Lecture 2 8.

321 Quantum Theory I, Fall 2017 5

Lecture 2 (Sep. 11, 2017)

2.1 More Relevant Math
2.1.1 Inner Products
Last time, we discussed the concept of a maximally linearly independent set, which is a set {|αj i}
of vectors that are linearly independent, and such that there exists no |βi such that {|αj i, |βi} is
linearly independent. In this case, {|αj i} forms a basis for V . If |α1 i, . . . , |αn i form a basis for V ,
then any vector |βi can be decomposed in terms of the basis vectors as
X
|βi = ci |αi i . (2.1)
i

Next, we discuss the concept of an inner product (we will assume that V is a vector space C).
A (Hermitian) inner product is a map (·, ·) : V × V → C with the following properties:

1. For all c ∈ C and |αi, |βi ∈ V , we have

(|αi, c|βi) = c(|αi, |βi) ,

(2.2)
(c|αi, |βi) = c∗ (|αi, |βi) ,

where c∗ is the complex conjugate of c.

2. For all |αi, |βi ∈ V ,

(|αi, |βi)∗ = (|βi, |αi) . (2.3)

3. For all |αi, |βi, |β 0 i ∈ V ,

(|αi, |βi + |β 0 i) = (|αi, |βi) + (|αi, |β 0 i) . (2.4)

4. For all |αi ∈ V ,

(|αi, |αi) ≥ 0 , (2.5)
and if (|αi, |αi) = 0, then |αi is the zero ket, |αi = 0.

Let’s consider some examples. First, consider V = Cn , whose vectors are of the form
 
z1
 .. 
|zi =  .  , z1 , . . . , zn ∈ C . (2.6)
zn

One definition of an inner product on this space is

(|zi, |wi) = z1∗ w1 + z2∗ w2 + · · · + zn∗ wn , (2.7)

which we can check satisfies all of the required properties of an inner product.
A more exciting example is the vector space of square-integrable functions on the interval [0, 1],
 ˆ 1 
dx f ∗ (x)f (x) < ∞ ,

V = f : [0, 1] → C  (2.8)
0
Lecture 2 8.321 Quantum Theory I, Fall 2017 6

as we discussed in the last lecture, which can be endowed with the inner product
ˆ 1
(f, g) = dx f ∗ (x)g(x) . (2.9)
0
Now we introduce a bit of terminology. The vectors we have been discussing are referred to as
kets. Two kets are orthogonal if (|αi, |βi) = 0. The norm of a ket is given by
p
N := (|αi, |αi) := k|αik . (2.10)
Using this norm, we can normalize any nonzero ket by defining
1
|α̃i = |α i , (2.11)
N
which has
(|α̃i, |α̃i) = 1 . (2.12)

2.1.2 Dual Space

Now we introduce the dual space. The space V that we have described so far is the space of kets
|αi. The dual space V ∗ is the space of objects called bras, which is “dual” to the ket space. To
every |αi ∈ V , we associate a bra hα| ∈ V ∗ , and define an operation between bras and kets,
hα|βi := (|αi, |βi) . (2.13)
This shows us the motivation for these names: when we combine a bra and a ket, we get a
“bra(c)ket.”
Consider the space of linear functionals γ : V → C. We introduce the notation γ → hγ|, such
that for any basis {|αj i}, the statement
γ : |αj i 7→ cj (2.14)
implies that
hγ|αj i = cj . (2.15)
We see that the dual space is just the space of linear functionals on V . The {γ} form an n-
dimensional vector space V ∗ , where n = dim V . We then have a duality V ↔ V ∗ , given by
|αi ↔ hα|. Note that scalars are conjugated under this duality, so that for all c ∈ C and |αi ∈ V ,
c|αi ↔ c∗ hα|.

2.1.3 Orthonormal Bases

An orthonormal basis is a basis {|φi i} such that
hφi |φj i = δij , (2.16)
where δij is the Kronecker delta. We know that we can write any ket in the form
X
|αi = ci |φi i . (2.17)
i

If the |φi i form an orthonormal basis, then we find that ci = hφi |αi. Thus, if we have an orthonormal
basis, then we can write any ket in the form
X
|αi = |φi ihφi |αi. (2.18)
i

These is an example of a completeness relation.

Lecture 2 8.321 Quantum Theory I, Fall 2017 7

2.1.4 Operators
Recall the second postulate of quantum mechanics: observables are represented in quantum me-
chanics as Hermitian operators acting on the Hilbert space H (from here on out, we will consider
states in the Hilbert space H rather than an arbitrary vector space V ). In order to understand this
statement, we first need to understand what an operator is, and then we must understand what a
Hermitian operator is.
An operator is an object that acts on a ket and returns another ket. Thus, if X is an operator,
then X acts on a ket |αi ∈ H as
X(|αi) := X|αi ∈ H , (2.19)
where the right-hand side is a naming convention. Two operators A and B are said to be equal
(A = B) if and only if A|αi = B|αi for all |αi ∈ H.
We introduce an addition operation ‘+’ on the space of operators, which is commutative and
associative: for all operators X, Y, Z, we have

X +Y =Y +X,
(2.20)
X + (Y + Z) = (X + Y ) + Z .

We will primarily be interested in linear operators: a linear operator X satisfies the property

X(cα |αi + cβ |βi) = cα X|αi + cβ X|βi (2.21)

for all cα , cβ ∈ F and |αi, |βi ∈ H. We can also define the notion of an anti-linear operator: an
anti-linear operator X satisfies

X(cα |αi + cβ |βi) = c∗α X|αi + c∗β X|βi (2.22)

for all cα , cβ ∈ F and |αi, |βi ∈ H.

We can also define an action of these same operators on the dual space, which will be a right
action of the form hβ|X for hβ| ∈ H∗ . We define this object via its inner product with kets, by
requiring that it satisfy the property

(hβ|X)|αi = hβ|(X|αi) (2.23)

for all |αi ∈ H. Using an orthonormal basis {|φj i}, we can then express this bra as
X
hβ|X = hβ |X |φj ihφj | . (2.24)
j

We also define multiplication of operators, such that for all operators X, Y and |αi ∈ H,

(XY )|αi = X(Y |αi) . (2.25)

This multiplication is not necessarily commutative, XY 6= Y X, but it is associative,

X(Y Z) = (XY )Z = XY Z . (2.26)

Lecture 2 8.321 Quantum Theory I, Fall 2017 8

2.1.5 Operators as Matrices in a Given Basis

In a given basis, the action of an operator can be expressed by a matrix. To see this, we first define
the identity operator 1, which satisfies
1|αi = |αi (2.27)
for all |αi ∈ H. Given an orthonormal basis |{a0 }i (notation from Sakurai), the identity operator
can be expressed in the form
1 = a0 a0  .
X 

(2.28)
a0
We can use Eq. (2.28) to write any operator X in the form
X = 1X 1
! !
X 
 X
= a0 a0  X |∗a00 iha00 |
(2.29)
a0 a00
X  
  

= a0 a0 X a00 a00  .
a0 ,a00

Note that ha0 |X|a00 i ∈ C is simply a number. Thus, all information about X is equivalently
contained in the complex numbers ha0 |X|a00 i, given any orthonormal basis |{a0 }i. This defines an
n × n matrix with complex entries corresponding to each operator X. For this reason, we will often
use the words “operator” and “matrix” interchangeably if the chosen basis is clear, even though
the concept of an operator is more fundamental.

2.1.6 Adjoint Operators

For each operator X, we define its adjoint operator (also called the Hermitian conjugate) X † to be
the operator satisfying
X † |αi ↔ hα|X (2.30)
under the duality H ↔ H∗ . If X † = X, then the operator X is called self-adjoint or Hermitian.
(Note that in infinite-dimensional spaces, the concepts of self-adjointness and Hermiticity are tech-
nically distinct. We may discuss this later in the course.) Note that (XY )† = Y † X † . The proof of
this fact is straightforward: For arbitrary |αi ∈ H, consider the kets defined by
|βi = Y |αi , |γi = X|βi . (2.31)
Their dual bras are then defined by
hα|Y † = hβ| , hβ|X † = hγ| . (2.32)
From these definitions, we have
XY |αi = |γi , hα|Y † X † = hγ| . (2.33)
Taking the dual of the first equation in (2.33) yields
hα|(XY )† = hγ| , (2.34)
which we can compare with the second equation in (2.33) to yield (XY )† = Y † X † .
This same concept has a well-defined meaning for matrices. As we have seen, any operator has
the same information content as an n × n matrix, where n = dim H. If M is a matrix corresponding
to the operator X, then X † corresponds to the matrix found by conjugating the entries of the
transpose M T . This matrix is denoted as M † , and is also called the Hermitian conjugate. A matrix
satisfying M † = M is called a Hermitian matrix.
Lecture 2 8.321 Quantum Theory I, Fall 2017 9

2.1.7 Operator Examples

As a first example of an operator, we define the outer product. Given a ket |αi ∈ H and a bra
hβ| ∈ H∗ , we define the operator
O = |αihβ| (2.35)
to be their outer product. This is an operator: for all |γi ∈ H, we have
O|γi = |αihβ|γi . (2.36)
Taking the dual of this equation, we find
hγ|O† = hβ|γi∗ hα|
(2.37)
= hγ|βihα| ,
and so we conclude that
O† = |βihα| . (2.38)
As a second example, consider the Hilbert space H of square-integrable functions on (−∞, ∞),
with inner product ˆ ∞
hf1 |f2 i = dx f1∗ (x)f2 (x) (2.39)
−∞
d
for |f1 i, |f2 i ∈ H. Now consider the operator A = dx . Using integration by parts, we see that
ˆ ∞ ˆ ∞  
∗ d d ∗
dx f1 (x) f2 (x) = dx − f1 (x) f2 (x) , (2.40)
−∞ dx −∞ dx
telling us that        ∗
d d
f1  f2 = − f2  f1 .
  (2.41)
dx dx
Thus,
 †
d d
=− . (2.42)
dx dx
Operators satisfying A† = −A are called anti-Hermitian. We can make this into a Hermitian
operator by multiplying it by the imaginary unit (or its negative):

d †
 
d
−i = −i . (2.43)
dx dx
You may recognize this as the momentum operator in one-dimensional quantum mechanics.
A third example is the identity operator 1, which we have already seen.
For some operators, we can define an inverse operator: for an operator A, its inverse operator
A−1 (if it exists) satisfies
A−1 A = AA−1 = 1 . (2.44)
Inverse operators are not guaranteed to exist.
Of special interest to us are unitary operators: a unitary operator U is one that satisfies
U −1 = U † . These are useful because they preserve the inner product:
 
hβ|U † (U |αi) = β U −1 U α = hβ|αi .

  
(2.45)

We are also interested in projection operators, which are operators A satisfying A2 = A. An

example of a projection operator is A = |αihα|, for some |αi ∈ H.
Lecture 2 8.321 Quantum Theory I, Fall 2017 10

2.1.8 Eigenstates and Eigenvalues

If, for some operator A and ket |αi ∈ H, we have

A|αi = a|αi , (2.46)

with a ∈ C, then |αi is an eigenket or eigenstate of the operator A, with eigenvalue a. The spectrum
of an operator A is the set of its eigenvalues {a}.
We now prove an important theorem regarding the eigenkets and eigenvalues of Hermitian
operators.
Theorem 1 (Spectral Theorem). If A is a Hermitian operator, A = A† , then
1. all eigenvalues ai of A are real, and

2. eigenkets of A with distinct eigenvalues are orthogonal.

Proof. Suppose that for some Hermitian operator A and kets |a0 i, |a00 i ∈ H we have

Aa0 = a0 a0 ,
   
(2.47a)

00 
a A = a00 (a00 )∗ .


(2.47b)

Multiplying Eq. (2.47a) by ha00 | on the left and multiplying Eq. (2.47b) by |a0 i on the right yields

00   0 
a Aa = a0 a00 a0 ,

 
(2.48a)
00 ∗ 00  0

00   0 

 
a Aa = a
  a a . (2.48b)

Comparing Eqs. (2.48a) and (2.48b), we see that

 00
∗
− a0 a00 a0 = 0 .

 
a (2.49)

If |a0 i = |a00 i, then this proves that (a0 )∗ = a0 , which means that a0 ∈ R. If |a0 i =
6 |a00 i and we have
0
a = 00 00 0 0 00
6 a , then this proves that ha |a i = 0, i.e., |a i and |a i are orthogonal. This completes the
proof.

If we normalize each eigenket |ai i of a Hermitian operator, so that hai |ai i = 1, and have chosen
the eigenkets so that distinct eigenkets with the same eigenvalue are orthogonal, then Thm. 1 tells
us that we have
hai |aj i = δij , (2.50)
i.e., the normalized eigenkets are orthonormal. We can then decompose
X
A= a|aiha| . (2.51)
a

We can check that for some eigenket |bi, we have

X X
A|bi = a|aiha|bi = a|aiδab = b|bi . (2.52)
a a

Note that the equation

1=
X
|aiha| (2.53)
a

is simply a special case of Eq. (2.51).

MIT OpenCourseWare
https://ocw.mit.edu

8.321 Quantum Theory I

Fall 2017

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.