You are on page 1of 14

8.

Transformation Theory, January 4, 2010

Chapter 8. Transformation Theory


Section 8.1. General theory 1 Introduction. We can describe the state of a system in a bound state | with the help of a complete, orthonormal basis set {|1 , |2 , . . . , |N } by using
N N

| =

i=1

|i i |

i=1

i |i

(1)

can be represented by Any operator A = A


N N

i=1 j =1

j |i i | A

j |

i=1 j =1

|i Aij j |

(2)

Suppose that I nd it advantageous to use two complete, orthonormal N basis sets, {|i }N i=1 and {|i }i=1 . Why would I do that? Perhaps the problem was easy to formulate by using the basis set {|i }N i=1 , but later I decided that the numerical evaluations are easier to perform with the basis set {|i }N i=1 . Or it might be easy to perform numerical calculations by using N {|i }i=1 but the physical interpretation is more transparent when {|i }N i=1 is used. Several questions are immediately apparent. How are the two basis sets through related? In other words, if I dene the operator U |i , i = 1, 2, . . . , N, |i U what are its properties? If a ket | is represented by
N

(3)

| = and by

i=1

i |i

(4)

| = where

i=1

i |i ,

(5)

i i | and i i | ,

(6) (7)

8. Transformation Theory, January 4, 2010

what is the relationship of i with i ? is represented by Finally, if an operator A = A and by = A where | Aij i | A j and | j , Aij i | A (11) how are the matrix elements Aij and Aij related? Kets and operators have multiple personalities and can take an innite number of representations. Are all of them consistent with each other? Do these multiple mathematical representations all lead to identical physical results? They had better, since otherwise the formalism is useless. must be unitary. In what follows, I make extensive use 2 The operator U of the properties of unitary and adjoint operators so I remind you of some of their properties derived in Chapter 4A. is unitary if and only if 1. U = U 1 U (12) (10)
N N i=1 j =1 N N

i=1 j =1

|i Aij j |

(8)

|i Aij j |,

(9)

and U 1 is the inverse of U is the Hermitian conjugate of U where U (a unitary operator always has an inverse). of an operator A satises the equations 2. The Hermitian conjugate A | = | A A and | = | A A for any kets | and | . (14) (13)

8. Transformation Theory, January 4, 2010

N 3. If {|i }N i=1 and {|i }i=1 are two complete, orthonormal basis sets, then dened by the operator U

|i |i , i = 1, 2, . . . , N U is unitary.

(15)

4. If {|i }N i=1 is a complete, orthonormal basis set and U is a unitary operator, then the set {|i }N i=1 dened by |i , i = 1, 2, . . . , N |i = U is a complete, orthonormal basis set. (16)

Exercise 1 An orthonormal set {|i }N i=1 is complete if and only if i | y = 0 for all i (17) |i , implies that |y = 0. Use the denition to show that the set |i = U N is unitary. i = 1, 2, . . . , N , is complete if {|i }i=1 is complete and U Exercise 2 Prove that the following statements are equivalent for an orthonormal set {|i }N i=1 : (a) {|i }N i=1 is complete. (b) For any ket |y ,
N i=1

i | y |i = |y .

(c) For any ket |y , if i | y = 0, i = 1, 2, . . . , N , then |y = 0. 3 After this brief review, we can return to the questions formulated in 1. Our rst result is the following.
N If {|i }N i=1 and {|i }i=1 are two complete, orthonormal basis sets, then the unitary operator U dened by

|i = |i , i = 1, . . . , N U

(18)

8. Transformation Theory, January 4, 2010

can be written as = U

N i=1

|i i |

(19)

, as given by Eq. 19, on |j : The proof is straightforward. Act with U |j = U


N i=1 N

|i i | j =

i=1

|i ij = |j

(20)

given by Eq. 19 has the property (Eq. 18) that denes U . This shows that U 4 The connection between i and i . Let us represent a ket | with N two complete, orthonormal basis sets {|i }N i=1 and {|i }i=1 that are related through |i |i = U (21) The two representations of a ket | are
N N

| = and | =

i=1 N i=1

|i i |

i=1 N

i |i

(22)

|i i |

i=1

i |i

(23)

The connection between i and i is calculated as follows. j = = = Here in the basis set {|i }N . is the matrix element of the operator U i=1 Using the notation of matrix algebra, 1 . = . . , N

j | N i=1 i j | i N i=1 i j | U i N i=1 Uji i

(used Eq. 7) (used Eq. 23 for | ) (used Eq. 21 for i )

(24)

| i Uji j | U

(25)

1 . = . . N

(26)

8. Transformation Theory, January 4, 2010

and

we can write Eq. 24 as


U =

U11 U21 . . .

U12 U22 . . .

UN 1 UN 2 UN N

U1N U2N . . .

(27)

1 . . . = U N

U11 U21 . . .

U12 U22 . . .

UN 1 UN 2 UN N

U1N U2N . . .

1 2 . . . N

(28)

The matrix elements Uij can be calculated from our knowledge of the two basis sets. Using Eq. 19 we have Uij = = =
=1

| j i | U
N

i |
N

=1

| | j (29)

i | j = i | j

hence Uij = i | j (30) We know the two basis sets and therefore we can calculate the matrix elements Uij from Eq. 30. This allows us to use Eq. 28 to calculate from . Or, taking advantage of the fact that U 1 = U , we can turn Eq. 28 into = U which can be used to calculate from . Exercise 3 Show that j =
i

(31)

| i i j | U

8. Transformation Theory, January 4, 2010

5 The connection between the matrix elements Aij and Aij . An operator can be represented as A = A or as = A where Aij Aij Aij = = = | j i | A | |A
i j N N

i=1 j =1 N N

|i Aij j |

(32)

i=1 j =1

|i Aij j |

(33)

(34) (35)

We can calculate the connection between Aij and Aij as follows. | i | A j U i | AU | j

(used Eq. 21) (used Eq. 14) (36)

A U | j i | U

in the basis set {| }N are equal to the matrix The matrix elements Aij of A i i=1 A U in the basis set {|i }N elements of U . i=1 = N By inserting I | | between the operators, we can turn Eq. 36 =1 into a matrix equation:
N N

Aij =
=1 =1 N N

| | A | i | U (U )i A U j

| j | U (37)

=
=1 =1

We can write Eq. 37 as A = U AU (38)

Exercise 4 Show that


N N

Aij =
=1 =1

(U )i A U j

(39)

8. Transformation Theory, January 4, 2010

6 Summary. We can now summarize our results. We can use two orthonormal, complete basis sets {|1 , |2 , . . . , |N } and {|1 , |2 , . . . , |N } dened by to represent kets or operators. The operator U |i , i = 1, 2, . . . , N |i = U is unitary. We can represent a ket by
N

(40) (41)

(42)

| = or by

i=1

i |i

(43)

| = by and an operator A = A or by = A We have i =


N N

i=1

i |i

(44)

i=1 j =1

|i Aij j |

(45)

i=1 j =1

|i Aij j |

(46)

Uij j
j =1

(47)

| j , and where Uij = i | U | A = (U AU )ij = i | A j ij


N N

(U )i A U j
=1 =1

(48)

8. Transformation Theory, January 4, 2010

7 The physics must be the same. It appears that we can take an innite number of basis sets and obtain an innite number of representations of a ket or an operator. This is a bit disconcerting but not tragic: kets and operators are not observables and we have no reason to demand that they have a unique representation. However, the magnitude of quantities that are measurable must have values independent of representation. Let us show that this is the case. We start with the probability amplitude an | where | is the state of the system and |an is a pure state of an observable A. The probability that a measurement of A gives the result an is P (an ) = | an | |2 (49)

The magnitude of P (an ) must be the same regardless of representation. Suppose | is represented by
N

| = and by

i=1

i |i

(50)

| =

i=1

i |i

(51)

If we use the basis set {|i }N i=1 to calculate an | , we have


N

an | = If we use the basis set

i=1

i an | i

(52)

|i }N {|i U i=1 , the same amplitude is written as


N

(53)

an | =

i=1

i an | i

(54)

The prime in an | reminds me that this is an | calculated with the basis set {|i }N i=1 . I will show that an | is identical to an | : changing the basis set does not change the probability amplitude.

8. Transformation Theory, January 4, 2010

I can write an | as an | = N i=1 an | i i | = N i=1 an | U i U i | N an | i i | U = i=1 U an | U = an | U U = an | = U (used (used (used (used Eq. 6) Eq. 53) N i=1 |i i | = I ) U = U and Eq. 12)

Thus we see that an | and an | are equal: the probability amplitude is independent of the basis set used to represent | . That is the way it should be! | , the average value of Let us look at another quantity: A | A an observable A when the system is in the state | . If we use the basis set {|i }N i=1 to represent | , we have A

| = = |A

i=1 j =1

| j j i i | A

(55)

If we use Eq. 44 to represent | in the basis set {|i }N i=1 then | A | becomes N N | j j A= i i | A (56)
i=1 j =1

Again, the prime reminds me that A is calculated by representing | in the basis set {|i }N i=1 . But, starting with Eq. 56, we have A

= = = = = = =

N i=1 N i=1 N i=1 N i=1

N j =1 N j =1 N j =1 N j =1

| j j | i | i | A i | U i | A |U j U j | U i U i | A |U j j | U |U j | U

| i i | U A U | j U U I U U | IU A U A U U |U | = A |A

(used (used (used and (used (used (used (used

i = i | , j = j | ) |i ) |i = U a|b = b|a j | = j | U ) U i = U | i ) |U k |k k | = I ) | y = | Uy ) U U = I ) U

The mean value calculated with the basis set {|i }N i=1 is the same as the one calculated with the basis set {|i }N . i=1 We expect that the eigenvalues of an operator representing an observable A should be independent of the basis set since the eigenvalues are the results

8. Transformation Theory, January 4, 2010

10

of measuring A. They cannot depend on our decision to use one basis set or another. We have shown that in the basis set {|i }N i=1 , the operator A is represent by the matrix A with elements Aij = i | A | j . This matrix is equal to A = U AU (57) | j represents the where the matrix A having the elements Aij i | A N is a unitary operator dened by | = operator in the basis set {|i }i=1 . U i U |i . I will show that the eigenvalues of the matrix A are the same as those of A. Let (i) be an eigenvector of the matrix A corresponding to the eigenvalue i , so that A (i) = i (i) (58) Since the matrix is represented in the basis {|i }N i=1 , the vector (i) consists of the coordinates of the eigenvector in the same basis. Using Eq. 57, we have U AU (i) = i (i) (59)

Since U is unitary, we know that U = U 1 and U U = I . Acting with U on Eq. 59 gives (60) U U AU (i) = U (i (i)) = i (U (i)) We nd that A(U (i)) = i (U (i)). This shows that if (i) is an eigenvector of A then U (i) is an eigenvector of A corresponding to the same eigenvalue. In particular, if i is an eigenvalue of A then i is an eigenvalue of A. This is what we expect if the theory is well constructed. is an arbitrary operator and U is an operator Exercise 5 Suppose that A 1 1 that has an inverse, U . Prove that U AU and A have the same eigenvalues U corresponding to eigenvalue , 1 A and that if | is an eigenvector of U | is an eigenvector of A corresponding to the eigenvalue . We say then U 1 that U AU is obtained from A by a similarity transformation.

8. Transformation Theory, January 4, 2010

11

Section 8.2. Two applications 8 Introduction. Transformation theory points out that there are many equivalent descriptions of a quantum process. They are interesting either because they illuminate dierent aspects in the behavior of the system, or because some representations make certain calculations simpler than other representations. In this section I show that you have already encountered applications of this theory. When you studied the hydrogen atom, you learned that the angular dependence of the wave function is given by the spherical harmonics Y m (, ) (see Metiu, Quantum Mechanics, p. 317). You were also told that for the purpose of describing the chemical bond, it is more useful to use the orbitals npx , npy , and npz . I will show here that the set 2px , 2py , 2pz , for n = 2, is obtained from Y1m (, ) according to transformation theory: the matrix converting Y11 , Y10 , Y11 into 2px , 2py , 2pz is unitary, and therefore {2px , 2py , 2pz } and {Y11 , Y10 , Y11 } are equivalent orthonormal basis sets. For example, the z for j = 1 can be calculated with either set to obtain the eigenvalues of L same results. The calculation is much easier in the basis set {Y11 , Y10 , Y11 }. Moreover, that set is more convenient when we examine the spectroscopy of the atom. However, the basis {2px , 2py , 2pz } is more convenient when we discuss the chemical bond in a diatomic. You encountered another example of transformation theory in general chemistry. You were told that the sp3 hybrid orbitals are more appropriate for describing the valence electrons of the carbon atom than are {2s, 2px , 2py , 2pz }. The reason: they point towards the corners of a tetrahedron, just like the bonds in CH4 (or other saturated bonds involving the carbon atom in organic compounds). I will show here that the sp3 orbitals are obtained by a unitary transformation of the vector {2s, 2px , 2py , 2pz }. In accordance with transformation theory, these two sets are equivalent. 9 2px , 2py , and 2pz . The orbitals 2px , 2py , and 2pz are dened by (see Metiu, Quantum Mechanics ) 1 2px (, ) = [Y11 (, ) Y11 (, )] 2 i 2py (, ) = [Y11 (, ) + Y11 (, )] 2 0 2pz (, ) = Y1 (, ) (61) (62) (63)

8. Transformation Theory, January 4, 2010

12

You can think of Yjm (, ) as the spherical coordinate representation of the abstract ket |j, m . Similarly, 2px (, ) is the spherical coordinate representation of the abstract ket |2px . The kets |2py and |2pz have similar interpretations. Because Eqs. 6163 contain only spherical harmonics having j = 1, the presence of j in the notation |j, m is superuous: Ill use instead |m , with m = 1, 0, or 1. In the abstract ket space, Eqs. 6163 become |2px |2py |2pz I can write this as

1 = (| 1 |1 ) 2 i = (| 1 + |1 ) 2 = |0
1 2 i 2

(64) (65) (66)


In Cell 1 of WorkBook8 Transformation theory.nb, I veried that indeed the matrix U dened by Eq. 67 does the conversion of {|1 , |0 , | 1 } to {|2px , |2py , |2pz } correctly and that the absolute value of all its eigenvalues are equal to 1. This proves that the matrix U is unitary. According to the theory developed in this chapter, any measurable quantity can be calculated by using {| 1 , |0 , |1 }1 as a basis set or by using z, {|2px , |2py , |2pz }. As an example, let us calculate the eigenstates of L by using these two basis sets. For this calculation, the basis set {| 1 , |0 , |1 } is particularly advantageous because z |m = h L m|m , m = 1, 0, 1 (68) z . The z , hence they are eigenstates of L The states |m are pure states of L o-diagonal matrix elements of Lz in this basis are all equal to zero. For example, z | 1 1|L = h 1| 1 =0 (used Eq. 68) (used orthogonality, 1 | 1 = 0) (69)

1 0 |2px 2 i 0 |2py = 2 |2pz 0 0

|1 |1 |0 |0 U |1 |1 1

(67)

The diagonal elements are given by z | m = h m|L m, m = 1, 0, 1 (70)

8. Transformation Theory, January 4, 2010

13

z in this basis is therefore The matrix of L

The eigenvalues are h , 0, and h . The matrix of Lz in the basis {|2px , |2py , |2pz } can be calculated by using Eqs. 68 and 6466. For example, 1 i z | 2py = z( 2px | L ( 1| 1|)L )(| 1 + |1 ) (used Eqs. 6466) 2 2 i z | 1 ) z | m = 0 if m = m ) = 2 ( 1 | Lz | 1 1 | L (used m | L i z | m = 1) =2 (h h ) (used Eq. 68 and m | L = h i (72) I calculated the matrix elements 2px | Lz | 2px , 2px | Lz | 2py , etc., in Cell 2b of WorkBook8 Transformation theory.nb. The matrix Lz , corre z , in the basis set {|2px , |2py , |2pz } is sponding to the operator L Lz = 2px 2py 2pz 2px 2py 2p z 0 i h 0 h 0 0 i 0 0 0

h 0 0 Lz = 0 0 0 0 0 h

(71)

The eigenvalues of this matrix are (see Cell 2b of WorkBook8) h , h , and 0. They are exactly the same as the eigenvalues calculated in the basis {| 1 , |0 , |1 }, as theory says they should be.

10 sp3 orbitals. The carbon atom has four valence electrons in the orbitals {2s, 2px , 2py , 2pz }. These have the same energy as the orbitals {Y00 , Y10 , Y11 , Y11 } z . Which one should we use if we want to but they are not eigenstates of L construct a simple theory of the chemical bonds in methane? It turns out that neither is as good as the set {h1 , h2 , h3 , h4 } of hybrid orbitals. These are dened by 1 h1 = (2s + 2px + 2py + 2pz ) (73) 2 1 (74) (2s + 2px 2py 2pz ) h2 = 2 1 h3 = (2s 2px + 2py 2pz ) (75) 2 1 h4 = (2s 2px 2py + 2pz ) (76) 2

8. Transformation Theory, January 4, 2010

14

The conversion from {2s, 2px , 2py , 2pz } to {h1 , h2 , h3 , h4 } is performed by the matrix U dened by

h1 h2 h3 h4

1 = 2

1 1 1 1 2s 1 1 1 1 2px 1 1 1 1 2py 2pz 1 1 1 1

2s 2px 2py 2pz

I show in Cell 3 of WorkBook8 Transformation theory.nb that U does the job (i.e. gives Eqs. 7376) and is unitary (the absolute values of its eigenvalues are equal to 1). Since the set {2s, 2px , 2py , 2pz } is orthonormal and U is unitary, {h1 , h2 , h3 , h4 } is orthonormal. Either set can be used for calculating the properties of the atom. However, h1 , h2 , h3 , and h4 have a special property: they point towards the corners of a tetrahedron that has the carbon atom at its center. It so happens that the H atoms in CH4 are also located at the corners of a tetrahedron with the C atom in the center. The orbitals h1 , h2 , h3 , h4 are much better for building a theory of bonding in methane than are 2s, 2px , 2py , 2pz . This statement should puzzle you. We spend quite a bit of time proving that two representations using two basis sets connected by a unitary transformation are completely equivalent. How, then, can one set be better than the other? The dierence is qualitative. The hybrid orbitals explain why carbon atoms form tetrahedral bonds. The other, equivalent, basis sets do not. However, if the structure of methane is calculated with any of the equivalent basis sets, the results will be the same. While hybrid orbitals are convenient for explaining the coordination of the carbon atoms, they are very cumbersome to use in explaining spectroscopy. Exercise 6 Find the matrix that converts {Y00 , Y11 , Y10 , Y11 } into {h1 , h2 , h3 , h4 }. Prove that it is unitary. Explain the consequences of this fact. z by using the basis set {h1 , h2 , h3 , Exercise 7 Find the eigenvalues of L h4 }.

You might also like