With Applications to Quantum Mechanics and Solid State Physics
Roland Winkler [email protected]
August 2011
(Lecture notes version: November 3, 2015)
Please, let me know if you find misprints, errors or inaccuracies in these notes. Thank you.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 General Literature
I J. F. Cornwell, Group Theory in Physics (Academic, 1987) general introduction; discrete and continuous groups
I W. Ludwig and C. Falter, Symmetries in Physics (Springer, Berlin, 1988). general introduction; discrete and continuous groups
I W.-K. Tung, Group Theory in Physics (World Scientific, 1985). general introduction; main focus on continuous groups
I L. M. Falicov, Group Theory and Its Physical Applications (University of Chicago Press, Chicago, 1966). small paperback; compact introduction
I E. P. Wigner, Group Theory (Academic, 1959). classical textbook by the master
I Landau and Lifshitz, Quantum Mechanics, Ch. XII (Pergamon, 1977) brief introduction into the main aspects of group theory in physics
I R. McWeeny, Symmetry (Dover, 2002) elementary, self-contained introduction
I and many others Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Specialized Literature
I G. L. Bir und G. E. Pikus, Symmetry and Strain-Induced Effects in Semiconductors (Wiley, New York, 1974) thorough discussion of group theory and its applications in solid state physics by two pioneers
I C. J. Bradley and A. P. Cracknell, The Mathematical Theory of Symmetry in Solids (Clarendon, 1972) comprehensive discussion of group theory in solid state physics
I G. F. Koster et al., Properties of the Thirty-Two Point Groups (MIT Press, 1963) small, but very helpful reference book tabulating the properties of the 32 crystallographic point groups (character tables, Clebsch-Gordan coefficients, compatibility relations, etc.)
I A. R. Edmonds, Angular Momentum in Quantum Mechanics (Princeton University Press, 1960) comprehensive discussion of the (group) theory of angular momentum in quantum mechanics
I and many others
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 These notes are dedicated to
Prof. Dr. h.c. Ulrich R¨ossler from whom I learned group theory
R.W.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Introduction and Overview Definition: Group A set G = {a, b, c,...} is called a group, if there exists a group multiplication connecting the elements in G in the following way (1) a, b ∈ G : c = a b ∈ G (closure) (2) a, b, c ∈ G :(ab)c = a(bc) (associativity) (3) ∃ e ∈ G : a e = a ∀a ∈ G (identity / neutral element) (4) ∀a ∈ G ∃ b ∈ G : a b = e, i.e., b ≡ a−1 (inverse element) Corollaries (a) e−1 = e (b) a−1a = a a−1 = e ∀a ∈ G (left inverse = right inverse) (c) e a = a e = a ∀a ∈ G (left neutral = right neutral) (d) ∀a, b ∈ G : c = a b ⇔ c−1 = b−1a−1 Commutative (Abelian) Group (5) ∀a, b ∈ G : a b = b a (commutatitivity) Order of a Group = number of group elements Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Examples
I integer numbers Z with addition (Abelian group, infinite order)
I rational numbers Q\{0} with multiplication (Abelian group, infinite order)
I complex numbers {exp(2πi m/n): m = 1,..., n} with multiplication (Abelian group, finite order, example of cyclic group)
I invertible (= nonsingular) n × n matrices with matrix multiplication (nonabelian group, infinite order, later important for representation theory!)
I permutations of n objects: Pn (nonabelian group, n! group elements)
I symmetry operations (rotations, reflections, etc.) of equilateral triangle ≡ P3 ≡ permutations of numbered corners of triangle – more later! n I (continuous) translations in R : (continuous) translation group ≡ vector addition in Rn
I symmetry operations of a sphere only rotations: SO(3) = special orthogonal group in R3 = real orthogonal 3 × 3 matrices
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics
Group theory is the natural language to describe symmetries of a physical system
I symmetries correspond to conserved quantities
I symmetries allow us to classify quantum mechanical states • representation theory • degeneracies / level splittings
I evaluation of matrix elements ⇒ Wigner-Eckart theorem e.g., selection rules: dipole matrix elements for optical transitions
I Hamiltonian Hˆ must be invariant under the symmetries of a quantum system ⇒ construct Hˆ via symmetry arguments
I ...
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Classical Mechanics
I Lagrange function L(q, q˙ ), d ∂L ∂L I Lagrange equations = i = 1,..., N dt ∂q˙ i ∂qi ∂L ∂L I If for one j : = 0 ⇒ pj ≡ is a conserved quantity ∂qj ∂q˙ j
Examples
I qj linear coordinate I qj angular coordinate • translational invariance • rotational invariance • linear momentum pj = const. • angular momentum pj = const. • translation group • rotation group
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Quantum Mechanics (1) Evaluation of matrix elements
I Consider particle in potential V (x) = V (−x) even
I two possiblities for eigenfunctions ψ(x)
ψe (x) even: ψe (x) = ψe (−x)
ψo (x) odd: ψo (x) = −ψo (−x) R ∗ I overlapp ψi (x) ψj (x) dx = δij i, j ∈ {e, o} R ∗ I expectation value hi|x|ii = ψi (x) x ψi (x) dx = 0 well-known explanation
I product of two even / two odd functions is even
I product of one even and one odd function is odd
I integral over an odd function vanishes
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Quantum Mechanics (1) Evaluation of matrix elements (cont’d)
Group theory provides systematic generalization of these statements
I representation theory ≡ classification of how functions and operators transform under symmetry operations
I Wigner-Eckart theorem ≡ statements on matrix elements if we know how the functions and operators transform under the symmetries of a system
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Quantum Mechanics (2) Degeneracies of Energy Eigenvalues ˆ ˆ I Schr¨odingerequation Hψ = Eψ or i~∂t ψ = Hψ ˆ ˆ ˆ ˆ ˆ I Let O with i~∂t O = [O, H] = 0 ⇒ O is conserved quantity ˆ ˆ ⇒ eigenvalue equations Hψ = Eψ and Oψ = λOˆ ψ can be solved simultaneously ˆ ⇒ eigenvalue λOˆ of O is good quantum number for ψ Example: H atom 2 ∂2 2 ∂ Lˆ2 e2 Hˆ = ~ + + − ⇒ group SO(3) I 2m ∂r 2 r ∂r 2mr 2 r 2 2 ⇒ [Lˆ , Hˆ ] = [Lˆz , Hˆ ] = [Lˆ , Lˆz ] = 0 2 ⇒ eigenstates ψnlm(r): index l ↔ Lˆ , m ↔ Lˆz
I really another example for representation theory
I degeneracy for 0 ≤ l ≤ n − 1: dynamical symmetry (unique for H atom) Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Quantum Mechanics (3) Solid State Physics in particular: crystalline solids, periodic assembly of atoms ⇒ discrete translation invariance
(i) Electrons in periodic potential V (r)
I V (r + R) = V (r) ∀ R ∈ {lattice vectors}
⇒ translation operator TˆR : TˆR f (r) = f (r + R)
[TˆR, Hˆ ] = 0
ik·r ⇒ Bloch theorem ψk(r) = e uk(r) with uk(r + R) = uk(r) ⇒ wave vector k is quantum number for the discrete translation invariance, k ∈ first Brillouin zone
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Quantum Mechanics (3) Solid State Physics
(ii) Phonons rotation by 90o I Consider square lattice
I frequencies of modes are equal
I degeneracies for particular propagation directions
(iii) Theory of Invariants
I How can we construct models for the dynamics of electrons or phonons that are compatible with given crystal symmetries?
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Physics Quantum Mechanics (4) Nuclear and Particle Physics
Physics at small length scales: strong interaction
Proton mp = 938.28 MeV rest mass of nucleons almost equal
Neutron mn = 939.57 MeV ∼ degeneracy
I Symmetry: isospin ˆI with [ˆI , Hˆstrong] = 0
1 1 1 1 I SU(2): proton | 2 2 i, neutron | 2 − 2 i
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Mathematical Excursion: Groups Basic Concepts
Group Axioms: see above
Definition: Subgroup Let G be a group. A subset U ⊆ G that is itself a group with the same multiplication as G is called a subgroup of G.
Group Multiplication Table: compilation of all products of group elements ⇒ complete information on mathematical structure of a (finite) group
P3 e a b c d f Example: permutation group P3 e e a b c d f 1 2 3 1 2 3 1 2 3 e = a = b = a a b e f c d 1 2 3 2 3 1 3 1 2 b b e a d f c 1 2 3 1 2 3 1 2 3 c c d f e a b c = d = f = 1 3 2 3 2 1 2 1 3 d d f c b e a f f c d a b e
I {e}, {e, a, b}, {e, c}, {e, d}, {e, f }, G are subgroups of G
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 P3 e a b c d f Conclusions from e e a b c d f Group Multiplication Table a a b e f c d b b e a d f c I Symmetry w.r.t. main diagonal ⇒ group is Abelian c c d f e a b d d f c b e a n I order n of g ∈ G: smallest n > 0 with g = e f f c d a b e
2 n I {g, g ,..., g = e} with g ∈ G is Abelian subgroup (a cyclic group)
I in every row / column every element appears exactly once because:
Rearrangement Lemma: for any fixed g 0 ∈ G, we have G = {g 0g : g ∈ G} = {gg 0 : g ∈ G} i.e., the latter sets consist of the elements in G rearranged in order. 0 0 0 proof: g1 6= g2 ⇔ g g1 6= g g2 ∀g1, g2, g ∈ G
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Goal: Classify elements in a group (1) Conjugate Elements and Classes
−1 I Let a ∈ G. Then b ∈ G is called conjugate to a if ∃ x ∈ G with b = x ax . Conjugation b ∼ a is equivalence relation: • a ∼ a reflexive • b ∼ a ⇔ a ∼ b symmetric • a ∼ co a = xcx −1 ⇒ c = x −1ax ⇒ a ∼ b transitive b ∼ c b = ycy −1 = (xy −1)−1a(xy −1)
I For fixed a, the set of all conjugate elements Example: P3 C = {x ax −1 : x ∈ G} is called a class. x e a b c d f e e a b c d f • identity e is its own class x e x −1 = e ∀x ∈ G a e a b d f c • Abelian groups: each element is its own class b e a b f c d x ax −1 = ax x −1 = a ∀a, x ∈ G c e b a c f d d e b a f d c • Each b ∈ G belongs to one and only one class f e b a d c f ⇒ decompose G into classes ⇒ classes {e}, {a, b}, • in broad terms: “similar” elements form a class {c, d, f }
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Goal: Classify elements in a group (2) Subgroups and Cosets I Let U ⊂ G be a subgroup of G and x ∈ G. The set x U ≡ {x u : u ∈ U} (the set Ux) is called the left coset(right coset) of U.
I In general, cosets are not groups. If x ∈/ U, the coset x U lacks the identity element: suppose ∃u ∈ U with xu = e ∈ x U ⇒ x −1 = u ∈ U ⇒ x = u−1 ∈ U 0 0 0 I If x ∈ x U, then x U = x U any x ∈ x U can be used to define coset x U
I If U contains s elements, then each coset also contains s elements (due to rearrangement lemma).
I Two left (right) cosets for a subgroup U are either equal or disjoint (due to rearrangement lemma).
I Thus: decompose G into cosets G = U ∪ x U ∪ y U ∪ ... x, y,.../∈ U Thus Theorem 1: Let h order of G h I ⇒ ∈ N Let s order of U ⊂ G s I Corollary: The order of a finite group is an integer multiple of the orders of its subgroups.
I Corollary: If h prime number ⇒ {e}, G are the only subgroups ⇒ G is isomorphic to cyclic group Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Goal: Classify elements in a group (3) Invariant Subgroups and Factor Groups connection: classes and cosets
I A subgroup U ⊂ G containing only complete classes of G is called invariant subgroup (aka normal subgroup).
I Let U be an invariant subgroup of G and x ∈ G ⇔ x Ux −1 = U ⇔ x U = Ux (left coset = right coset)
I Multiplication of cosets of an invariant subgroup U ⊂ G: x, y ∈ G :(x U)(y U) = xy U = z U where z = xy
well-defined: (x U)(y U) = x (U y) U = xy UU = z UU = z U
I An invariant subgroup U ⊂ G and the distinct cosets x U form a group, called factor group F = G/U • group multiplication: see above • U is identity element of factor group • x −1 U is inverse for x U
I Every factor group F = G/U is homomorphic to G (see below).
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Example: Permutation Group P3
e a b c d f invariant subgroup U = {e, a, b} e e a b c d f a a b e f c d ⇒ one coset cU = dU = f U = {c, d, f } b b e a d f c factor group P /U = {U, cU} c c d f e a b 3 d d f c b e a U cU f f c d a b e U U cU cU cU U
I We can think of factor groups G/U as coarse-grained versions of G.
I Often, factor groups G/U are a helpful intermediate step when working out the structure of more complicated groups G.
I Thus: invariant subgroups are “more useful” subgroups than other subgroups.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Mappings of Groups 0 0 I Let G and G be two groups. A mapping φ : G → G assigns to each g ∈ G an element g 0 = φ(g) ∈ G0, with every g 0 ∈ G0 being the image of at least one g ∈ G.
I If φ(g1) φ(g2) = φ(g1 g2) ∀g1, g2 ∈ G, then φ is a homomorphic mapping of G on G0.
• A homomorphic mapping is consistent with the group structures • A homomorphic mapping G → G0 is always n-to-one (n ≥ 1): The preimage of the unit element of G0 is an invariant subgroup U of G. G0 is isomorphic to the factor group G/U.
I If the mapping φ is one-to-one, then it is an isomomorphic mapping of G on G0. • Short-hand: G isomorphic to G0 ⇒ G ' G0 • Isomorphic groups have the same group structure.
I Examples: 0 • trivial homomorphism G = P3 and G = {e}
• isomorphism between permutation group P3 and symmetry group C3v of
equilateral triangle Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Products of Groups
I Given two groups G1 = {ai } and G2 = {bk }, their outer direct product is the group G1 × G2 with elements (ai , bk ) and multiplication
(ai , bk ) · (aj , bl ) = (ai aj , bk bl ) ∈ G1 × G2
• Check that the group axioms are satisfied for G1 × G2.
• Order of Gn is hn (n = 1, 2) ⇒ order of G1 × G2 is h1h2
• If G = G1 × G2, then both G1 and G2 are invariant subgroups of G. Then we have isomorphisms G2 'G/G1 and G1 'G/G2. • Application: built more complex groups out of simpler groups
I If G1 = G2 = G = {ai }, the elements
(ai , ai ) ∈ G ⊗ G define a group G˜ ≡ G ⊗ G called the inner product of G. • The inner product G ⊗ G is isomorphic to G (⇒ same order as G) • Compare: product representations (discussed below)
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Matrix Representations of a Group
Motivation C e i e = identity i I Consider symmetry group Ci = {e, i} i = inversion e e i i i e I two “types” of basis functions: even and odd
I more abstract: reducible and irreducible representations matrix representation (based on 1 × 1 and 2 × 2 matrices) Γ = {D = 1, D = 1} 1 e i Γ = {D = 1, D = −1} consistent with group 2 e i multiplication table 1 0 1 0 Γ3 = De = 0 1 , Di = 0 −1 where Γ1 : even function fe (x) = fe (−x) irreducible representations Γ2 : odd functions fo (x) = −fo (−x)
Γ3 : reducible representation: decompose any f (x) into even and odd parts 1 fe (x) = f (x) + f (−x) f (x) = f (x)+f (x) with 2 e o 1 fo (x) = 2 f (x) − f (−x) How to generalize these ideas for arbitrary groups? Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Matrix Representations of a Group I Let group G = {gi : i = 1,..., h}
I Associate with each gi ∈ G a nonsingular square matrix D(gi ). If the resulting set {D(gi ): i = 1,..., h} is homomorphic to G it is called a matrix representation of G.
• gi gj = gk ⇒ D(gi ) D(gj ) = D(gk ) • D(e) = 1 (identity matrix) −1 −1 • D(gi ) = D (gi )
I dimension of representation = dimension of representation matrices
Example (1): G = C∞ = rotations around a fixed axis (angle φ) I C∞ is isomorphic to group of orthogonal 2 × 2 matrices SO(2) cos φ − sin φ D (φ) = ⇒ two-dimensional (2D) representation 2 sin φ cos φ
I C∞ is homomorphic to group {D1(φ) = 1} ⇒ trivial 1D representation 1 0 higher-dimensional I C∞ is isomorphic to group ⇒ 0 D2(φ) representation
I Generally: given matrix representations of dimensions n1 and n2, we can construct (n1 + n2) dimensional representations Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Matrix Representations of a Group (cont’d)
Example (2): Symmetry group C3v of equilateral triangle (isomorphic to permutation group P3)
y ◦ ◦ ◦ ◦ y x = 120 = 240 = 240 = 120 ↔ − identity rotation φ rotation φ reflection y roto- reflection φ roto- reflection φ 2 P3 e a b = a c = ec d = ac f = bc 2 Koster E C3 C3 σv σv σv
Γ1 (1) (1) (1) (1) (1) (1)
Γ2 (1) (1) (1) (−1) (−1) (−1) √ √ √ √ 1 3 1 3 1 3 1 3 1 0 − 2 − 2 − 2 2 1 0 − 2 − 2 − 2 2 Γ √ √ √ √ 3 0 1 3 1 3 1 0 −1 3 1 3 1 2 − 2 − 2 − 2 − 2 2 2 2
P3 e a b c d f I mapping G → {D(gi )} homomorphic, e e a b c d f but in general not isomorphic (not faithful) multipli- a a b e f c d I consistent with group multiplication table cation b b e a d f c table c c d f e a b I Goal: characterize matrix representations of G d d f c b e a I Will see: G fully characterized by its “distinct” f f c d a b e matrix representations (only three for G = C3v !) Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Goal: Identify and Classify Representations
I Theorem 2: If U is an invariant subgroup of G, then every representation of the factor group F = G/U is likewise a representation of G. Proof: G is homomorphic to F, which is homomorphic to the representations of F. Thus: To identify the representations of G it helps to identify the representations of F.
I Definition: Equivalent Representations
Let {D(gi )} be a matrix representation for G with dimension n. Let X be a n-dimensional nonsingular matrix. 0 −1 The set {D (gi ) = X D(gi ) X } forms a matrix representation called equivalent to {D(gi )}. 0 Convince yourself: {D (gi )} is, indeed, another matrix representation.
Matrix representations are most convenient if matrices {D} are unitary. Thus
I Theorem 3: Every matrix representation {D(gi )} is equivalent to a unitary 0 0 † 0 −1 representation {D (gi )} where D (gi ) = D (gi )
I In the following, it is always assumed that matrix representations are unitary.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Theorem 3 (cf. Falicov)
Challenge: Matrix X has to be choosen such that it makes all matrices 0 D (gi ) unitary simultaneously.
I Let {D(gi ) ≡ Di : i = 1,..., h} be a matrix representation for G (dimension h). h P † I Define H = Di Di (Hermitean) i=1 I Thus H can be diagonalized by means of a unitary matrix U. −1 P −1 −1 P −1 −1 −1 d ≡ U HU = U Di Di U = U Di U U Di U i i | {z } | {z } † † =D˜i ˜ P ˜ ˜ =Di = Di Di with dµν = dµδµν diagonal i I Diagonal entries dµ are positive: PP ˜ ˜† P ˜ ˜∗ P ˜ 2 dµ = (Di )µλ(Di )λµ = (Di )µλ(Di )µλ = |(Di )µλ| > 0 i λ iλ iλ ±1/2 I Take diagonal matrix d˜± with elements (d˜±)µν ≡ dµ δµν 1 ˜ ˜ ˜ P ˜ ˜† ˜ I Thus = d− d d− = d− Di Di d− (identity matrix) i Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Theorem 3 (cont’d)
0 ˜ ˜ ˜ ˜ −1 ˜ I Assertion: Di = d− Di d+ = d−U Di U d+ are unitary matrices equivalent to Di
−1 • equivalent by construction: X = d˜−U
• unitarity: =1 z }| { 0 0 † ˜ ˜ ˜ ˜ P ˜ ˜† ˜ ˜ ˜† ˜ Di Di = d−Di d+ (d− Dk Dk d−) d+Di d− k ˜ P ˜ ˜ ˜† ˜† ˜ = d− Di Dk Dk Di d− k | {z } | {z } ˜ ˜† = Dj = Dj (rearrangement lemma) | {z } = d = 1 qed
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Reducible and Irreducible Representations (RRs and IRs)
I If for a given representation {D(gi ): i = 1,..., h}, an equivalent 0 representation {D (gi ): i = 1,..., h} can be found that is block diagonal 0 0 D1(gi ) 0 D (gi ) = 0 ∀gi ∈ G 0 D2(gi )
then {D(gi ): i = 1,..., h} is called reducible, otherwise irreducible.
I Crucial: the same block diagonal form is obtained for all representation matrices D(gi ) simultaneously. 0 0 I Block-diagonal matrices do not mix, i.e., if D (g1) and D (g2) are 0 0 0 block diagonal, then D (g3) = D (g1) D (g2) is likewise block diagonal. ⇒ Decomposition of RRs into IRs decomposes the problem into the smallest subproblems possible.
I Goal of Representation Theory Identify and characterize the IRs of a group.
I We will show The number of inequivalent IRs equals the number of classes. Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Schur’s First Lemma
Schur’s First Lemma: Suppose a matrix M commutes with all matrices D(gi ) of an irreducible representation of G
D(gi ) M = M D(gi ) ∀gi ∈ G (♠) then M is a multiple of the identity matrix M = c1, c ∈ C.
Corollaries
I If (♠) holds with M 6= c1, c ∈ C, then {D(gi )} is reducible.
I All IRs of Abelian groups are one-dimensional
Proof: Take gj ∈ G arbitrary, but fixed. G Abelian ⇒ D(gi ) D(gj ) = D(gj ) D(gi ) ∀gi ∈ G Lemma ⇒ D(gj ) = cj 1 with cj ∈ C, i.e., {D(gj ) = cj } is an IR.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Schur’s First Lemma (cf. Bir & Pikus)
† † † † I Take Hermitean conjugate of (♠): M D (gi ) = D (gi ) M † −1 † † Multiply with D (gi ) = D (gi ): D(gi ) M = M D(gi )
† I Thus: (♠) holds for M and M , and also the Hermitean matrices 0 1 † 00 i † M = 2 (M + M ) M = 2 (M − M )
0 00 I It exists a unitary matrix U that diagonalizes M (similar for M ) −1 0 d = U M U with dµν = dµδµν
0 0 0 −1 I Thus (♠) implies D (gi ) d = d D (gi ), where D (gi ) = U D(gi ) U 0 more explicitly: Dµν (gi )(dµ − dν ) = 0 ∀i, µ, ν
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Schur’s First Lemma (cont’d)
Two possibilities:
I All dµ are equal, i.e, d = c1. So M0 = UdU−1 and M00 are likewise proportional to 1, and so is M = M0 − iM00.
I Some dµ are different:
Say {dκ : κ = 1,..., r} are different from {dλ : λ = r + 1,..., h}. ∀κ = 1,..., r; Thus: D0 (g ) = 0 κλ i ∀λ = r + 1,..., h 0 Thus {D (gi ): i = 1,..., h} is block-diagonal, contrary to the assumption that {D(gi )} is irreducible qed
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Schur’s Second Lemma
Schur’s Second Lemma: Suppose we have two IRs {D1(gi ), dimension n1} and {D2(gi ), dimension n2}, as well as a n1 × n2 matrix M such that
D1(gi ) M = M D2(gi ) ∀gi ∈ G (♣)
(1) If {D1(gi )} and {D2(gi )} are inequivalent, then M = 0.
(2) If M 6= 0 then {D1(gi )} and {D2(gi )} are equivalent.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Schur’s Second Lemma (cf. Bir & Pikus)
† −1 −1 I Take Hermitean conjugate of (♣); use D (gi ) = D (gi ) = D(gi ), so † −1 −1 † M D1(gi ) = D2(gi )M −1 −1 I Multiply by M on the left; Eq. (♣) implies M D2(gi ) = D1(gi ) M, so † −1 −1 † −1 MM D1(gi ) = D1(gi )MM ∀gi ∈ G
† I Schur’s first lemma implies that MM is square matrix with MM† = c1 with c ∈ C (*)
I Case a: n1 = n2 • If c 6= 0 then det M 6= 0 because of (*), i.e., M is invertible.
So (♣) implies −1 M D1(gi ) M = D2(gi ) ∀gi ∈ G
thus {D1(gi )} and {D2(gi )} are equivalent.
• If c = 0 then MM† = 0, i.e., P † P ∗ P 2 Mµν Mνµ = Mµν Mµν = |Mµν | = 0 ∀µ ν ν ν so that M = 0.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Schur’s Second Lemma (cont’d)
I Case b: n1 6= n2 (n1 < n2 to be specific)
• Fill up M with n2 − n1 rows to get matrix M˜ with det M˜ = 0.
• However M˜ M˜ † = MM†, so that det(MM†) = det(M˜ M˜ †) = (det M˜ ) (det M˜ †) = 0
• So c = 0, i.e., MM† = 0, and as before M = 0. qed
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Orthogonality Relations for IRs Notation:
I Irreducible Representations (IR): ΓI = {DI (gi ): gi ∈ G} I nI = dimensionality of IR ΓI I h = order of group G
Theorem 4: Orthogonality Relations for Irreducible Representations (1) two inequivalent IRs ΓI 6= ΓJ h 0 0 P ∗ ∀ µ , ν = 1,..., nI DI (gi )µ0ν0 DJ (gi )µν = 0 i=1 ∀ µ, ν = 1,..., nJ
(2) representation matrices of one IR ΓI h nI P ∗ 0 0 DI (gi )µ0ν0 DI (gi )µν = δµ0µ δν0ν ∀ µ , ν , µ, ν = 1,..., nI h i=1 Remarks I [DI (gi )µν : i = 1,..., h] form vectors in a h-dim. vector space p I vectors are normalized to h/nI (because ΓI assumed to be unitary)
I vectors for different I , µν are orthogonal P 2 P 2 I in total, we have I nI such vectors; therefore I nI ≤ h Corollary: For finite groups the number of inequivalent IRs is finite. Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Theorem 4: Orthogonality Relations for IRs
(1) two inequivalent IRs ΓI 6= ΓJ I Take arbitrary nJ × nI matrix X 6= 0 (i.e., at least one Xµν 6= 0) P −1 I Let M ≡ DJ (gi ) X DI (gi ) i =M =1 z }| { z }| { P −1 −1 ⇒D J (gk ) M = DJ (gk ) DJ (gi ) X DI (gi ) DI (gk ) DI (gk ) i | {z } | {z } P −1 = DJ (gk gi ) X DI (gk gi ) DI (gk ) i |{z} |{z} =gj =gj P −1 = DJ (gj ) X DI (gj ) DI (gk ) j | {z } = M DI (gk ) ⇒ (Schur’s Second Lemma) 0 0 = Mµµ0 ∀µ, µ P P −1 in particular correct for = DJ (gi )µκ Xκλ DI (g )λµ0 i X = δ δ 0 i κ,λ κλ νκ λν P −1 = DJ (gi )µν DI (gi )ν0µ0 i P ∗ = DI (gi )µ0ν0 DJ (gi )µν qed i Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Theorem 4: Orthogonality Relations for IRs (cont’d)
(2) representation matrices of one IR ΓI First steps similar to case (1): P −1 I Let M ≡ DI (gi ) X DI (gi ) with nI × nI matrix X 6= 0 i ⇒D I (gk ) M = M DI (gk ) ⇒ (Schur’s First Lemma): M = c 1, c ∈ C
P P −1 I Thus c δµµ0 = DI (gi )µκ Xκλ DI (gi )λµ0 choose Xκλ = δνκ δλν0 i κ,λ P −1 = DI (gi )µν DI (gi )ν0µ0 = Mµµ0 i 1 P 1 P P −1 h I c = Mµµ = DI (gi )µν DI (gi )ν0µ = δνν0 qed nI µ nI i µ nI | {z } −1 DI (gi gi = e)ν0ν = δνν0
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Goal: Characterize different irreducible representations of a group Characters
I The traces of the representation matrices are called characters P χ(gi ) ≡ tr D(gi ) = i D(gi )µµ
I Equivalent IRs are related via a similarity transformation 0 −1 D (gi ) = X D(gi )X with X nonsingular 0 This transformation leaves the trace invariant: tr D (gi ) = tr D(gi ) ⇒ Equivalent representations have the same characters.
I Theorem 5: If gi , gj ∈ G belong to the same class Ck of G, then for every representation ΓI of G we have χI (gi ) = χI (gj )
Proof: −1 • gi , gj ∈ C ⇒ ∃ x ∈ G with gi = x gj x −1 • Thus DI (gi ) = DI (x) DI (gj ) DI (x )
• −1 (trace invariant under χI (gi ) = tr DI (x) DI (gj ) DI (x ) cyclic permutation) −1 = tr DI (x ) DI (x) DI (gj ) = χI (gk ) | {z } =1 Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Characters (cont’d) Notation I χI (Ck ) denotes the character of group elements in class Ck
I The array [χI (Ck )] with I = 1,..., N (N = number of IRs) k = 1,..., N˜ (N˜ = number of classes) is called character table.
Remark: For Abelian groups the character table is the table of the 1 × 1 representation matrices
Theorem 6: Orthogonality relations for characters
Let {DI (gi )} and {DJ (gi )} be two IRs of G. Let hk be the number of elements in class Ck and N˜ the number of classes. Then N˜ P hk ∗ Proof: Use orthogonality χI (Ck ) χJ (Ck ) = δIJ ∀ I , J = 1,..., N relation for IRs k=1 h
I Interpretation: rows [χI (Ck ): k = 1,... N˜ ] of character table are like N orthonormal vectors in a N˜ -dimensional vector space ⇒ N ≤ N˜ .
I If two IRs ΓI and ΓJ have the same characters, this is necessary and sufficient for Γ and Γ to be equivalent. I J Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Example: Symmetry group C3v of equilateral triangle
(isomorphic to permutation group P3)
y ◦ ◦ ◦ ◦ y x = 120 = 240 = 240 = 120 ↔ − identity rotation φ rotation φ reflection y roto- reflection φ roto- reflection φ 2 P3 e a b = a c = ec d = ac f = bc 2 Koster E C3 C3 σv σv σv
Γ1 (1) (1) (1) (1) (1) (1)
Γ2 (1) (1) (1) (−1) (−1) (−1) √ √ √ √ 1 3 1 3 1 3 1 3 1 0 − 2 − 2 − 2 2 1 0 − 2 − 2 − 2 2 Γ √ √ √ √ 3 0 1 3 1 3 1 0 −1 3 1 3 1 2 − 2 − 2 − 2 − 2 2 2 2
P3 e a b c d f Character table e e a b c d f P3 e a, b c, d, f multipli- a a b e f c d cation b b e a d f c C3v E 2C3 3σv table c c d f e a b Γ1 1 1 1 d d f c b e a Γ2 1 1 −1 f f c d a b e Γ3 2 −1 0
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Interpretation: Character Tables
I A character table is the uniquely defined signature of a group and its IRs ΓI [independent of, e.g., phase conventions for representation matrices DI (gi ) that are quite arbitrary].
I Isomorphic groups have the same character tables.
I Yet: the labeling of IRs ΓI is a matter of convention. – Customary: • Γ1 = identity representation: all characters are 1 • IRs are often numbered such that low-dimensional IRs come first; higher-dimensional IRs come later
• If G contains the inversion, a superscript ± is added to ΓI indicating ± the behavior of ΓI under inversion (even or odd) • other labeling schemes are inspired by compatibility relations (more later)
I Different authors use different conventions to label IRs. To compare such notations we need to compare the uniquely defined characters for each class of an IR. (See, e.g., Table 2.7 in Yu and Cardona: Fundamentals of Semiconductors; here we follow Koster et al.)
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Decomposing Reducible Representations (RRs) Into Irreducible Representations (IRs)
Given an arbitrary RR {D(gi )} the representation matrices {D(gi )} can be brought into block-diagonal form by a suitable unitary transformation
D1(gi ) ) . 0 .. a1 times D1(gi ) 0 . . D(gi ) → D (gi ) = .. . . DN (gi ) ) . .. aN times 0 DN (gi )
Theorem 7: Let aI be the multiplicity, with which the IR ΓI ≡ {DI (gi )} is contained in the representation {D(gi )}. Then N P (1) χ(gi ) = aI χI (gi ) I =1 h N˜ 1 P ∗ P hk ∗ (2) aI = χI (gi ) χ(gi ) = χI (Ck ) χ(Ck ) h i=1 k=1 h
We say: {D(gi )} contains the IR ΓI aI times.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof: Theorem 7 (1) due to invariance of trace under similarity transformations N h X 1 X ∗ (2) we have aJ χJ (gi ) = χ(gi ) χ (gi ) × h I J=1 i=1 N h h X 1 X 1 X ⇒ a χ∗(g ) χ (g ) = χ∗(g ) χ(g ) qed J h I i J i h I i i J=1 i=1 i=1 | {z } =δIJ
Applications of Theorem 7:
I Corollary: The representation {D(gi )} is irreducible if and only if h X 2 |χ(gi )| = h i=1 1 for one I Proof: Use Theorem 7 with a = I 0 otherwise
I Decomposition of Product Representations (see later)
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Where Are We?
We have discussed the orthogonality relations for
I irreducible representations
I characters These can be complemented by matching completeness relations. Proving those is a bit more cumbersome. It requires the introduction of the regular representation.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 The Regular Representation
Finding the IRs of a group can be tricky. Yet for finite groups we can derive the regular representation which contains all IRs of the group.
I Interpret group elements gν as basis vectors {|gν i : ν = 1,... h} for a h-dim. representation
⇒ Regular representation: νth column vector of DR (gi ) gives image |gµi = gi |gν i ≡ |gi gν i of basis vector |gν i 1 if g g −1 = g ⇒ D (g ) = µ ν i R i µν 0 otherwise
I Strategy: −1 −1 −1 g1 g2 g3 ... • Re-arrange the group multiplication table g1 e ... as shown on the right g2 e . • For each gi ∈ G we have DR (gi )µν = 1, g3 . e if the entry (µ, ν) in the re-arranged . . e group multiplication table equals gi , otherwise DR (gi )µν = 0. Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Properties of the Regular Representation {DR (gi )}
(1) {DR (gi )} is, indeed, a representation for the group G
(2) It is a faithful representation, i.e., {DR (gi )} is isomorphic to G = {gi }. h if g = e (3) χ (g ) = i R i 0 otherwise Proof:
(1) Matrices {DR (gi )} are nonsingular, as every row / every column contains “1” exactly once.
Show: if gi gj = gk , then DR (gi )DR (gj ) = DR (gk ) Take i, j, µ, ν arbitrary, but fixed −1 −1 DR (gi )µλ = 1 only for gµ gλ = gi ⇔ gλ = gi gµ −1 DR (gj )λν = 1 only for gλ gν = gj ⇔ gλ = gj gν P −1 ⇔ DR (gi )µλ DR (gj )λν = 1 only for gi gµ = gj gν λ −1 ⇔ gµ gν = gi gj = gk [definition of DR (gk )µν ]
(2) immediate consequence of definition of DR (gi ) 1 if g = g g −1 = e (3) D (g ) = i µ µ R i µµ 0 otherwise h if g = e ⇒ χ (g ) = PD (g ) = i R i R i µµ 0 otherwise µ Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Example: Regular Representation for P3
g e a b c d f g −1 e b a c d f e e a b c d f e e b a c d f a a b e f c d a a e b f c d b b e a d f c ⇒ b b a e d f c c c d f e a b c c f d e a b d d f c b e a d d c f b e a f f c d a b e f f d c a b e
Thus 1 0 0 0 0 1 0 1 0 0 1 0 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 e = a = b = 1 0 0 0 1 0 0 0 1 0 1 0 0 0 1 1 0 0 0 0 1 1 0 0 0 1 0
1 0 0 0 1 0 0 0 1 0 1 0 0 0 1 1 0 0 0 0 1 1 0 0 0 1 0 c = d = f = 1 0 0 0 0 1 0 1 0 0 1 0 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Completeness of Irreducible Representations
Lemma: The regular representation contains every IR nI times, where nI = dimensionality of IR ΓI . P Proof: Use Theorem 7: χR (gi ) = aI χI (gi ) where i 1 P ∗ 1 ∗ aI = h χI (gi ) χR (gi ) = h χI (e) χR (e) = nI i | {z } | {z } =nI =h Corollary (Burnside’s Theorem): For a group G of order h, the dimensionalities nI of the IRs ΓI obey P P 2 P 2 Proof: h = χR (e) = I aI χI (e) = I nI nI = h I serious constraint for dimensionalities of IRs Theorem 8: The representation matrices DI (gi ) of a group G of order h obey the completeness relation X X nI D∗(g ) D (g ) = δ ∀ i, j = 1,..., h (*) h I i µν I j µν ij I µ,ν Proof:
I Theorem 4: Interpret [DI (gi )µν : i = 1,..., h] as orthonormal row vectors of a matrix M ⇒ M is square matrix: unitary I Corollary: M has h columns ⇒ column vectors also orthonormal = completeness (*) Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Completeness Relation for Characters
Theorem 9: Completeness Relation for Characters If χI (Ck ) is the character for class Ck and irreducible representation I , then h k X ∗ 0 χ (C )χ (C 0 ) = δ 0 ∀ k, k = 1,..., N˜ h I k I k kk I
χ1(Ck ) . ˜ I Interpretation: columns . of character table [k = 1,..., N] χN (Ck ) are like N˜ orthonormal vectors in a N-dimensional vector space
˜ (from com- ) I Thus N ≤ N. pleteness) Number N of irreducible representations ˜ (from ortho- = Number N˜ of classes I Also N ≤ N gonality)
I Character table • square table • rows and column form orthogonal vectors
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Theorem 9: Completeness Relation for Characters
Lemma: Let {DI (gi )} be an nI -dimensional IR of G. Let Ck be a class of G with hk elements. Then X hk DI (gi ) = χI (Ck ) 1 nI i∈Ck The sum over all representation matrices in a class of an IR is proportional to the identity matrix. Proof of Lemma:
I For arbitrary gj ∈ G P −1 P −1 P DI (gj ) DI (gi ) DI (gj ) = DI (gj ) DI (gi ) DI (gj ) = DI (gi 0 ) x 0 i∈Ck i∈Ck | {z } i ∈Ck 0 =DI (g 0 ) with i ∈Ck i because gj maps gi 6= gi onto g 0 6= g 0 1 2 i1 i2 P ⇒ (Schur’s First Lemma): DI (gi ) = ck 1 i∈Ck
1 P hk I ck = tr DI (gi ) = χI (Ck ) qed nI i∈Ck nI
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Proof of Theorem 9: Completeness Relation for Characters
I Use Theorem 8 (Completeness Relations for Irreducible Representations) N P P nI ∗ P P D (gi )µν DI (gj )µν = δij h I I =1 µ,ν i∈Ck j∈Ck0
P nI P P ∗ P ⇒ DI (gi ) DI (gj ) = hk δkk0 I h µ,ν i∈C j∈C 0 k µν k µν | {z } | {z } hk ∗ hk0 χI (Ck ) δµν χI (Ck0 ) δµν (Lemma) nI nI | {z } hk hk0 ∗ P 2 χI (Ck ) χI (Ck0 ) δµν qed nI µ,ν | {z } =nI
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Summary: Orthogonality and Completeness Relations
Theorem 4: Orthogonality Relations for Irreducible Representations n h I , J = 1,..., N I X ∗ 0 0 D (g ) 0 0 D (g ) = δ δ 0 δ 0 µ , ν = 1,..., n h I i µ ν J i µν IJ µµ νν I i=1 µ, ν = 1,..., nJ
Theorem 8: Completeness Relations for Irreducible Representations N X X nI D∗(g ) D (g ) = δ ∀ i, j = 1,..., h h I i µν I j µν ij I =1 µ,ν
Theorem 6: Orthogonality Relations for Characters
N˜ X hk χ∗(C ) χ (C ) = δ ∀ I , J = 1,..., N h I k J k IJ k=1
Theorem 9: Completeness Relation for Characters h N k X ∗ 0 χ (C )χ (C 0 ) = δ 0 ∀ k, k = 1,..., N˜ h I k I k kk I =1 Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Unreducible h hhhh More on Irreduciblehhhh Problems
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Group Theory in Quantum Mechanics
Topics:
I Behavior of quantum mechanical states and operators under symmetry operations
I Relation between irreducible representations and invariant subspaces of the Hilbert space
I Connection between eigenvalue spectrum of quantum mechanical operators and irreducible representations
I Selection rules: symmetry-induced vanishing of matrix elements and Wigner-Eckart theorem
Note: Operator formalism of QM convenient to discuss group theory. Yet: many results also applicable in other areas of physics.
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Symmetry Operations in Quantum Mechanics (QM)
I Let G = {gi } be a group of symmetry operations of a qm system e.g., translations, rotations, permutation of particles
I Translated into the language of group theory: In the Hilbert space of the qm system we have a group of unitary 0 0 operators G = {Pˆ(gi )} such that G is isomorphic to G.
Examples
I translations Ta ˆ → unitary operator P(Ta) = exp (ipˆ · a/~) (pˆ = momentum) ˆ 1 2 P(Ta) ψ(r) = 1 + ∇ · a + 2 (∇ · a) + ... ψ(r) = ψ(r + a)
I rotations Rφ ˆ (L = angular momentum → unitary operator Pˆ(n, φ) = exp iLˆ · nφ/~ φ = angle of rotation n = axis of rotation)
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Transformation of QM States
I Let {|νi} be an orthonormal basis
I Let Pˆ(gi ) be the symmetry operator for the symmetry transformation gi with symmetry group G = {gi }.
P I Then Pˆ(gi ) |νi = |µi hµ| Pˆ(gi ) |νi ↑ µ | {z } 1 P D(gi )µν = µ |µihµ| matrix of a unitary P I So Pˆ(gi ) |νi = D(gi )µν |µi where D(gi )µν = representation of G µ because Pˆ(gi ) unitary
I Note: bras and kets transform according to complex conjugate representations ˆ † P ∗ hν|P(gi ) = hµ| D(gi )µν µ (consistent with I Let gi , gj ∈ G with gi gj = gk ∈ G. Then matrix multiplication) D(gi )κµ D(gi )µν z }| { z }| { P ˆ ˆ P |κihκ| P(gi ) |µi hµ| P(gi ) |νi = D(gi )κµD(gi )µν |κi κµ κµ Pˆ(gi ) Pˆ(gj ) |νi = P Pˆ(gk ) |νi = D(gk )κν |κi κ Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Transformation of (Wave) Functions ψ(r) ˆ 0 ˆ 0 −1 ˆ † ˆ −1 I P(gi )|ri = |r =gi ri ⇔ hr|P(gi ) = hr =gi r| b/c P(g) = P(g )
I Let ψ(r) ≡ hr|ψi ˆ ˆ −1 ⇒ P(gi ) ψ(r) ≡ hr|P(gi )|ψi = ψ(gi r) ≡ ψi (r)
I In general, the functions V = {ψi (r): i = 1,..., h} are linear dependent
⇒ Choose instead linear independent functions ψν (r) ≡ hr|νi spanning V
⇒ Expand images Pˆ(gi ) ψν (r) in terms of {ψν (r)}: P P Pˆ(gi ) ψ(r) = hr|Pˆ(gi )|ψi = hr|µihµ|Pˆ(gi )|νi = D(gi )µν ψµ(r) µ µ
ˆ −1 P ⇒ P(gi ) ψν (r) = ψν (gi r) = D(gi )µν ψµ(r) µ
I Thus: every function ψ(r) induces a matrix representation Γ = {D(gi )}
I Also: every representation Γ = {D(gi )} is completely characterized by a (nonunique) set of basis functions {ψν (r)} transforming according to Γ.
I Dirac bra-ket notation convenient for formulating group theory of functions. Yet: results applicable in many areas of physics beyond QM. Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Important Representations in Physics (usually reducible)
(1) Representations for polar and axial (cartesian) vectors
I generally: two types of point group symmetry operations
• proper rotations gpr = (n, θ) about axis n, angle θ
D[gpr = (n, θ)] = Rodrigues’ rotation formula 2 nx (1 − cos θ) + cos θ nx ny (1 − cos θ) − nz sin θ nx nz (1 − cos θ) + ny sin θ 2 ny nx (1 − cos θ) + nz sin θ ny (1 − cos θ) + cos θ ny nz (1 − cos θ) − nx sin θ 2 nz nx (1 − cos θ) − ny sin θ nz ny (1 − cos θ) + nx sin θ nz (1 − cos θ) + cos θ
I det D(gpr) = +1
I χ(gpr) = tr D(gpr) = 1 + 2 cos θ independent of n
• improper rotations gim ≡ i gpr = gpri where i = inversion
• inversion i: Dpol(i) = −13×3 I polar vectors • improper rotations gim = i gpr: • proper rotations gpr: I Dpol(gim) = −Dpol(gpr) I det Dpol(gpr) = +1 I det Dpol(gim) = −1 I tr Dpol(gpr) = 1 + 2 cos θ I tr Dpol(gim) = −(1 + 2 cos θ)
• Γpol = {Dpol(g)} ⊆ O(3) always a faithful representation (i.e., isomorphic to G) • examples: position r, linear momentum p, electric field E Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Important Representations in Physics (1) Representations for polar and axial (cartesian) vectors (cont’d)
I axial vectors • proper rotations gpr:
I Dax(gpr) = Dpol(gpr)
I det Dax(gpr) = +1
I tr Dax(gpr) = 1 + 2 cos θ
• inversion i: Dax(i) = +13×3
• improper rotations gim = i gpr:
I Dax(gim) = Dax(gpr) = −Dpol(gpr)
I det Dax(gim) = +1
I tr Dax(gim) = 1 + 2 cos θ
• Γax = {Dax(g)} ⊆ SO(3) • examples: angular momentum L, magnetic field B
I systems with discrete symmetry group G = {gi : i = 1,..., h}: We have a “universal recipe” to Γpol = {Dpol(gi ): i = 1,..., h} construct the 3 × 3 matrices Dpol(g) and Dax(g) for each group element Γax = {Dax(gi ): i = 1,..., h} gpr = (n, θ) and gim = i (n, θ) Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Important Representations in Physics (cont’d)
(2) Equivalence Representations Γeq I Consider symmetric object (symmetry group G) • vertices, edges, and faces of platonic solids are equivalent by symmetry • atoms / atomic orbitals |µi in a molecule may be equivalent by symmetry
I Equivalence representation Γeq describes mapping of equivalent objects P I Generally: Pˆ(g) |µi = Deq(g)νµ |νi ν
I Example: orbitals of equivalent H atoms in NH3 molecule (group C3v )
Equivalent to: permutations of corners of triangle (group P3)
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Example: Symmetry group C3v of equilateral triangle
(isomorphic to permutation group P3)
z ◦ ◦ ◦ ◦
y y 2 3 = 120 = 120 = 240 = 240 ↔ − 1 x identity rotation φ rotation φ reflection y roto- reflection φ roto- reflection φ 2 P3 e a b = a c = ec d = ac f = bc 2 Koster E C3 C3 σv σv σv
Γ1 (1) (1) (1) (1) (1) (1)
Γ2 (1) (1) (1) (−1) (−1) (−1) √ √ √ √ 1 3 1 3 1 3 1 3 1 0 − 2 − 2 − 2 2 1 0 − 2 − 2 − 2 2 Γ √ √ √ √ 3 0 1 3 1 3 1 0 −1 3 1 3 1 2 − 2 − 2 − 2 − 2 2 2 2 √ √ 3 1 3 1 n, θ (0, 0, 1), 0 (0, 0, 1), 2π/3 (0, 0, 1), 4π/3 (0, 1, 0), π (− 2 , − 2 , 0), π (− 2 , 2 , 0), π √ √ √ √ 1 3 ! 1 3 ! 1 3 ! 1 3 ! 1 − − − 1 − − − Γpol = √2 2 √2 2 √2 2 √2 2 −1 1 3 − 1 − 3 − 1 − 3 1 3 1 Γ1 + Γ3 1 2 2 2 2 1 2 2 2 2 1 1 1 1 √ √ √ √ 1 3 ! 1 3 ! 1 3 ! 1 3 ! 1 − − − −1 − Γax = √2 2 √2 2 √2 2 2√ 2 1 3 3 1 3 − 1 − 3 − 1 − 1 − − 1 Γ2 + Γ3 1 2 2 2 2 −1 2 2 2 2 1 1 −1 −1
Γeq = 1 1 1 1 1 1 1 1 1 1 1 1 Γ1 + Γ3 1 1 1 1 1 1 Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Transformation of QM States (cont’d)
I in general: representation {D(gi )} of states {|νi} is reducible
0 −1 I We have D (gi ) = U D(gi ) U
0 P −1 I More explicitly: D (gi )µ0ν0 = Uµ0µ D(gi )µν Uνν0 µν | {z } hµ| Pˆ(gi ) |νi P −1 ˆ = hµ|Uµ0µ P(gi ) Uνν0 |νi µν 0 0 = hµ | Pˆ(gi ) |ν i 0 P with |ν i = Uνν0 |νi ν
I Thus: block diagonalization 0 −1 {D(gi )} → {D (gi ) = U D(gi ) U} corresponds to change of basis 0 P {|νi} → {|ν i = Uνν0 |νi} ν
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Basis Functions for Irreducible Representations
I matrices {DI (gi )} are fully characterized by basis functions I {ψν (r): ν = 1,... nI } transforming according to IR ΓI ˆ I −1 P I P(gi ) ψν (r) = ψν (gi r) = DI (gi )µν ψµ(r) µ
I convenient if we need to spell out phase conventions for {DI (gi )} (→ Koster) I identify IRs for (components of) polar and axial vectors
Example: Symmetry group C3v z ◦ ◦ ◦ ◦
y y = 240 = 120 = 240 = 120 ↔ − x identity rotation φ rotation φ reflection y roto- reflection φ roto- reflection φ basis functions x 2 + y 2; z; Γ1 (1) (1) (1) (1) (1) (1) 2 2 Lx + Ly Γ2 (1) (1) (1) (−1) (−1) (−1) Lz √ √ √ √ 1 3 1 3 1 3 1 3 1 0 − 2 − 2 − 2 2 1 0 2 2 2 − 2 Γ √ √ √ √ x, y 3 0 1 3 1 3 1 0 −1 3 1 3 1 2 − 2 − 2 − 2 2 − 2 − 2 − 2 √ √ √ √ 1 3 1 3 1 3 1 3 1 0 − 2 − 2 − 2 2 −1 0 − 2 − 2 − 2 2 √ √ √ √ Lx , Ly 0 1 3 1 3 1 0 1 3 1 3 1 2 − 2 − 2 − 2 − 2 2 2 2 r = (x, y, z) = polar vector, L = (Lx , Ly , Lz ) = axial vector Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Relevance of Irreducible Representations Invariant Subspaces Definition:
I Let G = {gi } be a group of symmetry transformations. Let H = {|µi} be a Hilbert space with states |µi. A subspace S ⊂ H is called invariant subspace (with respect to G) if
Pˆ(gi ) |µi ∈ S ∀ gi ∈ G, ∀ |µi ∈ S I If an invariant subspace can be decomposed into smaller invariant subspaces, it is called reducible, otherwise it is called irreducible.
Theorem 10: An invariant subspace S is irreducible if and only if the states in S transform according to an irreducible representation.
Proof:
I Suppose {D(gi )} is reducible. 0 −1 I ∃ unitary transformation U with {D (gi ) = U D(gi ) U} block diagonal 0 0 P I For {D (gi )} we have the basis {|µ i = µ Uµµ0 |µi} 0 0 I The block diagonal form of {D (gi )} implies that {|µ i is reducible
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Invariant Subspaces (cont’d)
Corollary: Every Hilbert space H can be decomposed into irreducible invariant subspaces SI transforming according to the IR ΓI Remark: Given a Hilbert space H we can generally have multiple α (possibly orthogonal) irreducible invariant subspaces SI α SI = |I ν αi : ν = 1,..., nI transforming according to the same IR ΓI P Pˆ(gi ) |I ν αi = DI (gi )µν |I µαi µ Theorem 11: (1) States transforming according to different IRs are orthogonal (2) For states |I µαi and |I ν βi transforming according to the same IR ΓI we have hI µα | I ν βi = δµν hI α||I βi where the reduced matrix element hI α||I βi is independent of µ, ν.
Remark: This theorem lets us anticipate the Wigner-Eckart theorem Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Invariant Subspaces (cont’d)
Proof of Theorem 11 ˆ 1 ˆ † ˆ 1 Pˆ † ˆ I Use unitarity of P(gj ): = P(gj ) P(gj ) = h P(gi ) P(gi ) i I Then 1 P ˆ † ˆ hI µα |J ν βi = h hI µα| P(gi ) P(gi ) |J ν βi i | {z ∗ } | {z } P hI µ0 α|D (g ) P D (g ) 0 |J ν0 βi µ0 I i µ0µ ν0 J i ν ν P 0 0 1 P ∗ = hI µ α |J ν βi h i DI (gi )µ0µ DJ (gi )ν0ν µ0ν0 | {z } (1/nI ) δIJ δµν δµ0ν0 1 P 0 0 = δIJ δµν hI µ α |I µ βi nI µ0 | {z } ≡hI α||I βi qed
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Discussion Theorem 11
I ΓJ × ΓI contains the identity representation Γ1 if and only if the IR ΓJ is the ∗ ∗ complex conjugate of ΓI , i.e., ΓJ = ΓI ⇔ DJ (g) = DI (g) ∀g.
I If the ket |J µαi transforms according to the IR ΓJ , the bra hJ µα| ∗ transforms according to the complex conjugate representation ΓJ . I Thus: hJ µα|I ν βi= 6 0 equivalent to • bra and ket transform according to complex conjugate representations • hJ µα|I ν βi contains the identity representation
I Indeed, common theme of representation theory applied to physics: Terms are only nonzero if they transform according to a representation that contains the identity representation.
I Variant of Theorem 11 (Bir & Pikus): Z If fI (x) transforms according to some IR ΓI , then fI (x) dx 6= 0
only if ΓI is the identity representation.
I Applications • Wigner-Eckart Theorem • Nonzero elements of material tensors • Our universe would be zero “by symmetry” if the apparently trivial identity
representation did not exist. Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Decomposition into Irreducible Invariant Subspaces
I Goal: Decompose general state |ψi ∈ H into components from irreducible invariant subspaces SI
ˆ I nI P ∗ ˆ I Generalized projection operator Πµµ0 := h DI (gi )µµ0 P(gi ) i ˆ I I Theorem 12: (i) Πµµ0 |Jναi = δIJ δµ0ν |I µαi ˆ I ˆ J ˆ J (ii) Πµµ0 Πνν0 = δIJ δµ0ν Πµν0 P ˆ I 1 (iii) Πµµ = Proof: I µ I n ∗ n ∗ 0 ˆ I P ˆ P I P 0 (i) Πµµ0 |Jναi = h DI (gi )µµ0 P(gi ) |Jναi = h i DI (gi )µµ0 DJ (gi )ν ν |Jν αi i | {z } ν0 | {z } P 0 δ δ δ ν0 DJ (gi )ν0ν |Jν αi IJ µν0 µ0ν ˆ I ˆ J nI P nJ P ∗ ∗ ˆ ˆ (ii) Πµµ0 Πνν0 = h h DI (gi )µµ0 DJ (gj )νν0 P(gi ) P(gj ) subst. gi gj = gk i j nI P nJ P ∗ −1 ∗ ˆ = h h DI (gi )µµ0 DJ (gi gk )νν0 P(gk ) i k nJ P nI P ∗ −1 ∗ ∗ ˆ = h h DI (gi )µµ0 DJ (gi )νλ DJ (gk )λν0 P(gk ) kλ i | {z } DJ (gi )λν | {z } [or use (i)] δIJ δµλδµ0ν P ˆ I P 1 P P ∗ ˆ P ˆ ˆ 1 (iii) Πµµ = h µDI (gi )µµ0 nI P(gi ) = δgi e P(gi ) = P(e) ≡ I µ i I | {z } |{z} i χ∗(g ) χ (e) I i I Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Decomposition into Invariant Subspaces (cont’d)
Discussion P I Let |ψi = cJνα |Jναi general state with coefficients cJµα (*) Jνα ˆ I I Diagonal operator Πµµ projects |ψi on components |I µαi: ˆ I 2 ˆ I P • (Πµµ) |ψi = Πµµ|ψi = cI µα |I µαi α P ˆ I • Πµµ = 1 I µ
ˆ I Pˆ I nI P ∗ ˆ I Let Π ≡ Πµµ = h χI (gi ) P(gi ): µ i I P • Πˆ |ψi = cI να |I ναi να I • Πˆ projects |ψi on the invariant subspace SI (IR ΓI )
I For functions ψ(r) ≡ hr|ψi:
ˆ I nI P ∗ −1 we need not know Πµµ0 ψ(r) = h DI (gi )µµ0 ψ(gi r) the expansion (*) i
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Irreducible Invariant Subspaces (cont’d) Example: C e i e = identity i I Group Ci = {e, i} i = inversion e e i i i e
Ci e i
I character table Γ1 1 1 Γ2 1 −1
I Pˆ(e) ψ(x) = ψ(x), Pˆ(i) ψ(x) = ψ(−x)
ˆ I nI P ∗ ˆ I Projection operator Π = h χI (gi ) P(gi ) with nI = 1, h = 2 i
ˆ 1 1 ˆ ˆ ˆ 1 1 I Π = 2 [P(e) + P(i)] ⇒ Π ψ(x) = 2 [ψ(x) + ψ(−x)] even part ˆ 2 1 ˆ ˆ ˆ 2 1 Π = 2 [P(e) − P(i)] ⇒ Π ψ(x) = 2 [ψ(x) − ψ(−x)] odd part
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Product Representations
I Let {|I µi : µ = 1,... nI } and {|Jνi : ν = 1,... nJ } denote basis functions for invariant subspaces SI and SJ (need not be irreducible) Consider the product functions
{|I µi |Jνi : µ = 1,..., nI ; ν = 1,..., nJ }. How do these functions transform under G?
I Definition: Let DI (g) and DJ (g) be representation matrices for g ∈ G.
The direct product (Kronecker product) DI (g) ⊗ DJ (g) denotes the matrix whose elements in row (µν) and column (µ0ν0) are given by 0 µ, µ = 1,..., nI [DI (g) ⊗ DJ (g)]µν,µ0ν0 = DI (g)µµ0 DJ (g)νν0 0 ν, ν = 1,..., nJ x11 x12 y11 y12 I Example: Let DI (g) = and DJ (g) = x21 x22 y21 y22 x11y11 x11y12 x12y11 x12y12 x11 DJ (g) x12 DJ (g) x11y21 x11y22 x12y21 x12y22 DI (g) ⊗ DJ (g) = = x21 DJ (g) x22 DJ (g) x21y11 x21y12 x22y11 x22y12 x21y21 x21y22 x22y21 x22y22
I Details of the arrangement in the following not relevant Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Product Representations (cont’d)
I Dimension of product matrix
dim [DI (g) ⊗ DJ (g)] = dim DI (g) dim DJ (g)
I Let ΓI = {DI (gi )} and ΓJ = {DJ (gi )} be representations of G. Then
ΓI × ΓJ ≡ {DI (g) ⊗ DJ (g)} is a representation of G called product representation.
I ΓI × ΓJ is, indeed, a representation:
Let DI (gi ) DI (gj ) = DI (gk ) and DJ (gi ) DJ (gj ) = DJ (gk ) ⇒ [D (g ) ⊗ D (g )] [D (g ) ⊗ D (g )] I i J i I j J j µν,µ0ν0 P = DI (gi )µκ DJ (gi )νλ DI (gj )κµ0 DJ (gj )λν0 κλ → DJ (gk ) 0 → DI (gk )µµ0 νν
= [DI (gk ) ⊗ DJ (gk )]µν,µ0ν0
ˆ P 0 I Let P(g) |I µi = µ0 DI (g)µ0µ |I µ i ˆ P 0 P(g) |Jνi = ν0 DJ (g)ν0ν |Jν i P 0 0 Then Pˆ(g) |I µi|Jνi = [DI (g) ⊗ DJ (g)]µ0ν0,µν |I µ i|Jν i µ0ν0 Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Product Representations (cont’d)
I The characters of the product representation are χI ×J (gi ) = χI (gi ) χJ (gi )
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Decomposing Product Representations
I Let ΓI = {DI (gi )} and ΓJ = {DJ (gi )} be irreducible representations of G
The product representation ΓI × ΓJ = {DI ×J (gi )} is generally reducible
I According to Theorem 7, we have N˜ h Γ × Γ = P aIJ Γ where aIJ = P k χ∗ (C ) χ (C ) I J K K K h K k I ×J k K k=1 | {z } =χI (Ck ) χJ (Ck )
I The multiplication table for the ΓI × ΓJ Γ1 Γ2 ... irreducible representations ΓI of G Γ1 lists P aIJ Γ Γ k K K .2 .
I Example: Permutation group P3
χ(C) e a, b c, d, f ΓI × ΓJ Γ1 Γ2 Γ3 Γ1 1 1 1 Γ1 Γ1 Γ2 Γ3 Γ2 1 1 −1 Γ2 Γ1 Γ3 Γ3 2 −1 0 Γ3 Γ1 + Γ2 + Γ3
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 (Anti-) Symmetrized Product Representations
Let {|σµi} and {|τν i} be two sets of basis functions for the same n-dim. (again: need not be representation Γ = {D(g)} with characters {χ(g)}. irreducible)
(1) “Simple” Product: (discussed previously)
µ = 1,..., n o 2 I |ψµν i = |σµi|τν i, ν = 1,... n total: n n n P P I Pˆ(g)|ψµν i = D(g)µ0µD(g)ν0ν |σµ0 i|τν0 i µ0=1 ν0=1 n n P P ≡ [D(g) ⊗ D(g)]µ0ν0,µν |ψµ0ν0 i µ0=1 ν0=1
2 I Character tr[D(g) ⊗ D(g)] = χ (g)
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 (Anti-) Symmetrized Product Representations (cont’d)
(2) Symmetrized Product: µ = 1,..., n o s 1 total: 1 n(n + 1) I |ψµν i = 2 (|σµi|τν i + |σν i|τµi), ν = 1, . . . µ 2 n n ˆ s 1 P P I P(g)|ψµν i = 2 Dµ0µDν0ν (|σµi|τν i + |σν i|τµi) µ0=1 ν0=1 n µ0−1 P P s s = (Dµ0µDν0ν + Dµ0ν Dν0µ)|ψµ0ν0 i + Dµ0µDµ0ν |ψµ0µ0 i µ0=1 ν0=1 n µ0 P P (s) s ≡ [D(g) ⊗ D(g)]µ0ν0,µν |ψµ0ν0 i µ0=1 ν0=1 n µ−1 (s) P P I tr[D(g) ⊗ D(g)] = (DµµDνν + Dµν Dνµ) + DµµDµµ µ=1 ν=1 n n 1 P P = 2 [Dµµ(g)Dνν (g) + Dµν (g)Dνµ(g)] µ=1 ν=1 n n 1 P P 2 = 2 Dµµ(g) Dνν (g) + Dµµ(g ) µ=1 ν=1 1 2 2 = 2 [χ(g) + χ(g )]
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 (Anti-) Symmetrized Product Representations (cont’d)
(3) Antisymmetrized Product: µ = 1,..., n o a 1 total: 1 n(n − 1) I |ψµν i = 2 (|σµi|τν i − |σν i|τµi), ν = 1, . . . µ − 1 2 n n ˆ a 1 P P I P(g)|ψµν i = 2 Dµ0µDν0ν (|σµi|τν i − |σν i|τµi) µ0=1 ν0=1 n µ0−1 P P a = (Dµ0µDν0ν − Dµ0ν Dν0µ)|ψµ0ν0 i µ0=1 ν0=1 n µ0−1 P P (a) a ≡ [D(g) ⊗ D(g)]µ0ν0,µν |ψµ0ν0 i µ0=1 ν0=1 n µ−1 (a) P P I tr[D(g) ⊗ D(g)] = (DµµDνν − Dµν Dνµ) µ=1 ν=1 n n 1 P P = 2 [Dµµ(g)Dνν (g) − Dµν (g)Dνµ(g)] µ=1 ν=1 n n 1 P P 2 = 2 Dµµ(g) Dνν (g) − Dµµ(g ) µ=1 ν=1 1 2 2 = 2 [χ(g) − χ(g )]
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Intermezzo: Material Tensors to be added ...
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Discussion
I Representation – Vector Space The matrices {D(gi )} of an n-dimensional (reducible or irreducible) representation describe a linear mapping of a vector space V onto itself. D(gi ) 0 0 P u = (u1,..., un) ∈ V : u −−−→ u ∈ V with uµ = D(gi )µν uν ν I Irreducible Representation (IR) – Invariant Subspace The decomposition of a reducible representation into IRs ΓI corresponds to a decomposition of the vector space V into invariant subspaces SI such that DI (gi ) SI −−−−→ SI ∀gi ∈ G (i.e., no mixing)
This decomposition of V lets us break down a big physical problem into smaller, more tractable problems
I Product Representation – Product Space A product representation ΓI × ΓJ describes a linear mapping of the product space SI × SJ onto itself DI ×J (gi ) SI × SJ −−−−−→ SI × SJ ∀gi ∈ G
P IJ I The block diagonalization ΓI × ΓJ = K aK ΓK corresponds to a decomposition of SI × SJ into invariant subspaces SK Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Discussion (cont’d) Clebsch-Gordan Coefficients (CGC)
P IJ I The block diagonalization ΓI × ΓJ = K aK ΓK corresponds to a decomposition of SI × SJ into invariant subspaces SK ⇒ Change of Basis: unitary transformation IJ aK ( PP ` SI × SJ −→ SK K `=1 I J K` old basis {eµeν } −→ new basis {eκ }
K` P IJ K ` I J index ` not needed Thus eκ = eµ eν if often aIJ ≤ 1 µν µ ν κ K IJ K ` where µ ν κ = Clebsch-Gordan coefficients (CGC)
Clebsch-Gordan coefficients describe the unitary transformation for the ` decomposition of the product space SI × SJ into invariant subspaces SK
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Clebsch-Gordan Coefficients (cont’d) Remarks
I CGC are independent of the group elements gi
I CGC are tabulated for all important groups (e.g., Koster, Edmonds)
I Note: Tabulated CGC refer to a particular definition (phase convention) I for the basis vectors {eµ} and representation matrices {DI (gi )}
I Clebsch-Gordan coefficients C describe a unitary basis transformation C† C = C C† = 1
I Thus Theorem 13: Orthogonality and completeness of CGC ∗ X IJ K ` IJ K 0 `0 µ ν κ µ ν κ0 = δKK 0 δκκ0 δ``0 µν ∗ X IJ K ` IJ K ` µ ν κ µ0 ν0 κ = δµµ0 δνν0 K`κ
Roland Winkler, NIU, Argonne, and NCTU 2011−2015 Clebsch-Gordan Coefficients (cont’d)