Quantum Mechanics in Hilbert Spaces: 1.1 The Abstract Hilbert Space
Quantum Mechanics in Hilbert Spaces: 1.1 The Abstract Hilbert Space
Quantum Mechanics in Hilbert Spaces: 1.1 The Abstract Hilbert Space
j i+ j 'i = j + 'i 2 H
(1.1)
aj i =ja i 2H
(1.2)
h' j i 2 C :
(1.3)
j i+ j 'i = j 'i+ j i
2
(1.4)
The last two relations state the existence of a 0-vector and the existence of a negative
vector with respect to j i.
1j i = j i
a(b j i) = (ab) j i
(a + b) j i = a j i + b j i
a(j i+ j 'i) = a j i + a j 'i
(1.5)
II. With respect to the scalar product, H is a unitary vector space
h j i0
and
h j i=0 )j i=0
h' j i = h j 'i
h' j a i = ah' j i
h' j 1 + 2 i = h' j 1 i + h' j 2i
Because of (1.6) a Norm can be dened
k k= h j i ;
(1.6)
(1.7)
(1.8)
(1.9)
where the specic characteristics of the norm depend on the vector space.
One has
j h' j i j k ' k k k
k'+ k k'k + k k
(1.10)
and
ha' j i = ah' j i :
3
(1.11)
(1.12)
If fj 'nig is dense in H, we can expand every arbitrary vector according to this basis
j i=
and thus
h'm j i =
1
X
n=1
1
X
n=1
j 'n i a n
h'm j 'ni an =
1
X
n=1
(1.13)
mn an = am
(1.14)
j i=
1
X
n=1
j 'nih'n j i :
(1.15)
The relation (1.15) is only valid if fj 'nig is a complete set, and we have the completeness
relation
k k =
2
1
X
n=1
j h'n j i j2 :
h j i=h j1j i =
allows to represent
1=
1
X
n=1
1
X
n=1
h j 'nih'n j i
j 'nih'n j ;
4
(1.16)
(1.17)
(1.18)
which is called the spectral representation of the 1-operator. The representation of the
scalar product
hx j i =
1
X
n=1
hx j 'nih'n j i
(1.19)
through the components h'n j i and h'n j xi of the vectors j i and j xi with respect to
the orthonormal system fj 'nig obtained in (1.19) through "insertion" of the 1-operator
as given in (1.18). The chosen Bra-(h' j) and Ket-(j 'i) notation allows the representation
of vectors in orthonormal basis systems in a very economical way.
A j i =j A i =j 0 i
(1.20)
A(a j i + b j 'i) = aA j i + bA j i :
(1.21)
h' j A j i = hAy' j i
(1.22)
Ay, in general dierent from A, is called the to A adjoint operator if (1.22) is fullled. If
DA is dense in H; Ay is uniquely dened.
h' j A j i = hA' j i
5
(1.23)
Ay = A
(1.24)
at least on the range DA of A. It could be that Ay dened via (1.22) exists on a larger
range DA DA. If this is not the case, i.e., DA = DA , then A is called self-adjoint. It
should be mentioned that a hermitian operator only has a complete set of eigenvectors if
it is self-adjoint. Since eigenvectors (or eigenfunctions) characterize quantum mechanical
systems, the requirement that operators are self-adjoint (not only hermitian), if they are
supported to characterize physical observables, is quite relevant.
y
(1.25)
hAi = h h j Aj j i i = h j A j i :
(1.26)
h j A j i = hA j i = h j A i
from which follows that for hermitian operator A the quantity h j A j i is real.
If
A j ai = a j ai
(1.27)
ai
0
h a j ai = 0
0
for a0 6= a.
6
(1.28)
Proof: From
A j ai = a j ai
(1.29)
h a j A j a i = ah a j a i
(1.30)
follows
0
and
hA a j a i = a0h a j a i = ah a j a i
(1.31)
and thus
j ai = 0 :
(1.32)
Since according to the assumption a0 6= a, i.e., (a0 ? a) 6= 0, follows that h a j a i = 0.
(a0 ? a)h
h
' j
i = h' j i :
(1.33)
h' j
y
j i = h' j i = h' j 1 j i ;
(1.34)
y
= 1 :
(1.35)
Because of
they obviously fulll
In nite dimensional vector spaces, the relation (1.35) would imply
y = 1. However,
for innite dimensional vector spaces this is in general not the case. If an operator U
fullls
U yU = UU y = 1
(1.36)
U y = U ?1 :
7
(1.37)
A j i = j 0i
(1.38)
uses the representation of the 1-operator (1.18) and multiplies from the left with h'm j,
one obtains
X
n
(1.39)
Here h'n j i are the Fourier components of the expansion (1.15) of j i with respect to
a complete orthonormal system fj 'nig. If one denes
X
n
Amnan = a0m
(1.40)
(1.41)
which is the form of a linear map represented by matrices. Introducing a basis in the
abstract Hilbertspace assigns each vector a column vector of Fourier coecients:
0 h' j i 1 0 a 1
1
B
C BB a12 CC
h
'
B
1 j i C
j i ?! BB@ .. CCA BB@ ... CCA
.
an
h'n j i
(1.42)
(1.43)
1.4 "A"-Representation
The eigenvectors j 'ni of a self-adjoint operator A
A j ' n i = a n j 'n i
8
(1.44)
for distinct eigenvalues an do in general not form a complete orthonormal system. If they
do form a complete orthonormal system fj n ig, then one can use those vectors as basis
vectors for representing other operators.
In this particular case one can assign to an arbitrary map
B j i = j B i = j 0i
(1.45)
the representation
X
n
(1.46)
p(t) ?! P (t)
x(t) ?! X (t)
(1.48)
A = A(P; X ) ;
e.g., the Hamiltonian of the harmonic oscillator is given as
2
H = 2Pm + m2 !2X 2 :
9
(1.49)
(1.50)
(1.51)
A quantum mechanical state is characterized by a Hilbert space vector j i. The expectation values of operators in such a state are given as (provided k k= 1)
hAi = h j A j i :
(1.52)
The time dependence of an operator A is given through the Hamiltonian H (p; x) via
(1.53)
A_ = hi [H; A] :
hAi = h j A j i
(1.54)
(with k k= 1). The root-mean square deviation from this expectation value is given by
the mean value (A ? h j A j i2), i.e., through the non-negative expression
(A)2 = h j (A ? h j A j i)2 j i 0 :
(1.55)
Its square root (A) is called root-mean-square deviation or standard deviation. From
(1.55) follows
(A)2 = h j A2 j i ? 2h j Ah j A j i j i + h j A j i2
(1.56)
(A)2 = h j A2 j i ? h j A j i2 :
(1.57)
and thus
With this denition, one proves an essential theorem in quantum mechanics
A j ai = a j ai ;
(1.58)
h a j A j ai = a ;
(1.59)
then
10
(A)2 a = h a j (A ? a)2 j a i = 0 :
(1.60)
(1.63)
(H ) n = 0 ;
then, according to the above theorem, one has
H j n i = En j n i ;
(1.64)
i.e., bound states are obtained by solving the eigenvalue equation for the Hamiltonian H .
(1.65)
Here A h j A j i. The left-hand side of (1.65) can be estimated from below via
j Imh j (A ? A)(B ? B ) j ij j h j (A ? A)(B ? B ) j i j :
(1.66)
One also has
j (A ? A)(B ? B ) j i j
= 1 j h j (A ? A)(B ? B ) j i ? h(A ? A)(B ? B ) j i j
2
1
j h j (A ? A)(B ? B ) ? (B ? B )(A ? A) j i j
=
2
= 21 j h j AB ? BA j i j
j Im h
(1.67)
(1.69)
12
Chapter 2
Symmetries I
2.1 Constants of Motion
According to (1.53) the time dependence of an operator A is given through the Hamiltonian H (P; X ) via
A_ = @A
= i [H; A]
(2.1)
@t h
The observable A is called constant of motion of the system, if
(i) A is compatible with H, i.e.
(ii)
[H; A] = 0
(2.2)
@A = 0
@t
(2.3)
Condition (ii) states that A does not have any explicit time dependence. (More on time
dependence later.)
In general, symmetries or invariance properties lead to conservation laws. There are two
distinct kinds of symmetries:
discrete symmetries and continuous symmetries.
Let an innitesimal unitary transformation depend on a real parameter " and vary
only slightly from unity
^
U^"(G^ ) = 1 + i"G;
(2.4)
13
G^ = G^ y:
(2.5)
(2.6)
(2.7)
(2.8)
N
Y
G^ ) = lim (1 + iG^ )N = eiG^ :
U^ (G^ ) = Nlim
(
1
+
i
N !1
!1
N
k=1
(2.9)
iG^ y ?iG^ iG^ ?1
=e
= e
:
e
(2.10)
^ A] = 0.
only if [G;
Consider the following examples:
1. Time Translation
14
(2.11)
Set
G^ h1 H
" = t
^Ut = 1 + i tH
h
(2.12)
2. Space Translation
Consider for simplication one-dimensional space translations in x-direction. Then
G^ h1 P^x
U^" = 1 + hi "P^x
(2.15)
Application on a state vector yields
i
1 + h "P^x j (x)i = j (x)i + hi "P^xj (x)i
@ j (x)i
= j (x)i + " @x
j (x + "i;
(2.16)
where the coordinate space representation P^x = ?ih @x@ was used. For the operator X one
gets
i i
0
?
1 ^
^
^
^
X = U"(Px)X U" (Px) = 1 + h "P^x X 1 ? h "P^x
X + hi "[P^x; X ]
= X + ":
(2.17)
(2.18)
(2.19)
does not explicity depend on the time, i.e. G^ is a constant of motion. That is
d G^ = 1 [[G;
^ H ] + @ G^ = 0:
dt
ih
@t
(2.20)
2.2 Inversion
The rst example of a discrete symmetry is inversion.
(2.21)
(2.22)
P2 ,
2m
which is given as
P 2 = ? h 2 r2 = ? h ( @ 2 + @ 2 + @ 2 ):
(2.24)
2m
2m
2m @x2 @y2 @z2
Obviously, the kinetic energy operator is invariant if (x; y; z) ! (?x; ?y; ?z). If one
writes the Hamiltonian as
2
~ P~ 2) = P + V (X~ )
H (X;
(2.25)
2m
16
one only needs to consider the transformation properties of V (X~ ). Any central potential
V (X~ ) V (jjX~ jj) is invariant under inversion. A potential of the form
(2.26)
(2.28)
(2.29)
(2.30)
Since (2.30) holds for any ji, the operator P has to fulll
P 2 = 1:
(2.31)
Let be the eigenvalue of the operator P , then P 2 must have the eigenvalue 2 = 1.
Thus
P(~x) = (~x)
P(~x) = (?~x);
from which follows
(?~x) = (~x);
i.e. parity eigenstates with eigenvalue +1 (?1) are even (odd) functions of ~x.
>From P 2 = 1 follows
P ?1 = P
The specic choice of (~x) H (~x) (~x) gives
P(~x) = PH (~x; ::) (~x) = H (?~x; ::) (?~x)
= PH (~x; ::)P ?1P (~x) = PH (~x; ::)P ?1 (?~x):
17
(2.32)
(2.33)
(2.34)
(2.35)
(2.36)
which leads to
(2.38)
(2.41)
This means that the eigenfunctions to a non-degenerate eigenvalue E are either even or
odd functions of ~x, having either even ( = +1) or odd ( = ?1) parity.
a (~x) + b (?~x)
(2.42)
is also eigenfunction of H (~x; ::) to the same eigenvalue En. One can use this freedom to
choose linear combinations which are even or odd parity states, i.e.
(~x) (?~x)
18
(2.43)
(2.47)
Due to the operator character of A, the factorization of the Hamiltonian is slightly more
complicated, and one obtains an additive term 12 ;
H = 21 (P 2 + Q2) = Ay A + 12 1 :
(2.48)
The same factorization can be applied to the dierential equation for the harmonic oscillator
d2 x(t) + !2 x(t) = 0 ;
(2.49)
dt2
where the dierential operator can be decomposed as
d2 + !2 = d ? i!
dt2
dt
!
d + i! :
dt
(2.50)
Thus one obtains solutions to (2.49) if one solves one of the following rst-order dierential
equations:
!
!
d ? i! x(t) = 0 or
d + i! x(t) = 0 :
(2.51)
dt
dt
19
and
x(t) = x0 e?i!t :
(2.52)
Finally, there is a specic symmetry property in (2.44) if one considers that in the (x; p)
phase space Hclass describes the "length" of a vector in that space. Thus, Hclass does not
change if one carries out a rotation in the phase space given by
x0 = x cos + p sin
p0 = ?x sin + p cos :
(2.53)
This rotation leaves Hclass invariant, i.e., Hclass(x0 ; p0) = Hclass (x; p). For the quantities
a from (2.45), the corresponding symmetry transformation is given by
a 0 = ei a and a 0 = e?i a :
(2.54)
The analogous symmetry carries over into quantum mechanics, since N = AyA is invariant
under
A0 = ei A and Ay = e?i Ay :
(2.55)
According to the general symmetry principle in quantum mechanics, this symmetry operation must be generated by a unitary operator. Thus, a unitary operator U () must
exist, which carried A into A0,
A0 = U ()y A U () :
(2.56)
(2.57)
(2.58)
Important is that the transformation introduced in (2.56) forms a group, i.e., carrying
out transformations of this kind gives another one of the same kind,
U () U ( ) = U ( + ) :
(2.59)
Especially, the inverse element exists:
U (?) = U ?1 () :
(2.60)
20
The geometric interpretation of this group is obvious if one considers the (x; p) form of the
transformation in phase space. The operator U () describes a rotation in the (x; p) plane
and the group is called O(2) orthogonal group in two dimensions. If one considers
the complex form of (2.55), the underlying group is the unitary group with dimension
d = 1; U (1).
One obtains important insight into the quantum theory if one explicitly constructs the
operator U (). This procedure follows the general rule that one considers rst small
values of the parameter , i.e., so-called innitesimal transformations.
Because of (2.58), one can expand
X () = Y +
and
() = 1 + iY +
(2.61)
A0 = (1 ? iY ) A(1 + iY ) A + i (AY ? Y A) :
(2.62)
Thus
A0 = A + i [A; Y ] + : : :
(2.63)
This means that the innitesimal transformation is determined by the commutator [A; Y ].
Thus, Y is called the innitesimal generator of the group (here U (1)). Up to now we
have not used any specic properties of the group U (1). This comes into play when we
expand (2.55) in powers of
A0 = ei A A + i A + : : :
(2.64)
(2.65)
(2.66)
Y = N:
Thus, the number operator N is the innitesimal generator of the group U (1).
21
(2.67)
U () = e?iN
(2.68)
is an exact representation.
The results derived for the harmonic oscillator have numerous applications in modern
physics, since transformations of the form (2.56) and operators with the properties of
N appear in many areas of theoretical physics. Examples are particle number, electric
charge, baryon number, strangeness, etc. In all cases, the quantities are "quantized,"
i.e., have discrete values, which are, in general, integers. Thus, we have a paradigm for
understanding quantum numbers.
22