Comments
Description
Transcript
Choosing an appropriate formalism
REPRESENTATION THEORY Finally, for ozone, which is angular rather than linear, symmetry does not place such tight constraints. A dipole-moment component parallel to the axis BB (figure 29.1(c)) is possible, since there is no symmetry operation that reverses the component in that direction and at the same time carries the molecule into an indistinguishable copy of itself. However, a dipole moment perpendicular to BB is not possible, since a rotation of π about BB would both reverse any such component and carry the ozone molecule into itself – two contradictory conclusions unless the component is zero. In summary, symmetry requirements appear in the form that some or all components of permanent electric dipoles in molecules are forbidden; they do not show that the other components do exist, only that they may. The greater the symmetry of the molecule, the tighter the restrictions on potentially non-zero components of its dipole moment. In section 23.11 other, more complicated, physical situations will be analysed using results derived from representation theory. In anticipation of these results, and since it may help the reader to understand where the developments in the next nine sections are leading, we make here a broad, powerful, but rather formal, statement as follows. If a physical system is such that after the application of particular rotations or reflections (or a combination of the two) the final system is indistinguishable from the original system then its behaviour, and hence the functions that describe its behaviour, must have the corresponding property of invariance when subjected to the same rotations and reflections. 29.2 Choosing an appropriate formalism As mentioned in the introduction to this chapter, the elements of a finite group G can be represented by matrices; this is done in the following way. A suitable column matrix u, known as a basis vector,§ is chosen and is written in terms of its components ui , the basis functions, as u = (u1 u2 · · · un )T . The ui may be of a variety of natures, e.g. numbers, coordinates, functions or even a set of labels, though for any one basis vector they will all be of the same kind. Once chosen, the basis vector can be used to generate an n-dimensional representation of the group as follows. An element X of the group is selected and its effect on each basis function ui is determined. If the action of X on u1 is to produce u1 , etc. then the set of equations ui = Xui § (29.1) This usage of the term basis vector is not exactly the same as that introduced in subsection 8.1.1. 1078 29.2 CHOOSING AN APPROPRIATE FORMALISM generates a new column matrix u = (u1 u2 · · · un )T . Having established u and u we can determine the n × n matrix, M(X) say, that connects them by u = M(X)u. (29.2) It may seem natural to use the matrix M(X) so generated as the representative matrix of the element X; in fact, because we have already chosen the convention whereby Z = XY implies that the effect of applying element Z is the same as that of first applying Y and then applying X to the result, one further step has to be taken. So that the representative matrices D(X) may follow the same convention, i.e. D(Z) = D(X)D(Y ), and at the same time respect the normal rules of matrix multiplication, it is necessary to take the transpose of M(X) as the representative matrix D(X). Explicitly, D(X) = MT (X) (29.3) u = DT (X)u. (29.4) and (29.2) becomes Thus the procedure for determining the matrix D(X) that represents the group element X in a representation based on basis vector u is summarised by equations (29.1)–(29.4).§ This procedure is then repeated for each element X of the group, and the resulting set of n × n matrices D = {D(X)} is said to be the n-dimensional representation of G having u as its basis. The need to take the transpose of each matrix M(X) is not of any fundamental significance, since the only thing that really matters is whether the matrices D(X) have the appropriate multiplication properties – and, as defined, they do. In cases in which the basis functions are labels, the actions of the group elements are such as to cause rearrangements of the labels. Correspondingly the matrices D(X) contain only ‘1’s and ‘0’s as entries; each row and each column contains a single ‘1’. § An alternative procedure in which a row vector is used as the basis vector is possible. Defining equations of the form uT X = uT D(X) are used, and no additional transpositions are needed to define the representative matrices. However, row-matrix equations are cumbersome to write out and in all other parts of this book we have adopted the convention of writing operators (here the group element) to the left of the object on which they operate (here the basis vector). 1079 REPRESENTATION THEORY For the group S3 of permutations on three objects, which has group multiplication table 28.8 on p. 1055, with (in cycle notation) I = (1)(2)(3), C = (1)(2 3), A = (1 2 3), D = (3)(1 2), B = (1 3 2 E = (2)(1 3), use as the components of a basis vector the ordered letter triplets u1 = {P Q R}, u4 = {P R Q}, u2 = {Q R P}, u5 = {Q P R}, u3 = {R P Q}, u6 = {R Q P}. Generate a six-dimensional representation D = {D(X)} of the group and confirm that the representative matrices multiply according to table 28.8, e.g. D(C)D(B) = D(E). It is immediate that the identity permutation I = (1)(2)(3) leaves all ui unchanged, i.e. ui = ui for all i. The representative matrix D(I) is thus I6 , the 6 × 6 unit matrix. We next take X as the permutation A = (1 2 3) and, using (29.1), let it act on each of the components of the basis vector: u1 = Au1 = (1 2 3){P Q R} = {Q R P} = u2 u2 = Au2 = (1 2 3){Q R P} = {R P Q} = u3 .. .. . . u6 = Au6 = (1 2 3){R Q P} = {Q P R} = u5 . The matrix M(A) has to be such that u = M(A)u (here dots replace zeros to aid readability): u1 · 1 · · · · u2 u3 · · 1 · · · u2 u 1 · · · · · u3 u = 1 = ≡ M(A)u. u6 · · · · · 1 u4 u4 · · · 1 · · u5 · · · · 1 · u5 u6 D(A) is then equal to MT (A). The other D(X) are determined in a similar way. In general, if Xui = uj , then [M(X)]ij = 1, leading to [D(X)]ji = 1 and [D(X)]jk = 0 for k = i. For example, Cu3 = (1)(23){R P Q} = {R Q P} = u6 implies that [D(C)]63 · · · D(C) = 1 · · = 1 and [D(C)]6k = 0 for k = 1, 2, 4, 5, 6. When calculated in full · 1 · · · · · · 1 · · · · 1 · · · · · · 1 · · · · · 1 1 · · · · · , D(B) = , · · · · · · · · · · 1 · · · 1 · · 1 · · · · · · · · 1 · · 1 · · · · · · · · 1 · · · 1 · · · · · · 1 · D(E) = , · 1 · · · · · · 1 · · · 1 · · · · · 1080 29.2 CHOOSING AN APPROPRIATE FORMALISM R P P P 1 3 3 2 3 (a) Q R 2 1 Q (b) R 2 Q 1 (c) Figure 29.2 Diagram (a) shows the definition of the basis vector, (b) shows the effect of applying a clockwise rotation of 2π/3 and (c) shows the effect of applying a reflection in the mirror axis through Q. from which it can be verified that D(C)D(B) = D(E). Whilst a representation obtained in this way necessarily has the same dimension as the order of the group it represents, there are, in general, square matrices of both smaller and larger dimensions that can be used to represent the group, though their existence may be less obvious. One possibility that arises when the group elements are symmetry operations on an object whose position and orientation can be referred to a space coordinate system is called the natural representation. In it the representative matrices D(X) describe, in terms of a fixed coordinate system, what happens to a coordinate system that moves with the object when X is applied. There is usually some redundancy of the coordinates used in this type of representation, since interparticle distances are fixed and fewer than 3N coordinates, where N is the number of identical particles, are needed to specify uniquely the object’s position and orientation. Subsection 29.11.1 gives an example that illustrates both the advantages and disadvantages of the natural representation. We continue here with an example of a natural representation that has no such redundancy. Use the fact that the group considered in the previous worked example is isomorphic to the group of two-dimensional symmetry operations on an equilateral triangle to generate a three-dimensional representation of the group. Label the triangle’s corners as 1, 2, 3 and three fixed points in space as P, Q, R, so that initially corner 1 lies at point P, 2 lies at point Q, and 3 at point R. We take P, Q, R as the components of the basis vector. In figure 29.2, (a) shows the initial configuration and also, formally, the result of applying the identity I to the triangle; it is therefore described by the basis vector, (P Q R)T . Diagram (b) shows the the effect of a clockwise rotation by 2π/3, corresponding to element A in the previous example; the new column matrix is (Q R P)T . Diagram (c) shows the effect of a typical mirror reflection – the one that leaves the corner at point Q unchanged (element D in table 28.8 and the previous example); the new column matrix is now (R Q P)T . In similar fashion it can be concluded that the column matrix corresponding to element B, rotation by 4π/3, is (R P Q)T , and that the other two reflections C and E result in 1081 REPRESENTATION THEORY column matrices (P R Q)T and (Q P R)T respectively. The forms of the representative matrices Mnat (X), (29.2), are now determined by equations such as, for element E, P 0 1 0 Q P = 1 0 0 Q 0 0 1 R R implying that 0 D (E) = 1 0 nat 1 0 0 T 0 0 0 = 1 1 0 In this way the complete representation is obtained 1 0 0 0 0 nat nat 0 1 0 D (I) = , D (A) = 1 0 0 0 1 0 1 1 0 0 0 0 Dnat (C) = 0 0 1 , Dnat (D) = 0 1 0 1 0 1 0 1 0 0 0 0 . 1 as 1 0 , 0 1 0 , 0 0 D (B) = 0 1 0 Dnat (E) = 1 0 nat 1 0 0 1 0 0 0 1 , 0 0 0 . 1 It should be emphasised that although the group contains six elements this representation is three-dimensional. We will concentrate on matrix representations of finite groups, particularly rotation and reflection groups (the so-called crystal point groups). The general ideas carry over to infinite groups, such as the continuous rotation groups, but in a book such as this, which aims to cover many areas of applicable mathematics, some topics can only be mentioned and not explored. We now give the formal definition of a representation. Definition. A representation D = {D(X)} of a group G is an assignment of a nonsingular square n × n matrix D(X) to each element X belonging to G, such that (i) D(I) = In , the unit n × n matrix, (ii) D(X)D(Y ) = D(XY ) for any two elements X and Y belonging to G, i.e. the matrices multiply in the same way as the group elements they represent. As mentioned previously, a representation by n × n matrices is said to be an n-dimensional representation of G. The dimension n is not to be confused with g, the order of the group, which gives the number of matrices needed in the representation, though they might not all be different. A consequence of the two defining conditions for a representation is that the matrix associated with the inverse of X is the inverse of the matrix associated with X. This follows immediately from setting Y = X −1 in (ii): D(X)D(X −1 ) = D(XX −1 ) = D(I) = In ; hence D(X −1 ) = [D(X)]−1 . 1082 29.2 CHOOSING AN APPROPRIATE FORMALISM As an example, the four-element Abelian group that consists of the set {1, i, −1, −i} under ordinary multiplication has a two-dimensional representation based on the column matrix (1 i)T : 1 0 0 −1 D(1) = , D(i) = , 0 1 1 0 −1 0 0 1 D(−1) = , D(−i) = . 0 −1 −1 0 The reader should check that D(i)D(−i) = D(1), D(i)D(i) = D(−1) etc., i.e. that the matrices do have exactly the same multiplication properties as the elements of the group. Having done so, the reader may also wonder why anybody would bother with the representative matrices, when the original elements are so much simpler to handle! As we will see later, once some general properties of matrix representations have been established, the analysis of large groups, both Abelian and non-Abelian, can be reduced to routine, almost cookbook, procedures. An n-dimensional representation of G is a homomorphism of G into the set of invertible n × n matrices (i.e. n × n matrices that have inverses or, equivalently, have non-zero determinants); this set is usually known as the general linear group and denoted by GL(n). In general the same matrix may represent more than one element of G; if, however, all the matrices representing the elements of G are different then the representation is said to be faithful, and the homomorphism becomes an isomorphism onto a subgroup of GL(n). A trivial but important representation is D(X) = In for all elements X of G. Clearly both of the defining relationships are satisfied, and there is no restriction on the value of n. However, such a representation is not a faithful one. To sum up, in the context of a rotation–reflection group, the transposes of the set of n × n matrices D(X) that make up a representation D may be thought of as describing what happens to an n-component basis vector of coordinates, (x y · · · )T , or of functions, (Ψ1 Ψ2 · · · )T , the Ψi themselves being functions of coordinates, when the group operation X is carried out on each of the coordinates or functions. For example, to return to the symmetry operations on an equilateral triangle, the clockwise rotation by 2π/3, R, carries the threedimensional basis vector (x y z)T into the column matrix √ − 21 x + 23 y √ − 3x − 1y 2 2 z whilst the two-dimensional basis vector of functions (r 2 3z 2 − r 2 )T is unaltered, as neither r nor z is changed by the rotation. The fact that z is unchanged by any of the operations of the group shows that the components x, y, z actually divide (i.e. are ‘reducible’, to anticipate a more formal description) into two sets: 1083