Physics 324, Fall 2001

Dirac Notation

1 Vectors
1.1 Inner product
Recall from linear algebra: we can represent a vector V as a column vector; then V = (V T )
is a row vector, and the inner product (another name for dot product) between two vectors
is written as

A B = A1 B1 + A2 B2 + . (1)

In conventional vector notation, the above is just A ~ B.

~ Note that the inner product of a
vector with itself is positive definite; we can define the norm of a vector to be

|V | = V V , (2)

which is a non-negative real number. (In conventional vector notation, this is |V~ |, which
is the length of V~ ).

1.2 Basis vectors

We can expand a vector in a set of basis vectors {
ei }, provided the set is complete, which
means that the basis vectors span the whole vector space. The basis is called orthonormal
if they satisfy

ei ej = ij (orthonormality), (3)

and an orthonormal basis is complete if they satisfy

ei ei = I
X
(completeness), (4)
i

where I is the unit matrix. (note that a column vector ei times a row vector ei is a square
matrix, following the usual definition of matrix multiplication).
Assuming we have a complete orthonormal basis, we can write

ei ei V ei V ) .
X X
V = IV = Vi ei , Vi ( (5)
i i

The Vi are complex numbers; we say that Vi are the components of V in the {
ei } basis.

1
1.3 Eigenvectors as basis vectors
Sometimes it is convenient to choose as basis vectors the eigenvectors of a particular
matrix. In quantum mechanics, measurable quantities correspond to hermitian operators;
so here we will look at hermitian matrices. A hermitian matrix is one satisfying

M = M (MT ) (hermitian) . (6)

This just means that the components of a hermitian matrix satisfy Mij = Mji .
We say that the vector vn is an eigenvector of the matrix M if it satisfies

Mvn = n vn , (7)

where n is a number called an eigenvalue of M. If M is hermitian, the eigenvalues n

are all real, and the eigenvectors may be taken to be orthonormal:

vm vn = nm . (8)

So we can take the vn to be our basis vectors, and write an arbitrary vector A in this basis
as
X
A= An v n . (9)
n

where the An are in general complex numbers. This is a convenient choice if we wish to
know what is the action of the hermitian matrix M when it multiplies the vector A:
X X
MA = An Mvn = An n v n . (10)
n n

2 Dirac notation for vectors

Now let us introduce Dirac notation for vectors. We simply rewrite all the equations in the
above section in terms of bras and kets. We replace

V |V i , V hV | , A B hA|Bi . (11)

Suppose we have basis vector |ii, analogous to the ei , which form a complete orthonormal
set:

hi|ji = ij (orthonormality)
(12)
|iihi| = 1
P
i (completeness) ,

where 1 is the identity operator; it has the property 1|i = |i for any |i.
Then any vector |V i may be expanded in this basis as
X X
|V i = 1 |V i = |iihi|V i Vi |ii , Vi hi|V i . (13)
i i

2
Note that hV |ii = Vi .
As before, we can use the eigenvectors of a hermitian operator for our basis vectors.
Matrices become operators in this language, M M . Then the eigenvalue equation
becomes
|ni = n |ni ,
M (14)
where the n are real and we can take the |ni kets to be orthonormal: hm|ni = mn . Then
we can write

|V i = M |ni =
X X X
M Vn |ni = Vn M Vn n |ni . (15)
n n n

3 Dirac notation for quantum mechanics

Functions can be considered to be vectors in an infinite dimensional space, provided that
they are normalizable. In quantum mechanics, wave functions can be thought of as vectors
in this space. We will denote a quantum state as |i. This state is normalized if we make
it have unit norm: h|i = 1.
Measurable quantities, such as position, momentum, energy, angular momentum, spin,
is an operator corresponding
etc are all associated with operators which can act on |i. If O
to some measurable quantity, its expectation value is given by hOi = h|O|i.
Note that
since

hA|O|Bi
= hB|O |Ai ,
(check this in the case of finite length vectors and matrices!) it follows that if your mea-
= hOi
surable quantity is real (and they always are!) then hOi implies that

h|O|i |i ,
= h|O
=O
or that O .
Conclusion: Measurable quantities are associated with hermitian operators.
In order to compute expectation values for given quantum states, it is often useful to
choose a convenient basis. I will discuss three common bases that are often used: the
position eigenstate basis, the momentum eigenstate basis, and the energy eigenstate basis.

3.1 Position eigenstate basis: |xi

The position operator x = x is a hermitian operator, and we can use its eigenvectors as
an orthonormal basis. The state |xi is defined to be the eigenstate of x with eigenvalue x:
x|xi = x|xi . (16)
What is new here is that the eigenvalues x are not discrete, and so we use the Dirac
-function for normalization:
hx|x0 i = (x x0 ) (orthonormality). (17)

3
The states |xi form a complete basis for our space and
Z
dx |xihx| = 1 (completeness) (18)

Note that the sum over discrete basis vectors in eq. (12) has been replaced by an integral.
Also, the unit operator 1 has replaced the unit matrix I. Therefore we can expand our
state |i in terms of the |xi basis vectors:
Z Z
|i = 1 |i = dx |xihx|i = dx (x)|xi , (19)

where we have defined (x) hx|i. This (x) is nothing but our familiar wavefunction.
In the present language, (x) are the coordinates of the our state |i in the |xi basis.
Note that in eq. (19) we inserted the unit operator in the guise of a integral over |xihx|
this technique is very powerful, and is called inserting a complete set of states.
If we have normalized |i so that h|i = 1, it follows that
Z Z
1 = h|i = h|1|i = dx h|xihx|i = dx (x)(x) . (20)

This is the usual normalization condition we have seen.

Computing h xn i for our state is particularly easy in the |xi basis, since |xi is an eigen-
state of the operator xn :

xn i = Zh|
h xn |i
= dx h| xn |xihx|i
Z
= dx xn h|xihx|i
Z
= dx xn (x)(x) . (21)

In the next section I will discuss measuring h

pi, using the position eigenstate basis.

3.2 Momentum eigenstate basis: |pi

Another useful basis is formed by the eigenstates of the momentum operator p:
Z
p|pi = p|pi , hp|p0 i = (p p0 ) , dp |pihp| = 1 . (22)

We can expand |i in the momentum basis as

Z Z
|i = dp |pihp|i = dp (p)|pi , (p) hp|i . (23)

We see that (p) is just the wavefunction in the momentum basis. As in the above section,
we can easily compute expectation values such as hpn i using this basis.

4
 It is interesting to ask how we can translate between the |xi and the |pi bases. For this,
we need to know the quantity hx|pi. We can get this by knowing that (x) and (p) are
Fourier transforms of each other:
Z
dp
(x) = (p)eipx/h . (24)
2
h
We can rewrite this as
Z
dp
(x) = hx|i = hp|ieipx/h . (25)
2
h
Inserting a complete set of states on the left hand side of the above equation we get
Z Z
dp
dp hx|pihp|i = hp|ieipx/h , (26)
2
h
implying that
eipx/h
hx|pi = . (27)
2h
Using this result we can also compute (here |yi is an eigenvector of x with eigenvalue y):
Z
hy|
p|xi = dp hy|
p|pihp|xi
Z
= dp p hy|pihp|xi
Z
p ip(yx)/h
= dp e
2h
Z dp ip(yx)/h
= i e
x 2

h (x y) .
= i (28)
x
Therefore if we know (x) but not (p), we can still compute h
pi as
Z Z
h
pi = h|
p|i = dy dx h|yihy|
p|xihx|i
!
Z Z

= dy dx (y) i
h (x y) (x) . (29)
x
Integrating by parts with respect to x (ignoring the boundary terms at x = , which
vanish) we get
!
Z Z

h
pi = = dy dx (y)(x y) i h (x)
!
x
Z

= dx (x) i
h (x) (30)
x

So we see that in the x representation, p i
h x .

5
3.3 Energy eigenstates: |ni
which gives the energy. Up to
Finally, another operator of interest is the Hamiltonian H
now, the Hamiltonians we have seen take the form
2
= p + V (
H x) . (31)
2m
The time dependent Schrodinger equation can be written as

i
h |, ti = H|, ti . (32)
t
You can think of |, ti as a vector moving around in our vector space as a function of time,
and the above equation governs how it moves.
is also hermitian, provided that the potential V (x)
Since p and x are hermitian, then H
is a real function. Therefore we can use eigenstates |ni of H as basis vectors:

|nihn| = 1 .
X
H|ni = En |ni , hm|ni = mn , (33)
n

Note that this eigenvalue equation is simply the time-independent Schrodinger equation,
and that since H is hermitian, the eigenvalues En are real numbers. (Here I have assumed
that the energy eigenvalues En are discrete; this is correct for bound states but not scat-
tering states.
R
For states with continuous eigenvalue, replace the mn by (m n), and the
P
n by dn.)
Then we can expand |, ti in this basis, with time-dependent coefficients:
X
|, ti = cn (t)|ni . (34)
n

Plugging this into the Schrodinger equation eq. (32) we find

X dcn (t) X
n (t)|ni =
X
i
h |, ti = i
h |ni = Hc En cn (t)|ni (35)
t n dt n n

Since the |ni form an orthonormal basis, it is easy to show that the above equation implies

dcn (t)
i
h = En cn (t) = cn (t) = eiEn t/h cn (0) . (36)
dt
Therefore the solution for |, ti is

eiEn t/h cn (0)|ni .

X
|, ti = (37)
n

So all we need to know is what are cn (0) (the initial conditions at t = 0), and the solutions
|ni, En to the time-independent Schrodinger equation, eq. (33). Note that

eiEn t/h cn (0) hn| .

X
h, t| = (38)
n

6
 If |, ti is normalized, it follows that

cm (t)cn (t)hm|ni = |cn (t)|2 = |cn (0)|2 .

XX X X
1 = h, t|, ti = (39)
m n n n

Note that in the coordinate basis, where x x , p i

hd/dx.

hx|ni = un (x) , (40)

where
2 d2
" #
h
+ V (x) un (x) = En un (x) . (41)
2m dx2

Solving this sort of equation for different potentials V (x) (and generaliztions in 3 dimen-
sions) what we will be doing for the rest of this quarter.

3.4 Comments
Why are these bases we have discussed necessarily different from each other? For example,
cant we find a basis in which both p and x are simple? No. If we had a state |x, pi which
was simultaneously an eigenstate of x (with eigenvalue x) and p (with eigenvalue p) it
would follow that

x, p] |x, pi = xp|x, pi px|x, pi

[
= px|x, pi x p|x, pi
= (px xp)|x, pi = 0 (42)

since x and p are ordinary numbers, and commute both with each other and with operators
p and x. But we know that

[ h1
x, p] = i (43)

(we dont normally write the 1) which is inconsistent with the previous result, and so there
cannot be such as state as |x, pi.
The result is that a state |i cannot simultaneously be an eigenstate of two
operators that do not commute.
For the hydrogen atom, we will find that we can find basis sates which are simultaneously
L
eigenstates of H, L,
and L z , where L refers to the angular momentum vector. We will
discuss a normalized basis of angular momentum states later in the course.