100

00:21:15.480 --> 00:21:20.150
and a state of energy
101, you might say, well,

00:21:20.150 --> 00:21:22.510
if I take a
perturbation of size two

00:21:22.510 --> 00:21:25.550
or three, that's very small
compared to the energies.

00:21:25.550 --> 00:21:29.650
But it's not been small compared
to the difference of energies.

00:21:29.650 --> 00:21:32.920
And that can cause the
perturbation expansion

00:21:32.920 --> 00:21:34.280
to go wrong.

00:21:34.280 --> 00:21:38.320
So this gives you extra
insight that, in fact,

00:21:38.320 --> 00:21:42.160
being a small perturbation
not only means small compared

00:21:42.160 --> 00:21:45.040
to the energies, but
also small compared

00:21:45.040 --> 00:21:47.020
with the energy differences.

00:21:47.020 --> 00:21:49.970
Something that you
see here as well.

00:21:49.970 --> 00:21:52.480
Energy differences are
controlling things.

00:21:52.480 --> 00:21:57.750
And if the energy
differences are--

00:21:57.750 --> 00:22:02.110
if the perturbation is not small
compared to energy differences,

00:22:02.110 --> 00:22:06.520
then these ratios
can be rather large.

00:22:06.520 --> 00:22:09.610
And the perturbation
terms are very large.

00:22:09.610 --> 00:22:14.300
And nothing is very simple.

00:22:14.300 --> 00:22:20.380
So that's what I wanted to
say about the convergence

00:22:20.380 --> 00:22:22.270
of a perturbation expansion.

00:22:22.270 --> 00:22:24.310
More rigorous
statements can be made,

00:22:24.310 --> 00:22:27.540
but we're not going to
try to make them here.