Lecture 1: Identical Particles, Introduction (9/20/2005) : Symmetrizing The Multi-Particle Wavefunctions
Lecture 1: Identical Particles, Introduction (9/20/2005) : Symmetrizing The Multi-Particle Wavefunctions
Lecture 1: Identical Particles, Introduction (9/20/2005) : Symmetrizing The Multi-Particle Wavefunctions
introduction (9/20/2005)
The main unifying theme of this course will be time-dependence (and dynamics), but we will start with identical particles.
In classical physics, particles are distinguishable. Even if two electrons
happen to have exactly the same mass, charge, and other quantities, you
may trace them and distinguish which electron is Peter and which electron
is Paul.
However, particles in quantum mechanics dont have any ID. If two electrons scatter, it is impossible (even in principle) to decide which of them is
Peter and which is Paul. Actually, both possibilities contribute to the final
wavefunction that decides about the probabilities of various results. Draw a
picture of two possible histories of electrons that scatter elastically; one cant
say which of them occured.
p21
p2
+ 1 + V (~r1 ) + V (~r2 ) + Vrel (~r1 ~r2 )
2m 2m
1
where p2i =
h2 2i enters the kinetic energy, V (~r) is the potential energy
between an electron and the nucleus, and Vrel (~r1 ~r2 ) is the mutual interactions of the electrons. Obviously it looks nicely symmetric in both electrons.
What do we mean mathematically? Define the exchange operator P12 that
exchanges the electrons. It means that it acts as follows:
(P12 )(~r1 , ~r2 ) = (~r2 , ~r1 ).
If the electrons wavefunction depended on other quantities, such as their
spin, the spins would have to be exchanged on the RHS, too. Note that
2
P12
= 1; it squares to the identity operator:
2
(P12
)(~r1 , ~r2 ) = (P12 (P12 ))(~r1 , ~r2 ) = (P12 )(~r2 , ~r1 ) = 1 (~r1 , ~r2 ).
Now its important to realize that the exchange operator commutes with
the Hamiltonian:
[P12 , H] P12 H H P12 = 0.
It does not matter whether you first act with the Hamiltonian and then you
switch the dependence of the resulting wavefunction on ~r1 , ~r2 , or you first
exchange the electrons and then act with the Hamiltonian. Youll get the
same results. Fine. So now, because these two operators commute, we can
simultaneously diagonalize them. Because one of the two operators is the
Hamiltonian that determines time evolution, the eigenvalues of P12 will be
conserved.
More concretely, it is useful to consider eigenstates of P12 . Because weve
proved that the exchange operator squares to 1, the same thing holds for the
eigenvalues. If we find an eigenfunction (lets start to omit some awkward
parentheses) i.e. if
P12 (~r1 , ~r2 ) = p(~r1 , ~r2 ),
then also
2
P12
(~r1 , ~r2 ) = p2 (~r1 , ~r2 ).
But we know that the latter equals (~r1 , ~r2 ) which means that p2 = 1;
the eigenvalues must satisfy the same algebraic equation as the operator.
Obviously this is solved by p = 1. How do the eigenfunctions look like?
p = +1 :
p = 1 :
We say that the wavefunctions are symmetric in the first case, and antisymmetric in the second case.
2
These relations between the spin and the symmetry of the wavefunction
follow from the assumption that the energy is bounded from below and the
norm of the states (and thus the probabilities) must be non-negative.
The wavefunction of two Helium atoms is easily seen to be symmetric
(p = +1) because the exchange of two Helium atoms (He42 ) is nothing else
than the exchange of two pairs of electrons, two pairs of neutrons, and two
pairs of protons. And p2e p2n p2p = +1 regardless of the individual signs. But
more generally, one needs quantum field theory to prove the spin-statistics
relation above.
(MA, MA) = 0.
1
1
1
exp(E/kT ) =
Z
Z exp(E/kT ) + 0.
4
1
,
exp(Ef /kT ) + 1
p(Nf = 1) =
exp(Ef /kT )
.
exp(Ef /kT ) + 1
1
exp(Ef /kT ) + 1
In Maxwell-Boltzmann statistics, the +1 term in the denominator was missing; the number of particles in a different state was simply proportional to
the probability distribution of each individual distinguishable particle.
Bose-Einstein distribution
Consider the bosons. The number Nb for a given state is either 0 or 1 or 2
or any non-negative integer. The probability that there are Nb bosons must
depend on Nb like exp(Nb Eb /kT ). But the sum of these numbers must add
up to one. You can see that the unique solution is
p(Nb = nb ) = exp(nb Eb /kT ) exp((nb + 1)Eb /kT )
5
p(Nb = nb )nb =
nb =0
1
exp(Eb /kT ) 1
2 (r),
3 (r)
...
n (r).
permutations
X
permutations
X
where p is the sign of the permutation for fermions. The latter wavefunction
can also be written as the Slater determinant:
A det i (rj ) =
i,j
1 (r1 ) 2 (r1 )
1 (r2 ) 2 (r2 )
..
..
.
.
1 (rn ) 2 (rn )
n (r1 )
n (r2 )
..
..
.
.
n (rn )
(r1 , r2 ) =
.
2
Exchange forces
Because two fermions cant occupy the same state, you may think that
they dont even want to be too close. And you would be right. Even
non-interacting fermions tend to be further apart than if they were distinguishable. And bosons tend to be closer together than non-interacting
distinguishable particles would be.
To see it, start with distinguishable particles whose wavefunction is
(x1 , x2 ) = a (x1 )a (x2 ).
Calculate
h(x1 x2 )2 i = hx21 i 2hx1 x2 i + hx22 i
because the second integral in the product equals one (normalization). The
same thing applies to hx2 ib . Also, the integral calculating 2hx1 x2 i factorizes
as indicated.
For identical particles we must change a couple of things, starting from
the wavefunction. The sign below refers to bosons or fermions, respectively.
1
(x1 , x2 ) = [a (x1 )b (x2 ) b (x1 )a (x2 )].
2
In this case,
hx21 i =
1
2Z
1
+
2
1Z
2Z
1
2
=
(1)
(2)
(3)
(4)
(5)
The last two terms (among the four) vanish because their latter integral
factor equals zero by orthogonality of a and b . The first two terms give
you
1
. . . = (hx2 ia + hx2 ib )
2
2
Obviously, the result for hx2 i is the same thing because of an a-b symmetry.
The last term in the expectation value of the distance, namely hx1 x2 i, requires
an extra calculation:
hx1 x2 i =
+
Z
1Z
2
dx1 x1 |a (x1 )|
dx2 x2 |b (x2 )|2
2Z
Z
1
2
dx1 x1 |b (x1 )|
dx2 x2 |a (x2 )|2
2Z
Z
1
(6)
(7)
(8)
(9)
(10)
dx x a (x)b (x)
as the first factor in the third term (with x1 x, and it also appears in
the conjugated form, both in the third and the fourth term). This overlap
integral is only nonzero if the particles overlap in space; only in this case,
their indistinguishability affects the expectation value (and thus the force).
Otherwise we get back the previous result. The qualitative conclusion of
the calculated expectation values is, of course, that bosons tend to be closer
together while the fermions tend to be further apart.
The Fermi statistics requires the total wavefunction that can be usefully
written in a factorized form as long as the Hamiltonian does not directly
depend on the spins (or if it at least splits to two Hamiltonians that depend
on the positions and spins separately)
(x1 , x2 ; s1 , x2 ) = (x1 , x2 )(s1 , s2 )
to be antisymmetric
(x1 , x2 ; s1 , x2 ) = (x2 , x1 ; s2 , x1 ).
But because the spin wavefunction is already antisymmetric in our singlet
state
(s1 , s2 ) = (s2 , s1 ),
the spatial wavefunction for the electrons must actually be symmetric
(x1 , x2 ) = +(x2 , x1 ).
This effectively causes the electrons to behave as bosons. Both of them
occupy the space in between the nuclei, and the bound of H2 becomes strong.
If at least one of the atoms is helium, it contains electrons both with spin
up and spin down (the helium is a singlet itself). Therefore the other atom
either H or He must have at least one electron that shares the same
spin state with an electron of the first helium atom. Thats enough of an
antibinding to destroy chances for a He2 or HeH bound state.
We need to perform a more careful analysis for the more complex atoms;
see Gasiorowicz.
Nuclei and isospin
Spin 1/2 particles can be spinning up or down. Later we will see that
they are spinors a representation of the SU(2) group. The same mathematics may be applied in other contexts with different interpretations than
the actual spinning in space.
Consider protons and neutrons. They have a different charge. But the
electric charge is not so important in the nuclei where the strong nuclear force
is much more significant. From this perspective, a proton and a neutron look
like twins; a doublet. They may be described as nucleons that only differ
by the isospin (isotopical spin): the proton is a nucleon with spin up while
10
the neutrons isospin points down (these directions have nothing to do with
the real space).
Why is there a deuteron - a bound state of one proton and one neutron
np (heavy Hydrogen nucleus) but no pp state (neutron-free helium) or nn
(double-neutron)? And why does the deuteron have spin one? The total
wavefunction of two nucleons may be written as
= space spin isospin
and it must be antisymmetric for nucleons that are fermions because of their
spin 1/2. The ground state of the deuteron tries to minimize the angular momentum, much like the ground state of the Hydrogen atom. This implies that
space is symmetric. This means that spin isospin must be antisymmetric.
Were left with two possibilities:
1. spin antisymmetric, isospin symmetric:
J = 0 : spin
| i| i | i| i
=
,
2
I = 1 : iso
|pi|pi
= (|pi|ni + |ni|pi)/ 2
|ni|ni
| i| i
2 ,
(|
i|
i
+
|
i|
i)/
| i| i
I = 0 : iso =
|pi|ni |ni|pi
The explanation why two of these rules work is the following: the first rule
tries to make the spin wavefunction as symmetric as possible, which means
that the spatial wavefunction is antisymmetric as possible. Such an antisymmetry reduces the probability that the electrons are too close to each other,
and it reduces the positive energy contribution from their self-interactions.
The third rules is implied by a spin-orbital interaction proportional to l s
which is a linear function of j 2 for fixed l, s. The rule tries to minimize this
term if the shell is at most half-occupied; if it is more than half-full, it is
better to think in terms of holes for which the conclusion reverses.
12