This action might not be possible to undo. Are you sure you want to continue?

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

I. INTRODUCTION

The formalism of quantum mechanics is discussed in chapter 3 in Griﬃths. There is also

an Appendix on Linear Algebra. You might read these sections in addition to these notes,

and in a few places I will include comments on Griﬃths’ presentation. Although these notes

are self contained, they cover a more general case than Griﬃths, so I recommend that you

study section 3.4 on the statistical interpretation of the wave function. Section 3.5 on the

generalized uncertainty relation is not covered in these notes.

Here you will directly be introduced to the Dirac formalism, which Griﬃths treats in

section 3.6 but never uses fully. It is a very powerful formalism and not diﬃcult to learn.

It brings out the structure of quantum mechanics in a simple way and has also become

increasingly important in many modern application of quantum mechanics such as quantum

information.

The mathematics of quantum mechanics is linear algebra. However, there are a couple

of diﬀerences compared to the linear algebra that you have learned in your mathematics

courses. These are:

1. A vector is denoted as [α) (or ¸β[). This is a trivial change of notation, but it is

surprisingly useful.

2. The vectors are complex—i.e. they have complex components. This is a simple

generalization. Just some

∗

here and there (well....).

3. The vectors may have an inﬁnite number of components, a

i

, i = 1, 2, 3, ....∞ and,

more importantly, they may have a continuum of components, a(x), −∞ < x < ∞

note that here x is a real number. The set of components is in this case a function of

the real variable x (and we use the standard notation for functions a(x) instead of a

x

,

which would be more analogous to a

i

).

Item 3 above is in general a highly non-trivial generalization of linear algebra, and some of the

properties that hold in ﬁnite dimensional vector spaces may no longer be true. Fortunately,

however, for the vectors that occur in quantum mechanics this does not happen and the

1

generalization is simple! Thus we need not worry about the things a mathematician would

start to think about. We will just generalize the standard linear algebra to vectors whose

components are functions. Why is it that the vectors of quantum mechanics have this nice

property? Well, it is really something required by physics. The properties in question are

namely related to the Born interpretation of the wave function as a probability amplitude,

and what can go wrong is that probabilities would not add up to one. This is of course not

acceptable for a physical theory, so in quantum mechanics we only use those vector spaces

for which this holds true. It might then be a tricky mathematical problem to prove these

properties, i.e. to prove that we are using a consistent mathematical model of a physical

system. For the cases you will encounter in this course you can rest assured that these

problems have been sorted out.[1]

How linear algebra enters quantum mechanics will be explored and explained later, but

it might be good to know from the start what to look for. Here is a brief preview and an

illustrative example:

1. The state of a system is described by a vector.

2. Observables, i.e. things that can be measured, correspond to operators that imple-

ment linear transformations of the vectors .

3. A crucial feature of quantum mechanics is the principle of superposition, the linear

combination aψ(x) + bφ(x) of two wave functions is a possible wave function for the

system. In linear algebra this is just the basic property of a vector space, which says

that a linear combination of vectors is a new vector in the space (see (6) below).

Example: You have already encountered the eigenfunctions ψ

n

(x) and energy eigenvalues

E

n

= ω(n +

1

2

) for the harmonic oscillator hamiltonian. It turned out to be very useful to

introduce the ladder operators, a and a

†

satisfying the commutation relation:

[a, a

†

] = 1 (1)

Their action on the normalized eigenfunctions were,

a[n) =

√

n[n −1) (2)

a

†

[n) =

√

n + 1[n + 1)

2

From these we formed the number operator

ˆ

N = a

†

a. Note that all these operators are

diﬀerential operators acting on wave functions ψ(x), and that the origin of the commutation

relation (1) is the operator identity

d

dx

x = x

d

dx

+ 1. Now, there is another type of objects

that in general do not commute, namely matrices. So it is quite natural to ask whether the

commutation relation (1) could be represented by matrices. It is not hard to prove that this

is in fact not possible using ﬁnite matrices. However, if we allow for inﬁnite matrices and

deﬁne,

a =

_

_

_

_

_

_

_

_

_

_

_

_

_

0 1

0

√

2

0

√

3

0 .

. .

.

_

_

_

_

_

_

_

_

_

_

_

_

_

a

†

=

_

_

_

_

_

_

_

_

_

_

_

_

_

0

1 0

√

2 0

√

3 0

. .

.

_

_

_

_

_

_

_

_

_

_

_

_

_

, (3)

it is clear (just do the multiplication!) that (1) is satisﬁed, if the ”1” on the LHS is interpreted

as the inﬁnite diagonal unit matrix, 1 = diag(1, 1, . . . ). The number operator takes the form,

ˆ

N =

_

_

_

_

_

_

_

_

_

_

_

_

_

0

1

2

3

.

.

_

_

_

_

_

_

_

_

_

_

_

_

_

(4)

If we now let these matrices act on the column vectors

ψ

0

=

_

_

_

_

_

_

_

_

_

_

1

0

0

0

0

_

_

_

_

_

_

_

_

_

_

ψ

1

=

_

_

_

_

_

_

_

_

_

_

0

1

0

0

0

_

_

_

_

_

_

_

_

_

_

ψ

2

=

_

_

_

_

_

_

_

_

_

_

0

0

1

0

0

_

_

_

_

_

_

_

_

_

_

etc. (5)

the relations (2) are fulﬁlled. Since the hamiltonaian is given by the operator relation

H = ω(

ˆ

N +

1

2

) we see that the energy eigenvalue for the state ψ

n

is E

n

= ω(n +

1

2

) as

expected.

3

We thus have two alternative descriptions of the harmonic oscillator states, either in terms

of wave functions ψ

n

(x) and diﬀerential operators acting on these functions, or in terms of

(inﬁnite) vectors, ψ

n

and matrices acting upon them. We say that these two descriptions are

diﬀerent representations of the quantum mechanical system. Mathematically, the function

ψ

n

(x) and the inﬁnite column vector ψ

n

are just the components of the same abstract state

vector [n) in diﬀerent basis. The ﬁrst correspond to the position basis [x) and the second

to the energy basis [E). All this will be explained in detail in the following.

II. DIRAC’S NOTATION

A vector is denoted as [α) (called a ket-vector or simply a ket), or, as ¸β[ (called a

bra-vector or simply a bra). This is standard quantum mechanics notation introduced by

Dirac.[2] Note that a bra and ket together gives a bracket ¸α[β) (this is a physicists sense

of a joke). ¸α[β) denotes the scalar product of the vectors [α) and [β).

III. COMPLEX VECTORS

A vector space consists of vectors and scalars. Two vectors [α), [β) can be added and

multiplied by scalars a, b and this gives a new vector [γ)

[γ) = a[α) + b[β) (6)

in the vector space. As noted above this is the mathematical formulation of the quantum

mechanical superposition principle. A vector can be expanded in a set of linearly independent

basis vectors [e

i

), i = 1, 2, ...n,

[α) =

n

i=1

a

i

[e

i

) . (7)

The coeﬃcients a

i

are complex scalars and are called the components of the vector, they

describe the vector completely [3] [4]. n is the dimension of the vector space. For now we

shall take n to be ﬁnite, but later we shall also discuss vector space with inﬁnitely many

components, as the one considered in the example in section I. You are probably used to

the scalars being real numbers. In the vector spaces that occur in quantum mechanics the

scalars are, in general, complex (they must be since the wave function is complex, and this

is an example of components of a vector). This leads to some simple changes.

4

The scalar product of two vectors [α), [β) is denoted ¸α[β) and has the properties

¸β[α) = ¸α[β)

∗

(8)

¸α[α) ≥ 0 , and ¸α[α) = 0 if and only if [α) = 0 (9)

¸γ[(a[α) + b[β)) = a¸γ[α) + b¸γ[β) (10)

Note the complex conjugation that enters in the ﬁrst relation.

Let [e

i

) be an orthonormal basis,

¸e

i

[e

j

) = δ

ij

. (11)

The components of [α) =

n

i=1

a

i

[e

i

) are then

a

i

= ¸e

i

[α) (12)

and the scalar product of [α) and [β) =

n

i=1

b

i

[e

i

) is

¸α[β) =

n

i=1

a

∗

i

b

i

, (13)

and the square of the norm of the vector [α) is

¸α[α) =

n

i=1

a

∗

i

a

i

. (14)

To each ket [α) =

n

i=1

a

i

[e

i

), there is a corresponding bra

¸α[ =

n

i=1

¸e

i

[a

∗

i

. (15)

Note the complex conjugation that enters on the components of ¸α[, this follows from (8)

and makes the norm, i.e. the length of the vector, real and non-negative.

I will frequently use the simpliﬁed notation [i) = [e

i

).

IV. LINEAR TRANSFORMATIONS

A linear transformation

ˆ

T takes a vector [α) and transforms it into another vector [α

) =

ˆ

T[α) in such a way that the linearity condition

ˆ

T(a[α) + b[β)) = a

ˆ

T[α) + b

ˆ

T[β) . (16)

5

is fulﬁlled.

ˆ

T is called an operator. Because of this linearity relation, the action of the

operator

ˆ

T on an arbitrary vector can be obtained from its action on the basis vectors [i):

ˆ

T[j) =

n

i=1

T

ij

[i) Note order of indices! (17)

T

ij

= ¸i[

ˆ

T[j) . (18)

T

ij

are the components of the operator

ˆ

T, they form an n n-matrix T. If [α) =

i

a

i

[i)

and [α

) =

i

a

i

[i) we introduce the notation a and a

**for the column vectors (i.e. the
**

n 1-matrices) with components a

i

and a

i

respectively. Then we can write [α

) =

ˆ

T[α) in

matrix language as

i

a

i

[i) =

ˆ

T

j

a

j

[j) =

i,j

a

j

T

ij

[i) (19)

a

i

=

j

T

ij

a

j

(20)

a

= Ta . (21)

Note that matrix equations, such as (19), depend on the particular basis [i) used.

V. HERMITIAN TRANSFORMATIONS

If

ˆ

T is an operator then one deﬁnes the hermitian conjugate operator

ˆ

T

†

as follows

¸α[

ˆ

T

†

[β) = ¸β[

ˆ

T[α)

∗

(22)

(for all vectors [α), [β)).

[The deﬁnition (22) is equivalent to Griﬃths [A.87] ¸

ˆ

T

†

α[β) = ¸α[

ˆ

Tβ) ([3.83] in the ﬁrst edition). His

notation is confusing and I will try to avoid it, but you need to know it in order to read and understand the

textbook. What stands inside [...) is just a name for the vector, and a natural notation is then [

ˆ

Tβ) =

ˆ

T[β).

Using this notation one ﬁnds

¸

ˆ

T

†

α[β) = ¸β[

ˆ

T

†

α)

∗

= ¸β[

ˆ

T

†

[α)

∗

¸α[

ˆ

Tβ) = ¸α[

ˆ

T[β) (23)

which, when inserted in Griﬃths’ eq. [A.87] gives (the complex conjugate of) our deﬁnition (22). Note that

¸

ˆ

T

†

α[ is not equal to ¸α[

ˆ

T

†

, whereas [

ˆ

Tα) =

ˆ

T[α)). ]

6

A very important type of operators in quantum mechanics are those which hermitian

conjugate equals the operator itself:

ˆ

T

†

=

ˆ

T . (24)

Such an operator is said to be hermitian or self-adjoint. It follows form (22) that the

components of a hermitian operator obey (T

ji

)

∗

= T

ij

. The matrix T is then said to be

hermitian and this condition is written as T

†

= T.

For a product of operators one has the useful rule

(

ˆ

S

ˆ

T)

†

=

ˆ

T

†

ˆ

S

†

. (25)

This can be proven using (22), or by using components, i.e. matrix notation. Note the

change of order of the operators.

Hermitian operators are of crucial importance in quantum mechanics since any observable,

such as e.g. position, momentum, angular momentum, energy, spin, etc. ... for any physical

system corresponds to a hermitian operator. [5]

VI. EIGENVALUES AND EIGENVECTORS

Given an operator

ˆ

T, consider the equation

ˆ

T[α) = λ[α) , (26)

where λ is a scalar. This is an eigenvalue equation. Vectors [α) that solve this equation are

called eigenvectors of the operator

ˆ

T since they are mapped into themselves by

ˆ

T, and λ are

called eigenvalues. In matrix notation this equation becomes

Ta = λa or, in components,

n

j=1

T

ij

α

j

= λα

i

. (27)

You have encountered this equation in your linear algebra courses. Rewriting it as

(T − λ1)a = 0, one sees that it has solutions a ,= 0 only if det(T − λ1) = 0. Solving

this characteristic equation (’sekularekvationen’), which is an nth order polynomial in λ,

determines the n eigenvalues λ

i

, i = 1 . . . n. Having obtained the eigenvalues one can deter-

mine the corresponding eigenvectors by solving (26). If T happened to be a diagonal matrix

then, obviously, the eigenvalues are simply the diagonal elements. In general, to ﬁnd the

7

eigenvalues is equivalent to diagonalizing the matrix T by a change of basis vectors. T is

diagonal when the basis vectors are the eigenvectors.

The set of all eigenvalues ¦λ

i

¦ and eigenvectors ¦[i

s

)¦ for an operator is called the spectrum

of the operator. If there is only one eigenvector, [i) with a particular eigenvalue λ

i

then this

eigenvalue is said to be non-degenerate, if there are d linearly independent eigenvectors,

[i

s

), s = 1 . . . d with eigenvalue λ

i

this eigenvalue is said to be d-fold degenerate. From now

on we shall consider the non-degenerate case.

If

ˆ

T is a hermitian operator, then the following holds

1. All eigenvalues are real.

2. Eigenvectors corresponding to diﬀerent eigenvalues λ ,= λ

are orthogonal.

3. The eigenvectors span the vector space.

Item 3 means that the eigenvectors of a hermitian operator can be chosen as a basis

[i), i = 1, 2...n in the vector space. If all eigenvalues are non-degenerate, the corresponding

eigenvectors are automatically orthogonal because of 2. For a degenerate eigenvalue, one

may have to take linear combinations of the obtained eigenvectors to obtain orthonormal

basis vectors: ¸i[j[) = δ

ij

.

Properties 1 and 2 are simple to prove, see Griﬃths. The property 3 is obvious if all

eigenvalues are diﬀerent and can be shown to hold in general.

In quantum mechanics, an observable, e.g. the energy, is represented by a hermitian

operator and the possible results of the measurement of this observable are the eigenvalues

of this operator. These must be real and they are since the operator is hermitian. After one

has measured the observable and obtained one of the eigenvalues the state of the system

is described by the corresponding eigenvector—this is the so called collapse of the wave

function.

8

VII. SUMMARY OF RESULTS AND THE RESOLUTION OF THE IDENTITY

For a hermitian operator we have from the above

ˆ

T =

ˆ

T

†

(28)

ˆ

T[i) = λ

i

[i) (29)

[i) , i = 1, 2, ...n basis of eigenvectors, λ

i

real eigenvalues (30)

¸i[j) = δ

ij

orthonormal basis (31)

[ψ) =

n

i=1

ψ

i

[i) , ψ

i

= ¸i[ψ) (32)

[ψ) =

n

i=1

[i)¸i[ψ) = (

n

i=1

[i)¸i[)[ψ) . (33)

From the last equation we read oﬀ the operator identity (called the ’resolution of the iden-

tity’)

1 =

n

i=1

[i)¸i[ . (34)

Note that here ”1” is an operator (and should perhaps be denoted by

ˆ

1). The identity

(34) is a consequence of the vectors [i) forming a basis. This identity is very useful, it can

be inserted anywhere in an equation since it is just the identity operator and many useful

formulae can be derived almost without eﬀort, for example the components of [ψ) above.

As a further example we derive the components of an arbitrary operator

ˆ

A in the basis ¦[i)¦

deﬁned by the hermitian operator

ˆ

T:

ˆ

A = 1

ˆ

A 1 =

i,j

[i)¸i[

ˆ

A[j)¸j[ =

i,j

[i)A

ij

¸j[ (35)

A

ij

= ¸i[

ˆ

A[j) .

Note that

ˆ

A does not have to be hermitian. In particular, if we calculate the components

of the operator

ˆ

T itself in its own basis we get,

T

ij

= ¸i[

ˆ

T[j) = λ

i

¸i[j) = λ

i

δ

ij

(36)

which just expresses the trivial fact that the operator

ˆ

T is diagonal in the basis deﬁned by

its own eigenvectors.

9

So far everything was simple, every hermitian operator in a ﬁnite dimensional vector space

has a complete set of eigenvectors that can be chosen as basis vectors (and the eigenvalues

are real). As mentioned above the only new things compared to the linear algebra you have

met in your mathematics course is that the vector space is complex but this was easily taken

care of.

We now want to extend the theory of linear algebra to inﬁnite dimensional vector spaces.

The vector index i may run from 1 to ∞ (i.e. n → ∞ in the formulae above) but it may also

be that the vector index becomes a continuous variable x which takes for example all real

values. The components of a vector ψ

i

then becomes a function of x, ψ(x)—thus functions

can be thought of as components of vectors.

Formally, the generalization to inﬁnite dimensional vector spaces is simple. However,

mathematical diﬃculties may arise. For example, it may happen that the scalar product of

two vectors is inﬁnite. Moreover, the theorems that hold for a hermitian operator in a ﬁnite

dimensional vector space—that the eigenvectors span the space—do not in general hold

for a hermitian operator in an inﬁnite dimensional vector space. Thus inﬁnite dimensional

vector spaces are mathematically much trickier than their ﬁnite dimensional counterparts.

See Griﬃths for a disscussion of these problems. However, it so happens that the inﬁnite

dimensional vector spaces that occur in quantum mechanics are simple and behave exactly

as their ﬁnite dimensional brethren! In particular, the hermitian operators that occur in

quantum mechanics do have eigenvectors that span the vector space and hence can be used

as a basis. These eigenvectors may have an inﬁnite norm, however this inﬁnity is of the

same fairly innocent type as we encountered for the free particle.

Thus, for the purposes of quantum mechanics one can generalize linear algebra to inﬁnite

dimensional vector spaces almost trivially without worrying about whether things are well-

deﬁned or not. Quantum mechanics guarantees that this is the case. By using the bra, ket

notation of Dirac the formulae and the calculations will look almost the same also when

there is a continuum of components.

In my opinion, Griﬃths puts too much emphasis on the potential mathematical diﬃculties

in inﬁnite dimensional vector spaces and he also stops using the Dirac notation that he has

introduced. This obscures the close analogy with the ﬁnite dimensional case. Here, we

will use Dirac’s vector notation and trust that the mathematics works just as in the ﬁnite

dimensional case as is required for the consistency of quantum mechanics.

10

VIII. THE INFINITE DIMENSIONAL DISCRETE SPECTRUM: i = 1, 2, ....∞

As noted above this generalization is trivial, just make the replacement n → ∞ in the

formulae above.

IX. THE CONTINUOUS SPECTRUM: i →x

There are two new things: The sum is replaced by an integral and the Kronecker delta

is replaced by Dirac’s delta function[6]

n

i=1

→

_

∞

−∞

dx (37)

δ

ij

→ δ(x −x

) . (38)

The equations (28-33) become

ˆ

T =

ˆ

T

†

(39)

ˆ

T[x) = λ

x

[x) (40)

[x) , −∞ < x < ∞ basis of eigenvectors, λ

x

real eigenvalues (41)

¸x[x

) = δ(x −x

) orthonormal basis (42)

[ψ) =

_

∞

−∞

dx ψ(x)[x) , ψ(x) = ¸x[ψ) (43)

[ψ) =

_

∞

−∞

dx [x)¸x[ψ) =

__

∞

−∞

dx [x)¸x[

_

[ψ) (44)

x here can be the position of a particle, but can also be any other continuum variable such

as the momentum or the energy for the scattering states.[7] From the last equation we read

oﬀ the operator identity (called the ’resolution of the identity’)

1 =

_

∞

−∞

dx [x)¸x[ (45)

Compare this to the ﬁnite dimensional case above! The resolution of the identity can be

used to derive results just as above, using the deﬁning property of Dirac’s delta function

_

dx

δ(x −x

)f(x

) = f(x) . (46)

Note that Dirac’s delta function vanishes when x ,= x

and is ∞ when x = x

. Thus

eigenvectors corresponding to diﬀferent eigenvalues are orthogonal. However, the norm of

the eigenvectors is inﬁnite. (Compare the notes on the free particle.)

11

Comment: It is also possible to have an operator whose spectrum (i.e. set of eigenvalues)

consists of both a discrete part and a continuous part—the Hamiltonian for the Hydrogen

atom is an example of this and so is the delta-potential well. In these cases, one will have

both an integral

_

and a

in the formulae.

X. THE SCALAR PRODUCT

It is straightforward to derive the expression for the scalar product in terms of the com-

ponents of the vectors, Eq [3.6] in Griﬃths:

[ψ) =

_

dx [x)ψ(x) , [φ) =

_

dx [x)φ(x) (47)

¸φ[ψ) = (

_

dx φ

∗

(x)¸x[)(

_

dx

[x

)ψ(x

) (48)

=

_

dx

_

dx

φ

∗

(x)ψ(x

)¸x[x

) (49)

using ¸x[x

) = δ(x − x

**) and the deﬁning property of the delta function (46), we arrive at
**

the important result

¸φ[ψ) =

_

dx φ

∗

(x)ψ(x) . (50)

Compare this expression to the ﬁnite dimensional one eq.(13).

Note that Griﬃths takes the expression (50), in terms of the components ψ(x) of the vector ψ as the

deﬁnition of the scalar product. This is possible but obscures the general structure and the similarities to

the ﬁnite dimensional case. The integral in (50) must exist, thus this leads to restrictions on the allowed

functions ψ(x), see Griﬃths.

XI. CHANGE OF BASIS

Using the resolution of the identity it is very easy to perform a change of basis. We

illustrate this here with the discrete case, but the formulae translate immediately to the

general case.

Let [e

i

), i = 1, ...n and [f

i

), i = 1, ...n be two sets of normalized basis vectors (they may

for example be the eigenvectors of two operators corresponding to two physical observables).

12

There are now two resolutions of the identity

1 =

i

[e

i

)¸e

i

[ =

i

[f

i

)¸f

i

[ . (51)

Using these one ﬁnds immediately the linear transformation that relates the two sets of basis

vectors

[e

i

) =

j

[f

j

)¸f

j

[e

i

) . (52)

A vector [ψ) can be expanded as

[ψ) =

i

[e

i

)¸e

i

[ψ) =

i

[f

i

)¸f

i

[ψ) (53)

and its components are related as

¸e

i

[ψ) =

j

¸e

i

[f

j

)¸f

j

[ψ) , (54)

again using the resolution of the identity. Deﬁning the matrix S with components S

ij

=

¸e

i

[f

j

), this can be written in matrix notation as ψ

(e)

= Sψ

(f)

, where ψ

(e/f)

is the vector

with elements ¸e

i

/f

i

[ψ). For the components of an operator

ˆ

T one ﬁnds

¸e

i

[

ˆ

T[e

j

) =

k,m

¸e

i

[f

k

)¸f

k

[

ˆ

T[f

m

)¸f

m

[e

i

) . (55)

In matrix notation this becomes

T

(e)

= ST

(f)

S

−1

, (56)

where S

−1

is the matrix with components S

−1

ij

= ¸f

i

[e

j

). Using the resolution of the identity

one veriﬁes that S

−1

S = 1, hence S

−1

is the inverse of the matrix S.

The transformation (56) is called a similarity transformation. Diagonalizing the matrix

T

(f)

by ﬁnding a similarity transformation S such that T

(e)

is diagonal is equivalent to

transforming to a basis consisting of the eigenvectors of

ˆ

T. (We assume that

ˆ

T corresponds

to a physical observable so that a basis of eigenvectors exists.)

XII. THE SCHR

¨

ODINGER EQUATION

The time dependent Schr¨odinger equation in Dirac notation reads

ˆ

H[Ψ) = i

∂

∂t

[Ψ) , (57)

13

where

ˆ

H is the Hamiltonian describing the system, and [Ψ(t)) is the state vector that

contains all information about the system. Written in this abstract vector form it applies

to any physical system: a particle moving along the x-axis, a particle moving in three

dimensions, a spin, several particles, a molecule, a solid, a superconductor...... you name it!

Just as we did before we can derive a time-independent equation by separing variables

(we assume that

ˆ

H has no explicit t dependence). Making the ansatz

[Ψ(t)) = e

−iEt/

[ψ) , (58)

where [ψ) is t-independent, we ﬁnd the time-independent Schr¨odinger equation

ˆ

H[ψ) = E[ψ) . (59)

The above is the Schr¨odinger equation in vector form, we can obtain it in component

form by inserting the resolution of the identity in suitable places. In the discrete case, with

i

[i)¸i[ = 1, we obtain

i,j

[i)¸i[

ˆ

H[j)¸j[ψ) = E

i

[i)¸i[ψ) , (60)

which gives the component equation

j

H

ij

ψ

j

= Eψ

i

, (61)

where we have deﬁned, in the standard way,

H

ij

= ¸i[

ˆ

H[j), ψ

i

= ¸i[ψ) . (62)

The Schr¨odinger equation for, for example, a spin is of this form.

In the continuum case, with

_

dx [x)¸x[ = 1, we obtain instead

_

dx

_

dy [x)¸x[

ˆ

H[y)¸y[ψ) = E

_

dx [x)¸x[ψ) , (63)

which gives the component equation

_

dy H(x, y)ψ(y) = Eψ(x) (64)

where

H(x, y) = ¸x[

ˆ

H[y), ψ(x) = ¸x[ψ) . (65)

14

The Schr¨odinger equation [−

2

2m

d

2

dx

2

+ V (x)]ψ(x) = Eψ(x) that describes the motion of

a particle along the x-axis is of the form (64), i.e., it is given in component form. We see

immediately that the right hand sides of the two equations agree. The left hand sides will

also agree if we make the identiﬁcation H(x, y) = δ(x −y)[−

2

2m

d

2

dy

2

+ V (y)].

The variable x in the continuum case need not be the position of a particle moving along

the x-axis, but can be any continuous observable, such as the momentum. The formulae can

also easily be generalized to the case of several continuous variables, which is needed for,

for example, one particles motion in three dimensional space or several particles or...... All

one has to do is to replace x above by a set of variables and the integral

_

dx by a multiple

integral. Formally, everything will be the same.

XIII. THE OPERATORS ˆ x AND ˆ p

Here we will discuss the position, ˆ x, and momentum, ˆ p, operators for a particles motion.

Consider the eigenvalue equations

ˆ x[x) = x[x) (66)

ˆ p[p) = p[p) . (67)

Here, ˆ x, ˆ p are operators, x, p are eigenvalues and [x), [p) are eigenvectors. There is a unique

eigenvector [x) ([p)) for each real x (p) and they are complete:

_

∞

−∞

dx [x)¸x[ = 1 (68)

_

∞

−∞

dp [p)¸p[ = 1 . (69)

This is always true for operators in quantum mechanics that correspond to physical ob-

servables. Sometimes it can be proven, but in general, we simply assume this to be true,

without proof, for any quantum mechanical operator. The completeness is intimately tied

to the foundations of quantum mechanics as we will see when discussing the statistical

interpretation below.

A general state (describing a particles motion in one dimension) can be expanded in

either the [x) or the [p) basis:

[ψ) =

_

∞

−∞

dx [x)¸x[ψ) =

_

∞

−∞

dp [p)¸p[ψ) . (70)

15

This gives two wave functions that describe the system: ψ(x) = ¸x[ψ) and ψ(p) = ¸p[ψ).

These are just components of the same vector [ψ) in diﬀerent bases, i.e. they are related by

a change of basis, see the section above on this topic. ψ(x) is the wave function in position

space, this is the usual wave function which enters the ususal Schr¨odinger equation and

which gives the probability of ﬁnding the particle at position x. ψ(p) on the other hand, is

the wave function in momentum space and gives the probability to ﬁnd the particle with

momentum p. We will see below that ψ(x) and ψ(p) are related by a Fourier transformation.

Let us determine the components of [x) and [p) in the x-basis. For [x), we have

[y) =

_

dx [x)¸x[y) =

_

dx [x)δ(x −y) (71)

hence,

ψ

y

(x) ≡ ¸x[y) = δ(x −y) . (72)

Thus, the wave function in x-space for the state [x) is Dirac’s delta-function. This makes

sense, since, according to the probability interpretation of the wave function, it means that

the particle can only be found at x = y. The operator ˆ x becomes in x-basis

ˆ x =

_

dx

_

dy [x)¸x[ˆ x[y)¸y[ =

_

dx

_

dy [x) xδ(x −y) ¸y[ (73)

=

_

dx x[x)¸x[ ,

where we have used (66), (72) and the properties of the δ-function. Notice the similarity with

the ﬁnite dimensional expression eq.(35), and that in its own eigenbasis, ˆ x is represented by

a diagonal ’matrix’ xδ(x −y) just as in eq.(36).

We now turn to the components of [p) =

_

dx [x)¸x[p). ψ

p

(x) ≡ ¸x[p) is the wave function

for the [p) state. To proceed we need to know what ˆ p is. We deﬁned this operator in the

x-basis as

ˆ p

x

ψ(x) = −i

∂

∂x

ψ(x) . (74)

(Here, we use ˆ p for the abstract operator, e.g. the one in (67), whereas ˆ p

x

denotes the

representation of this operator in the x-basis (ˆ p

x

= −i∂/∂x). Usually, we use ˆ p also

for the latter operator.) Let us recall the solution of the eigenvalue equation for ˆ p in the

16

x-representation:

ˆ p

x

ψ

p

(x) = p ψ

p

(x) (75)

−i

∂

∂x

ψ

p

(x) = p ψ

p

(x) (76)

ψ

p

(x) = Ae

ipx/

, p real . (77)

Thus we have

[p) =

1

√

2π

_

dx [x)e

ipx/

, ψ

p

(x) ≡ ¸x[p) =

1

√

2π

e

ipx/

, (78)

where we have chosen A = 1/

√

2π. We can now evaluate the scalar product of [p) states:

¸p

[p) =

1

2π

_

dx

_

dx

¸x

[x)e

i(px−p

x

)/

(79)

=

1

2π

_

dx e

ix(p−p

)/

= δ(p

−p) .

In the last step we used the useful relation (see Problem 2.26 in Griﬃths)

1

2π

_

∞

−∞

dk e

ikx

= δ(x) . (80)

This is proven as follows. Let F(k) be the Fourier transform of f(x):

F(k) =

1

√

2π

_

dx f(x)e

ikx

(81)

f(x) =

1

√

2π

_

dk F(k)e

−ikx

. (82)

One then ﬁnds

_

dx f(x)

_

1

2π

_

dk e

ikx

_

=

1

√

2π

_

dk F(k) = f(0) , (83)

from which (80) follows.

We summarize some of the important results of this section:

[y) =

_

dx δ(y −x)[x) (84)

¸x[y) = δ(x −y) (85)

[p) =

1

√

2π

_

dx e

ipx/

[x) (86)

¸x[p) =

1

√

2π

e

ipx/

(87)

¸p

[p) = δ(p

−p) . (88)

Comments

17

• [p) (as well as [x)) are non-normalizable but note that they are orthogonal and almost

normalized in the sense that δ

ij

→ δ(x − y): ¸x[y) = δ(x − y) and ¸p

[p) = δ(p

− p)

(Griﬃths calls this ’delta-function normalized’), and this is good enough, see the notes

on the free particle.

• The wave functions in x and p space ψ(x) = ¸x[ψ) and ψ(p) = ¸p[ψ) are related by a

Fourier transform:

¸x[ψ) =

_

dp ¸x[p)¸p[ψ) =

1

√

2π

_

dp e

ipx/

¸p[ψ) . (89)

We end this section by proving that the operators ˆ x and ˆ p are hermitian. For ˆ x we have:

¸φ[ˆ x[ψ) =

_

dx

_

dy ¸φ[x)¸x[ˆ x[y)¸y[ψ) =

_

dx φ

∗

(x)xψ(x) (90)

¸φ[ˆ x

†

[ψ) ≡ ¸ψ[ˆ x[φ)

∗

= (

_

dx ψ

∗

xφ)

∗

=

_

dx ψxφ

∗

= ¸φ[ˆ x[ψ) , (91)

hence ˆ x is hermitian, see (22).

For ˆ p, we ﬁnd

¸φ[ ˆ p[ψ) =

_

dx φ

∗

(x)(−i

d

dx

)ψ(x) (92)

¸φ[ ˆ p

†

[ψ) ≡ ¸ψ[ ˆ p[φ)

∗

=

__

dx ψ

∗

(x)(−i

d

dx

)φ(x)

_

∗

=

_

−iψ

∗

φ[

∞

−∞

_

∗

+

__

dx i

dψ

∗

dx

φ

_

∗

=

_

dx φ

∗

(x)(−i

d

dx

)ψ(x) + (....)[

∞

−∞

= ¸φ[ ˆ p[ψ) , (93)

provided the wave functions go to zero fast enough at inﬁnity so that the surface term

vanishes (this is the case for normalizable wave functions). This shows that ˆ p is hermitian

(22).

XIV. THE HILBERT SPACE

The set of possible vectors [ψ) describing the quantum mechanical state of a particular

system form a Hilbert space. Diﬀerent systems have diﬀerent Hilbert spaces. We here

discuss brieﬂy the properties of such a space.

18

A space is said to be complete if all convergent series of elements in the space also belong

to the space. (Note that this has nothing to do with the completeness of the eigenvectors

to an operator discussed above.)

A Hilbert space H is a vector space, with a scalar product, that is complete:

[α) =

∞

j=1

a

j

[α

j

) (94)

is an element in H provided [α

i

) are elements in H and the series is convergent

¸α[α) =

∞

j=1

[a

j

[

2

< ∞ . (95)

For a particle moving along the x-axis, the Hilbert space is L

2

(−∞, ∞), this is the space

of all functions ψ(x), such that

_

∞

−∞

dx [ψ(x)[

2

< ∞ . (96)

These functions are called square-integrable. (Note that ψ(x) are the components of the

vector and that the Hilbert space here is deﬁned in terms of the components.)

For one spin 1/2, the Hilbert space is a two dimensional vector space, for spin one 1 it is

a three dimensional vector space, for N spin 1/2 it is a 2

N

dimensional vector space etc.

XV. THE STATISTICAL INTERPRETATION

Griﬃths discussion of the statistical interpretation in Section 3.4 is restricted to the

motion of one particle in one dimension. Here we consider the general case.

1. The system is whatever we describe by quantum mechanics. It can be a particle moving

in one dimension, in three dimensions, several particles, a spin, particles with spin, a

molecule, a solid etc.

2. The system is completely described by a vector [ψ) in a Hilbert space H. Each vector

in H describes a possible state for the system.

• Note that the principle of superposition is implied by this.

• A vector in H is called a state vector since it describes a quantum state.

19

3. Each observable (i.e. each measurable quantity) Q corresponds to a hermitian operator

ˆ

Q acting on the vectors in H.

• For the particle: Q(x, p, t) →

ˆ

Q(ˆ x, ˆ p, t).

4. The expectation value of Q is:

¸Q) =

¸ψ[

ˆ

Q[ψ)

¸ψ[ψ)

(97)

= ¸ψ[

ˆ

Q[ψ) , if ¸ψ[ψ) = 1 . (98)

• Repeated measurements on identically prepared systems (i.e. each one of them

in state [ψ)) gives this mean value.

Let

ˆ

Q have eigenvectors [e

i

) with eigenvalues λ

i

(We assume here that the spectrum

¦λ

i

¦ is discrete, for the generalization, see below.):

ˆ

Q[e

i

) = λ

i

[e

i

) , i = 1, 2, ... (99)

• λ

i

are real (since

ˆ

Q is hermitian).

• ¦[e

i

)¦ is a basis: ¸e

i

[e

j

) = δ

ij

,

i

[e

i

)¸i[ = 1

5. A measurement of the observable Q gives one of the eigenvalues λ

i

.

If [ψ) =

i

[e

i

)¸e

i

[ψ) then the probability, P

i

to get λ

i

as a result of the measurement

is

P

i

=

[¸e

i

[ψ)[

2

¸ψ[ψ)

(100)

= [¸e

i

[ψ)[

2

if ¸ψ[ψ) = 1. (101)

• Compare this to the statistical interpretation of ψ(x) = ¸x[ψ).

• The probability for the system to be found in the general (normalized) state [χ)

is obtained by replacing [e

i

) by [χ) above.

20

6. Immediately after the measurement of the observable Q that resulted in λ

i

, the state

of the system is [ψ

after

) = [e

i

).

• This is the collapse of the wave function (or rather the state vector) [ψ).

• Repeated measurement of Q imediately after the ﬁrst measuremet gives the same

result λ

i

with probability one.

Continuous spectrum

For a general hermitian operator

ˆ

Q with a continuous spectrum one is, in general, not

guaranteed that the eigenvectors span the space. From the above we infer that this property

is of fundamental importance in quantum mechanics—it is closely tied to the statistical

interpretation of the state vector. One assumes that all hermitian operators that correspond

to physical observables have the property that their eigenvectors span the Hilbert space. In

some cases, this can be proven, in other cases one simply takes this as an assumption.

In the discussion above we assumed a discrete spectrum, but 1-6 hold also when the

spectrum is continuous, provided one makes the standard replacements:

i

→

_

dx and

δ

ij

→ δ(x −y).[8]

[1] For the mathematically interested, more advanced texts on quantum mechanics will contain

some discussion of these points.

[2] Dirac was one of the most famous physicists of the twentieth century. His perhaps most im-

portant contribution was the Dirac equation which combines quantum mechanics and special

relativity to give a very accurate description of electrons. His 1930 textbook, The Principles of

Quantum Mechanics, is still recommended reading for the serious student of quantum mechan-

ics.

[3] The vector space contains a null vector, 0, i.e. a vector with the property that [α) +0 = [α) for

any [α). Obviously, this is the vector with all components 0, I will denote this vector simply by

0. Griﬃths uses the notation [0) for this 0-vector, see [A.4]—this is very bad notation. Normally,

in physics texts [0) is used to mean a vector that is not the 0-vector. Avoid using [0) for the

0-vector!

21

[4] Griﬃths sometimes puts the vector equal to its components. This is OK if you know what it

means but I will try to avoid this.

[5] Another important type of operator in quantum mechanics are operators that obey

ˆ

U

ˆ

U

†

= 1.

This means that the hermitian conjugate operator is equal to the inverse operator

ˆ

U

†

=

ˆ

U

−1

,

where the inverse is deﬁned as the operator that obeys

ˆ

U

ˆ

U

−1

= 1, where 1 is the identity

operator. In matrix language this becomes UˆU

†

= 1.

[6] Sometimes, like in the inﬁnite square well, we are interested in wave functions deﬁned on a

ﬁnite segment of the line, then the correspondence becomes

n

i=1

→

_

b

a

dx and to deﬁne the

theory one must supply proper boundary conditions.

[7] It can also easily be generalized to the case of several continuous variables, which is necessary,

eg for describing a particle in three dimensional space or several particles. All one has to do is to

replace x by a set of variables ¦x

1

, x

2

, ....¦ and

_

dx by a multiple integral over these variables

_

dx

1

_

dx

2

. . . .

[8] There are also operators where the spectrum has both a discrete part and a continuous part, the

energy for a the hydrogen atom is an example, this can also be accomodated in the formalism

here with minor changes.

22

generalization is simple! Thus we need not worry about the things a mathematician would start to think about. We will just generalize the standard linear algebra to vectors whose components are functions. Why is it that the vectors of quantum mechanics have this nice property? Well, it is really something required by physics. The properties in question are namely related to the Born interpretation of the wave function as a probability amplitude, and what can go wrong is that probabilities would not add up to one. This is of course not acceptable for a physical theory, so in quantum mechanics we only use those vector spaces for which this holds true. It might then be a tricky mathematical problem to prove these properties, i.e. to prove that we are using a consistent mathematical model of a physical system. For the cases you will encounter in this course you can rest assured that these problems have been sorted out.[1] How linear algebra enters quantum mechanics will be explored and explained later, but it might be good to know from the start what to look for. Here is a brief preview and an illustrative example: 1. The state of a system is described by a vector. 2. Observables, i.e. things that can be measured, correspond to operators that implement linear transformations of the vectors . 3. A crucial feature of quantum mechanics is the principle of superposition, the linear combination aψ(x) + bφ(x) of two wave functions is a possible wave function for the system. In linear algebra this is just the basic property of a vector space, which says that a linear combination of vectors is a new vector in the space (see (6) below). Example: You have already encountered the eigenfunctions ψn (x) and energy eigenvalues

1 En = ω(n + 2 ) for the harmonic oscillator hamiltonian. It turned out to be very useful to

introduce the ladder operators, a and a† satisfying the commutation relation: [a, a† ] = 1 Their action on the normalized eigenfunctions were, a|n = a† |n = √ √ n|n − 1 n + 1|n + 1 (2) (1)

2

3 . if we allow for inﬁnite matrices and deﬁne. . Now. . . . 1 0 √ 2 0 † a = √ 3 0 . ). a= 0 1 0 √ √ 3 0 0 . . there is another type of objects that in general do not commute. 2 0 .ˆ From these we formed the number operator N = a† a. (3) it is clear (just do the multiplication!) that (1) is satisﬁed. It is not hard to prove that this is in fact not possible using ﬁnite matrices. However. The number operator takes the form. . . . Since the hamiltonaian is given by the operator relation 1 ˆ H = ω(N + 1 ) we see that the energy eigenvalue for the state ψn is En = ω(n + 2 ) as 2 expected. namely matrices. Note that all these operators are diﬀerential operators acting on wave functions ψ(x). 1 = diag(1. ˆ = N 0 1 2 3 . (5) the relations (2) are fulﬁlled. . and that the origin of the commutation relation (1) is the operator identity d x dx d = x dx + 1. if the ”1” on the LHS is interpreted as the inﬁnite diagonal unit matrix. So it is quite natural to ask whether the commutation relation (1) could be represented by matrices. If we now let these matrices act on the column vectors 1 0 0 0 1 0 0 0 ψ2 = 1 ψ0 = ψ1 = 0 0 0 0 0 0 (4) etc. 1.

Mathematically..[2] Note that a bra and ket together gives a bracket α|β (this is a physicists sense of a joke). α|β denotes the scalar product of the vectors |α and |β . n is the dimension of the vector space. i = 1. All this will be explained in detail in the following. but later we shall also discuss vector space with inﬁnitely many components. or. they describe the vector completely [3] [4]. the function ψn (x) and the inﬁnite column vector ψn are just the components of the same abstract state vector |n in diﬀerent basis. For now we shall take n to be ﬁnite. as β| (called a bra-vector or simply a bra). ψn and matrices acting upon them. DIRAC’S NOTATION A vector is denoted as |α (called a ket-vector or simply a ket).. You are probably used to the scalars being real numbers.n. The ﬁrst correspond to the position basis |x and the second to the energy basis |E . 4 . |β can be added and multiplied by scalars a. In the vector spaces that occur in quantum mechanics the scalars are. b and this gives a new vector |γ |γ = a|α + b|β (6) in the vector space. This is standard quantum mechanics notation introduced by Dirac. COMPLEX VECTORS A vector space consists of vectors and scalars. We say that these two descriptions are diﬀerent representations of the quantum mechanical system.We thus have two alternative descriptions of the harmonic oscillator states. This leads to some simple changes. and this is an example of components of a vector). 2. As noted above this is the mathematical formulation of the quantum mechanical superposition principle. as the one considered in the example in section I. II. or in terms of (inﬁnite) vectors. Two vectors |α . A vector can be expanded in a set of linearly independent basis vectors |ei . n |α = i=1 ai |ei . (7) The coeﬃcients ai are complex scalars and are called the components of the vector. in general. either in terms of wave functions ψn (x) and diﬀerential operators acting on these functions. complex (they must be since the wave function is complex. . III.

(11) ai |ei are then ai = ei |α (12) is (13) and the scalar product of |α and |β = n i=1 bi |ei n α|β = i=1 a∗ bi .e. i. |β is denoted α|β and has the properties β|α = α|β α|α ≥ 0 . the length of the vector. real and non-negative.The scalar product of two vectors |α . (16) . LINEAR TRANSFORMATIONS ˆ A linear transformation T takes a vector |α and transforms it into another vector |α = ˆ T |α in such a way that the linearity condition ˆ ˆ ˆ T (a|α + b|β ) = aT |α + bT |β 5 . Let |ei be an orthonormal basis. there is a corresponding bra n α| = i=1 ei |a∗ i . (15) Note the complex conjugation that enters on the components of α|. this follows from (8) and makes the norm. i and the square of the norm of the vector |α is n α|α = i=1 a∗ ai . IV. ∗ (8) α|α = 0 if and only if |α = 0 (9) (10) and γ|(a|α + b|β ) = a γ|α + b γ|β Note the complex conjugation that enters in the ﬁrst relation. I will frequently use the simpliﬁed notation |i = |ei . i (14) To each ket |α = n i=1 ai |ei . ei |ej = δij The components of |α = n i=1 .

the action of the ˆ operator T on an arbitrary vector can be obtained from its action on the basis vectors |i : n ˆ T |j = i=1 Tij |i . |β ). such as (19). and a natural notation is then |T β = T |β . Using this notation one ﬁnds ˆ T † α|β ˆ α|T β ˆ = β|T † α ˆ = α|T |β ∗ ∗ (22) ˆ = β|T † |α ∗ (23) which.ˆ is fulﬁlled. Because of this linearity relation.83] in the ﬁrst edition). Note that matrix equations. V. What stands inside |. T is called an operator. when inserted in Griﬃths’ eq. they form an n × n-matrix T .87] gives (the complex conjugate of) our deﬁnition (22). Then we can write |α = T |α in i matrix language as ˆ ai |i = T i j aj |j = i. but you need to know it in order to read and understand the ˆ ˆ textbook.87] T † α|β = α|T β ([3. the ˆ n × 1-matrices) with components ai and a respectively..j aj Tij |i (19) (20) (21) ai = j Tij aj a = Ta . [A. whereas |T α = T |α ). Note that ˆ ˆ ˆ ˆ T † α| is not equal to α|T † . Note order of indices! (17) (18) i ˆ Tij = i|T |j ˆ Tij are the components of the operator T . is just a name for the vector. His notation is confusing and I will try to avoid it. depend on the particular basis |i used. HERMITIAN TRANSFORMATIONS ˆ ˆ If T is an operator then one deﬁnes the hermitian conjugate operator T † as follows ˆ ˆ α|T † |β = β|T |α (for all vectors |α . ] 6 . If |α = and |α = i ai |i ai |i we introduce the notation a and a for the column vectors (i. ˆ ˆ [The deﬁnition (22) is equivalent to Griﬃths [A..e.

for any physical system corresponds to a hermitian operator. (27) Rewriting it as You have encountered this equation in your linear algebra courses. one sees that it has solutions a = 0 only if det(T − λ1) = 0. This is an eigenvalue equation. etc. momentum. (25) This can be proven using (22). .A very important type of operators in quantum mechanics are those which hermitian conjugate equals the operator itself: ˆ ˆ T† = T . [5] VI. and λ are called eigenvalues. obviously.e. (24) Such an operator is said to be hermitian or self-adjoint. n. . i. It follows form (22) that the components of a hermitian operator obey (Tji )∗ = Tij . In matrix notation this equation becomes n T a = λa or. Solving this characteristic equation (’sekularekvationen’). Note the change of order of the operators. If T happened to be a diagonal matrix then. the eigenvalues are simply the diagonal elements. EIGENVALUES AND EIGENVECTORS ˆ Given an operator T . such as e. or by using components. angular momentum. position. (26) where λ is a scalar. in components. In general. Hermitian operators are of crucial importance in quantum mechanics since any observable. consider the equation ˆ T |α = λ|α . matrix notation. which is an nth order polynomial in λ. determines the n eigenvalues λi . For a product of operators one has the useful rule ˆˆ ˆ ˆ (S T )† = T † S † . (T − λ1)a = 0.g. Vectors |α that solve this equation are ˆ ˆ called eigenvectors of the operator T since they are mapped into themselves by T . . to ﬁnd the 7 . j=1 Tij αj = λαi .. spin. energy.. The matrix T is then said to be hermitian and this condition is written as T † = T . Having obtained the eigenvalues one can determine the corresponding eigenvectors by solving (26). i = 1 .

|is . is represented by a hermitian operator and the possible results of the measurement of this observable are the eigenvalues of this operator.n in the vector space. The eigenvectors span the vector space. see Griﬃths. In quantum mechanics. All eigenvalues are real. |i with a particular eigenvalue λi then this eigenvalue is said to be non-degenerate.. 2. 2.. If all eigenvalues are non-degenerate. The property 3 is obvious if all eigenvalues are diﬀerent and can be shown to hold in general. Eigenvectors corresponding to diﬀerent eigenvalues λ = λ are orthogonal. d with eigenvalue λi this eigenvalue is said to be d-fold degenerate.eigenvalues is equivalent to diagonalizing the matrix T by a change of basis vectors. For a degenerate eigenvalue. one may have to take linear combinations of the obtained eigenvectors to obtain orthonormal basis vectors: i|j| = δij . the corresponding eigenvectors are automatically orthogonal because of 2. if there are d linearly independent eigenvectors. ˆ If T is a hermitian operator. e. After one has measured the observable and obtained one of the eigenvalues the state of the system is described by the corresponding eigenvector—this is the so called collapse of the wave function. 8 . then the following holds 1. The set of all eigenvalues {λi } and eigenvectors {|is } for an operator is called the spectrum of the operator. . s = 1 . These must be real and they are since the operator is hermitian. Item 3 means that the eigenvectors of a hermitian operator can be chosen as a basis |i . 3. . Properties 1 and 2 are simple to prove. the energy. i = 1. an observable. From now on we shall consider the non-degenerate case. T is diagonal when the basis vectors are the eigenvectors.g. If there is only one eigenvector.

This identity is very useful. (34) Note that here ”1” is an operator (and should perhaps be denoted by ˆ The identity 1).VII. 2.. if we calculate the components ˆ of the operator T itself in its own basis we get.j |i Aij j| (35) ˆ Aij = i|A|j . . In particular.j ˆ |i i|A|j j| = i. n (28) (29) (30) (31) (32) . λi real eigenvalues orthonormal basis ψi |i .. SUMMARY OF RESULTS AND THE RESOLUTION OF THE IDENTITY For a hermitian operator we have from the above ˆ ˆ T = T† ˆ T |i = λi |i |i .n basis of eigenvectors. 9 . (33) i = 1. (34) is a consequence of the vectors |i forming a basis. i=1 n i|j = δij |ψ = |ψ = i=1 ψi = i|ψ n |i i|ψ = ( i=1 |i i|)|ψ From the last equation we read oﬀ the operator identity (called the ’resolution of the identity’) n 1= i=1 |i i| . it can be inserted anywhere in an equation since it is just the identity operator and many useful formulae can be derived almost without eﬀort. ˆ Note that A does not have to be hermitian. ˆ Tij = i|T |j = λi i|j = λi δij (36) ˆ which just expresses the trivial fact that the operator T is diagonal in the basis deﬁned by its own eigenvectors. ˆ As a further example we derive the components of an arbitrary operator A in the basis {|i } ˆ deﬁned by the hermitian operator T : ˆ ˆ A = 1·A·1= i. for example the components of |ψ above.

it may happen that the scalar product of two vectors is inﬁnite. Griﬃths puts too much emphasis on the potential mathematical diﬃculties in inﬁnite dimensional vector spaces and he also stops using the Dirac notation that he has introduced. every hermitian operator in a ﬁnite dimensional vector space has a complete set of eigenvectors that can be chosen as basis vectors (and the eigenvalues are real). In my opinion. See Griﬃths for a disscussion of these problems. 10 . Here. The components of a vector ψi then becomes a function of x. for the purposes of quantum mechanics one can generalize linear algebra to inﬁnite dimensional vector spaces almost trivially without worrying about whether things are welldeﬁned or not.So far everything was simple. mathematical diﬃculties may arise. the theorems that hold for a hermitian operator in a ﬁnite dimensional vector space—that the eigenvectors span the space—do not in general hold for a hermitian operator in an inﬁnite dimensional vector space. Formally. The vector index i may run from 1 to ∞ (i. ψ(x)—thus functions can be thought of as components of vectors.e. Thus. Moreover. However. These eigenvectors may have an inﬁnite norm. As mentioned above the only new things compared to the linear algebra you have met in your mathematics course is that the vector space is complex but this was easily taken care of. it so happens that the inﬁnite dimensional vector spaces that occur in quantum mechanics are simple and behave exactly as their ﬁnite dimensional brethren! In particular. Thus inﬁnite dimensional vector spaces are mathematically much trickier than their ﬁnite dimensional counterparts. We now want to extend the theory of linear algebra to inﬁnite dimensional vector spaces. the generalization to inﬁnite dimensional vector spaces is simple. For example. n → ∞ in the formulae above) but it may also be that the vector index becomes a continuous variable x which takes for example all real values. however this inﬁnity is of the same fairly innocent type as we encountered for the free particle. However. we will use Dirac’s vector notation and trust that the mathematics works just as in the ﬁnite dimensional case as is required for the consistency of quantum mechanics. the hermitian operators that occur in quantum mechanics do have eigenvectors that span the vector space and hence can be used as a basis. By using the bra. This obscures the close analogy with the ﬁnite dimensional case. ket notation of Dirac the formulae and the calculations will look almost the same also when there is a continuum of components. Quantum mechanics guarantees that this is the case.

However. 2.. just make the replacement n → ∞ in the formulae above. (46) Note that Dirac’s delta function vanishes when x = x and is ∞ when x = x .[7] From the last equation we read oﬀ the operator identity (called the ’resolution of the identity’) ∞ 1= −∞ dx |x x| (45) Compare this to the ﬁnite dimensional case above! The resolution of the identity can be used to derive results just as above. λx real eigenvalues ∞ (39) (40) (41) (42) (43) (44) = δ(x − x ) orthonormal basis dx ψ(x)|x .∞ As noted above this generalization is trivial. IX.. but can also be any other continuum variable such as the momentum or the energy for the scattering states. THE CONTINUOUS SPECTRUM: i → x There are two new things: The sum is replaced by an integral and the Kronecker delta is replaced by Dirac’s delta function[6] n ∞ → i=1 −∞ dx (37) (38) δij → δ(x − x ) .VIII. . −∞ < x < ∞ basis of eigenvectors. Thus eigenvectors corresponding to diﬀferent eigenvalues are orthogonal.) 11 . using the deﬁning property of Dirac’s delta function dx δ(x − x )f (x ) = f (x) . −∞ ∞ |ψ = |ψ = −∞ ψ(x) = x|ψ ∞ dx |x x|ψ = −∞ dx |x x| |ψ x here can be the position of a particle. the norm of the eigenvectors is inﬁnite.. The equations (28-33) become ˆ ˆ T = T† ˆ T |x = λx |x |x x|x . (Compare the notes on the free particle. THE INFINITE DIMENSIONAL DISCRETE SPECTRUM: i = 1.

i = 1. see Griﬃths. we arrive at the important result φ|ψ = dx φ∗ (x)ψ(x) . in terms of the components ψ(x) of the vector ψ as the deﬁnition of the scalar product.(13). . thus this leads to restrictions on the allowed functions ψ(x). set of eigenvalues) consists of both a discrete part and a continuous part—the Hamiltonian for the Hydrogen atom is an example of this and so is the delta-potential well. 12 . Eq [3.. This is possible but obscures the general structure and the similarities to the ﬁnite dimensional case. i = 1.. XI. We illustrate this here with the discrete case. |φ = dx |x φ(x) (47) φ|ψ = ( = dx φ∗ (x) x|)( dx dx |x ψ(x ) (48) (49) dx φ∗ (x)ψ(x ) x|x using x|x = δ(x − x ) and the deﬁning property of the delta function (46).n be two sets of normalized basis vectors (they may for example be the eigenvectors of two operators corresponding to two physical observables). but the formulae translate immediately to the general case.e. one will have both an integral and a in the formulae. (50) Compare this expression to the ﬁnite dimensional one eq. The integral in (50) must exist. Note that Griﬃths takes the expression (50)..Comment: It is also possible to have an operator whose spectrum (i. CHANGE OF BASIS Using the resolution of the identity it is very easy to perform a change of basis. Let |ei .. THE SCALAR PRODUCT It is straightforward to derive the expression for the scalar product in terms of the components of the vectors.6] in Griﬃths: |ψ = dx |x ψ(x) . In these cases. . X.n and |fi .

m ˆ ei |fk fk |T |fm fm |ei .) XII. Using the resolution of the identity one veriﬁes that S −1 S = 1. For the components of an operator T one ﬁnds ˆ ei |T |ej = k. hence S −1 is the inverse of the matrix S. (55) In matrix notation this becomes T (e) = ST (f ) S −1 . Diagonalizing the matrix T (f ) by ﬁnding a similarity transformation S such that T (e) is diagonal is equivalent to ˆ ˆ transforming to a basis consisting of the eigenvectors of T . where ψ (e/f ) is the vector ˆ with elements ei /fi |ψ . (56) −1 where S −1 is the matrix with components Sij = fi |ej . (57) . (We assume that T corresponds to a physical observable so that a basis of eigenvectors exists. (52) A vector |ψ can be expanded as |ψ = i |ei ei |ψ = i |fi fi |ψ (53) and its components are related as ei |ψ = j ei |fj fj |ψ . The transformation (56) is called a similarity transformation. (54) again using the resolution of the identity.There are now two resolutions of the identity 1= i |ei ei | = i |fi fi | . this can be written in matrix notation as ψ (e) = Sψ (f ) . ¨ THE SCHRODINGER EQUATION The time dependent Schr¨dinger equation in Dirac notation reads o ∂ ˆ |Ψ H|Ψ = i ∂t 13 . Deﬁning the matrix S with components Sij = ei |fj . (51) Using these one ﬁnds immediately the linear transformation that relates the two sets of basis vectors |ei = j |fj fj |ei .

ψi = i|ψ .. a spin is of this form. 14 ψ(x) = x|ψ . in the standard way.. we ﬁnd the time-independent Schr¨dinger equation o ˆ H|ψ = E|ψ .. and |Ψ(t) is the state vector that contains all information about the system. (59) The above is the Schr¨dinger equation in vector form. (61) where we have deﬁned. a superconductor. y)ψ(y) = Eψ(x) where ˆ H(x. several particles. a particle moving in three dimensions. ˆ Hij = i|H|j . (58) where |ψ is t-independent. Written in this abstract vector form it applies to any physical system: a particle moving along the x-axis. (60) which gives the component equation Hij ψj = Eψi j .j i |i i|ψ . with i |i i| = 1. a molecule.ˆ where H is the Hamiltonian describing the system. y) = x|H|y . a spin. we obtain instead dx |x x|ψ . Making the ansatz |Ψ(t) = e−iEt/ |ψ . we obtain ˆ |i i|H|j j|ψ = E i. o In the continuum case. for example. a solid. (62) The Schr¨dinger equation for. (63) ˆ dy |x x|H|y y|ψ = E which gives the component equation dy H(x. In the discrete case.. (65) (64) .. with dx dx |x x| = 1. you name it! Just as we did before we can derive a time-independent equation by separing variables ˆ (we assume that H has no explicit t dependence). we can obtain it in component o form by inserting the resolution of the identity in suitable places.

.. p. dx by a multiple XIII. 2 2 The variable x in the continuum case need not be the position of a particle moving along the x-axis. x. operators for a particles motion. The completeness is intimately tied to the foundations of quantum mechanics as we will see when discussing the statistical interpretation below. All one has to do is to replace x above by a set of variables and the integral integral. ˆ ˆ Consider the eigenvalue equations x|x = x|x ˆ p|p = p|p ˆ .. p are operators. it is given in component form.. one particles motion in three dimensional space or several particles or. Sometimes it can be proven. without proof. (66) (67) Here. We see immediately that the right hand sides of the two equations agree. A general state (describing a particles motion in one dimension) can be expanded in either the |x or the |p basis: ∞ ∞ |ψ = −∞ dx |x x|ψ = −∞ dp |p p|ψ . x. p are eigenvalues and |x . everything will be the same. but in general.e. for any quantum mechanical operator. i.. The formulae can also easily be generalized to the case of several continuous variables. y) = δ(x − y)[− 2m dy2 + V (y)]. x. but can be any continuous observable. and momentum. |p are eigenvectors.d The Schr¨dinger equation [− 2m dx2 + V (x)]ψ(x) = Eψ(x) that describes the motion of o 2 2 a particle along the x-axis is of the form (64). which is needed for. Formally. we simply assume this to be true. −∞ This is always true for operators in quantum mechanics that correspond to physical observables. such as the momentum.. THE OPERATORS x AND p ˆ ˆ Here we will discuss the position. The left hand sides will d also agree if we make the identiﬁcation H(x. There is a unique ˆ ˆ eigenvector |x (|p ) for each real x (p) and they are complete: ∞ dx |x x| = 1 −∞ ∞ (68) (69) dp |p p| = 1 . for example. (70) 15 .

Let us determine the components of |x and |p in the x-basis. i. and that in its own eigenbasis. We now turn to the components of |p = dx |x x|p .) Let us recall the solution of the eigenvalue equation for p in the ˆ 16 . whereas px denotes the ˆ representation of this operator in the x-basis (ˆx = −i ∂/∂x). ψ(x) is the wave function in position space. according to the probability interpretation of the wave function. ˆ the one in (67). where we have used (66). Notice the similarity with the ﬁnite dimensional expression eq. we have |y = hence. it means that the particle can only be found at x = y. (72) and the properties of the δ-function. We deﬁned this operator in the ˆ x-basis as px ψ(x) = −i ˆ ∂ ψ(x) . For |x . ψp (x) ≡ x|p is the wave function for the |p state.This gives two wave functions that describe the system: ψ(x) = x|ψ and ψ(p) = p|ψ . (72) dx |x x|y = dx |x δ(x − y) (71) Thus. since. we use p for the abstract operator. e.(36). see the section above on this topic.g. ψ(p) on the other hand.(35). we use p also p ˆ for the latter operator. this is the usual wave function which enters the ususal Schr¨dinger equation and o which gives the probability of ﬁnding the particle at position x. We will see below that ψ(x) and ψ(p) are related by a Fourier transformation. ∂x (74) (Here. they are related by a change of basis. This makes sense.e. the wave function in x-space for the state |x is Dirac’s delta-function. These are just components of the same vector |ψ in diﬀerent bases. ψy (x) ≡ x|y = δ(x − y) . Usually. x is represented by ˆ a diagonal ’matrix’ xδ(x − y) just as in eq. The operator x becomes in x-basis ˆ x = ˆ = dx dy |x x|ˆ|y y| = x dx dy |x xδ(x − y) y| (73) dx x|x x| . is the wave function in momentum space and gives the probability to ﬁnd the particle with momentum p. To proceed we need to know what p is.

x-representation: px ψp (x) = p ψp (x) ˆ ∂ −i ψp (x) = p ψp (x) ∂x ψp (x) = A eipx/ . We can now evaluate the scalar product of |p states: p |p = 1 dx dx x |x ei(px−p x )/ 2π 1 dx eix(p−p )/ = 2π = δ(p − p) . Comments 17 . (77) √ where we have chosen A = 1/ 2π . We summarize some of the important results of this section: |y = dx δ(y − x)|x (84) (85) (86) (87) (88) 1 2π 1 dk eikx = √ 2π dk F (k) = f (0) . Thus we have |p = √ 1 2π dx |x eipx/ . −∞ (80) This is proven as follows. (79) In the last step we used the useful relation (see Problem 2. (78) (75) (76) p real . Let F (k) be the Fourier transform of f (x): 1 F (k) = √ 2π 1 f (x) = √ 2π One then ﬁnds dx f (x) from which (80) follows. (83) dx f (x)eikx dk F (k)e−ikx . (81) (82) x|y = δ(x − y) 1 dx eipx/ |x |p = √ 2π 1 eipx/ x|p = √ 2π p |p = δ(p − p) .26 in Griﬃths) 1 2π ∞ dk eikx = δ(x) . ψp (x) ≡ x|p = √ 1 eipx/ 2π .

. 18 . see the notes on the free particle. For x we have: ˆ ˆ ˆ φ|ˆ|ψ = x dx ∗ dy φ|x x|ˆ|y y|ψ = x =( dx ψ ∗ xφ)∗ = dx φ∗ (x)xψ(x) . ˆ For p.. (90) (91) φ|ˆ† |ψ ≡ ψ|ˆ|φ x x hence x is hermitian.)|∞ −∞ dx φ|ˆ|ψ . we ﬁnd ˆ φ|ˆ|ψ = p φ|ˆ |ψ ≡ p = = = † dx ψxφ∗ = φ|ˆ|ψ x dx φ∗ (x)(−i ∗ d )ψ(x) dx ∗ ∗ (92) d )φ(x) dx ψ (x)(−i ψ|ˆ|φ = p dx ∗ dψ ∗ ∗ + dx i −i ψ ∗ φ|∞ φ −∞ dx d dx φ∗ (x)(−i )ψ(x) + (.. and this is good enough. XIV. (89) We end this section by proving that the operators x and p are hermitian. Diﬀerent systems have diﬀerent Hilbert spaces. p (93) provided the wave functions go to zero fast enough at inﬁnity so that the surface term vanishes (this is the case for normalizable wave functions). We here discuss brieﬂy the properties of such a space. This shows that p is hermitian ˆ (22). • The wave functions in x and p space ψ(x) = x|ψ and ψ(p) = p|ψ are related by a Fourier transform: x|ψ = dp x|p p|ψ = √ 1 2π dp eipx/ p|ψ . see (22). THE HILBERT SPACE The set of possible vectors |ψ describing the quantum mechanical state of a particular system form a Hilbert space.• |p (as well as |x ) are non-normalizable but note that they are orthogonal and almost normalized in the sense that δij → δ(x − y): x|y = δ(x − y) and p |p = δ(p − p) (Griﬃths calls this ’delta-function normalized’).

The system is whatever we describe by quantum mechanics. the Hilbert space is L2 (−∞. for spin one 1 it is a three dimensional vector space.A space is said to be complete if all convergent series of elements in the space also belong to the space. (Note that ψ(x) are the components of the vector and that the Hilbert space here is deﬁned in terms of the components. a molecule. several particles. ∞). with a scalar product.) For one spin 1/2. particles with spin. such that ∞ dx |ψ(x)|2 < ∞ . Here we consider the general case. The system is completely described by a vector |ψ in a Hilbert space H. It can be a particle moving in one dimension.) A Hilbert space H is a vector space.4 is restricted to the motion of one particle in one dimension. (Note that this has nothing to do with the completeness of the eigenvectors to an operator discussed above. in three dimensions. Each vector in H describes a possible state for the system. the Hilbert space is a two dimensional vector space. XV. • Note that the principle of superposition is implied by this. a solid etc. THE STATISTICAL INTERPRETATION Griﬃths discussion of the statistical interpretation in Section 3. a spin. 2. −∞ (96) These functions are called square-integrable. this is the space of all functions ψ(x). 19 . (95) For a particle moving along the x-axis. that is complete: ∞ |α = j=1 aj |αj (94) is an element in H provided |αi are elements in H and the series is convergent ∞ α|α = j=1 |aj |2 < ∞ . 1. • A vector in H is called a state vector since it describes a quantum state. for N spin 1/2 it is a 2N dimensional vector space etc.

Each observable (i. p. 20 . i = 1. A measurement of the observable Q gives one of the eigenvalues λi . each one of them in state |ψ ) gives this mean value..e. p. for the generalization. if ψ|ψ = 1 .. see below. (101) i |ei ei |ψ then the probability. If |ψ = is Pi = | ei |ψ |2 ψ|ψ = | ei |ψ |2 (100) if ψ|ψ = 1. • {|ei } is a basis: ei |ej = δij .e. (99) |ei i| = 1 5. ˆ x ˆ • For the particle: Q(x. i . . The expectation value of Q is: Q = ˆ ψ|Q|ψ ψ|ψ ˆ = ψ|Q|ψ .3.): ˆ Q|ei = λi |ei ˆ • λi are real (since Q is hermitian). 4. • The probability for the system to be found in the general (normalized) state |χ is obtained by replacing |ei by |χ above. each measurable quantity) Q corresponds to a hermitian operator ˆ Q acting on the vectors in H. (97) (98) • Repeated measurements on identically prepared systems (i. Pi to get λi as a result of the measurement • Compare this to the statistical interpretation of ψ(x) = x|ψ . 2. t) → Q(ˆ. ˆ Let Q have eigenvectors |ei with eigenvalues λi (We assume here that the spectrum {λi } is discrete. t).

4]—this is very bad notation.[8] i → dx and [1] For the mathematically interested. Normally. in physics texts |0 is used to mean a vector that is not the 0-vector. Griﬃths uses the notation |0 for this 0-vector. The Principles of Quantum Mechanics. His perhaps most important contribution was the Dirac equation which combines quantum mechanics and special relativity to give a very accurate description of electrons. see [A. i. In the discussion above we assumed a discrete spectrum. • Repeated measurement of Q imediately after the ﬁrst measuremet gives the same result λi with probability one. • This is the collapse of the wave function (or rather the state vector) |ψ . a vector with the property that |α + 0 = |α for any |α . this is the vector with all components 0. more advanced texts on quantum mechanics will contain some discussion of these points. Continuous spectrum ˆ For a general hermitian operator Q with a continuous spectrum one is.6. [3] The vector space contains a null vector.e. the state of the system is |ψaf ter = |ei . in general. in other cases one simply takes this as an assumption. From the above we infer that this property is of fundamental importance in quantum mechanics—it is closely tied to the statistical interpretation of the state vector. I will denote this vector simply by 0. Immediately after the measurement of the observable Q that resulted in λi . provided one makes the standard replacements: δij → δ(x − y). 0. Avoid using |0 for the 0-vector! 21 . [2] Dirac was one of the most famous physicists of the twentieth century. Obviously. this can be proven. One assumes that all hermitian operators that correspond to physical observables have the property that their eigenvectors span the Hilbert space. but 1-6 hold also when the spectrum is continuous. not guaranteed that the eigenvectors span the space. is still recommended reading for the serious student of quantum mechanics. In some cases. His 1930 textbook.

dx by a multiple integral over these variables n i=1 → b a dx and to deﬁne the [8] There are also operators where the spectrum has both a discrete part and a continuous part. like in the inﬁnite square well. we are interested in wave functions deﬁned on a ﬁnite segment of the line. . This is OK if you know what it means but I will try to avoid this. ˆˆ where the inverse is deﬁned as the operator that obeys U U −1 = 1. ˆˆ [5] Another important type of operator in quantum mechanics are operators that obey U U † = 1.. then the correspondence becomes theory one must supply proper boundary conditions. . eg for describing a particle in three dimensional space or several particles.. the energy for a the hydrogen atom is an example. [7] It can also easily be generalized to the case of several continuous variables. which is necessary. x2 . All one has to do is to replace x by a set of variables {x1 . 22 . ˆ ˆ This means that the hermitian conjugate operator is equal to the inverse operator U † = U −1 ..[4] Griﬃths sometimes puts the vector equal to its components.} and dx1 dx2 . . where 1 is the identity operator. this can also be accomodated in the formalism here with minor changes. . U [6] Sometimes. In matrix language this becomes Uˆ † = 1.

- 3
- Maths Probability Expectation and Conditional Expectation lec6/8.pdf
- Complex Numbers
- ch5aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
- Roots and Weights
- Gudmundsson - An Introduction to Gaussian Geometry - 1.010 - Jan 2009
- Complex v1
- Convolution
- sequences
- Pouliot
- Product Notation
- Chartrand 3e Chapter15
- MATH219 Lecture 6
- Complex Numbers _ Mathematics of the DFT
- MT5ComplexBook_Feb2013
- Generating Functions
- Linear differential Equations
- Unitary Matrices
- eigen
- Absolute Primes
- 10.1.1.53.7776
- Math 891
- Advanced PDEs - Results and Definitions
- ch2
- Hermite Polynomials
- Chapter-1
- Polynomial Invariant Theory of the Classical Groups
- EDP
- algebraic geometry notes
- Momkus (1)

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd