Introduction to Group Theory

With Applications to Quantum Mechanics and Solid State Physics

Roland Winkler [email protected] August 2011
(Lecture notes version: January 27, 2012)

Please, let me know if you nd misprints or mistakes in these notes. Thank you.

Roland Winkler, NIU, Argonne, and NCTU 2011

General Literature
J. F. Cornwell, Group Theory in Physics, (Academic, 1987)
general introduction; discrete and continuous groups

W. Ludwig and C. Falter, Symmetries in Physics, (Springer, Berlin, 1988).

general introduction; discrete and continuous groups

W.-K. Tung, Group Theory in Physics, (World Scientic, 1985).

general introduction; main focus on continuous groups

L. M. Falicov, Group Theory and Its Physical Applications, (University of Chicago Press, Chicago, 1966).
small paperback; compact introduction

E. P. Wigner, Group Theory, (Academic, 1959).

classical textbook by the master

Landau and Lifshitz, Quantum Mechanics, Chap. XII (Pergamon, 1977)

brief introduction into the main aspects of group theory in physics

and many others

Roland Winkler, NIU, Argonne, and NCTU 2011

Specialized Literature
G. L. Bir und G. E. Pikus, Symmetry and Strain-Induced Eects in Semiconductors, (Wiley, New York, 1974)
thorough discussion of group theory and its applications in solid state physics by two pioneers

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

G. F. Koster, J. O. Dimmock, R. G. Wheeler und H. Statz, Properties of the Thirty-Two Point Groups, (MIT Press, 1963)
small, but extremely helpful reference book tabulating the properties of the 32 crystallographic point groups (character tables, Clebsch-Gordan coecients, compatibility relations, etc.)

Edmonds, Angular Momentum in Quantum Mechanics, (Princeton University Press, 1960)

comprehensive discussion of the (group) theory of angular momentum in quantum mechanics

and many others

Roland Winkler, NIU, Argonne, and NCTU 2011

Special Thanks to my PhD Advisor Prof. Dr. h.c. Ulrich R ossler

from whom I learned group theory

R.W.

Roland Winkler, NIU, Argonne, and NCTU 2011

Introduction and Overview

Denition: Group
A set G = {a, b , c , . . .} is called a group, if there exists a group multiplication connecting any pair a, b G in the following way (1) (2) (3) (4) a, b G : a, b , c G : e G : a G e 1 = e
(left inverse = right inverse) (left neutral = right neutral)

c = ab G (ab )c = a(bc ) ae = a b G : a G a b = e, i.e., b = a

1

(closure) (associativity)
(identity / neutral element)

(inverse element)

Corollaries
(a) (c) (5) (b) a1 a = a a1 = e e a = ae = a a, b G : a G ab = ba

Commutative (Abelian) Group

(commutatitivity)

Order of a Group = number of group elements

Roland Winkler, NIU, Argonne, and NCTU 2011

Examples
integer numbers with addition:
(Abelian group, innite order)

Z, addition Q\{0}, multiplication

rational numbers with multiplication:

(Abelian group, innite order)

complex numbers {exp(2 i m/n) : m = 1, . . . , n} with multiplication

(Abelian group, nite order, example of cyclic group)

invertible (= nonsingular) n n matrices with matrix multiplication

(nonabelian group, innite order, later important for representation theory!)

permutations of n objects: Pn
(nonabelian group, n! group elements)

symmetry operations (rotations, reections, etc.) of equilateral triangle

P3 permutations of numbered corners of triangle more later!

(continuous) translations in
vector addition in

Rn : (continuous) translation group

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

Group Theory in Physics

Group symmetry is the natural language to describe symmetry properties symmetries correspond to conserved quantities symmetry allows us to classify quantum mechanical states
representation theory degeneracies / level splittings

evaluation of matrix elements Wigner-Eckart theorem

e.g., selection rules: dipole matrix elements for optical transitions

...

Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Physics

Classical Mechanics
), Lagrange function L(q, q Lagrange equations If for one j : d dt L q i = L qi i = 1, . . . , N is a conserved quantity

L L = 0 pj qj q j

Examples
qj linear coordinate translational invariance linear momentum pj = const. translation group qj angular coordinate rotational invariance angular momentum pj = const. rotation group
Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Physics

Quantum Mechanics
(1) Evaluation of matrix elements
Consider particle in potential V (x ) = V (x ) even two possiblities e (x ) even o (x ) odd overlapp e (x ) = e (x ) o (x ) = o (x ) i (x ) j (x ) dx = ij i |x |i = i , j = e, o i (x ) dx = 0 i (x ) x

expectation value well-known explanation

product of two even / two odd functions is even product of one even and one odd function is odd integral over an odd function vanishes

Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Physics

Quantum Mechanics
(1) Evaluation of matrix elements (contd)
Group theory provides systematic generalization of these statements representation theory classication of how functions and operators transform under symmetry operations Wigner-Eckart theorem statements on matrix elements if we know how the functions and operators transform under the symmetries of a systems

Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Physics

Quantum Mechanics
(2) Degeneracies of Energy Eigenvalues
= E or i t = H Schr odinger equation H = [O with i t O ,H ] = 0 Let O is conserved quantity O

= E and O = eigenvalue equations H O can be solved simultaneously eigenvalue O of O is quantum number for

Example: H atom
= H
2

2m

2 2 + 2 r r r

2 L e2 2mr 2 r

group SO (3)

2 , H ] = [L z , H ] = [L 2 , L z ] = 0 [L eigenstates nlm (r): 2 , index l L z mL

Roland Winkler, NIU, Argonne, and NCTU 2011

really another example for representation theory

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) V (r + R) = V (r) R {lattice vectors} R : translation operator T R f (r) = f (r + R) T R , H ] = 0 [T

Bloch theorem k (r) = ei kr uk (r) with uk (r + R) = uk (r) wave vector k is quantum number for the discrete translation invariance k rst Brillouin zone

Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Physics

Quantum Mechanics
(3) Solid State Physics
(ii) Phonons Consider square lattice
rotation by 90 o

frequencies of modes are equal degeneracies for particular propagation directions (iii) Theory of Invariants 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

Group Theory in Physics

Quantum Mechanics
(4) Nuclear and Particle Physics
Physics at small length scales: strong interaction Proton mp = 938.28 MeV Neutron mn = 939.57 MeV Symmetry: isospin I with SU(2): proton |1 2
1 2

rest mass of nucleons almost equal degeneracy strong ] = 0 [ I,H neutron

1 |1 2 2

Roland Winkler, NIU, Argonne, and NCTU 2011

Mathematical Excursion: Groups

Basic Concepts
Group Axioms: see above Denition: 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 (nite) group Example: permutation group P3 e= c= 1 2 3 1 2 3 1 2 3 1 3 2 a= d= 1 2 3 3 1 2 1 2 3 3 2 1 b= f = 1 2 3 2 3 1 1 2 3 2 1 3 P3 e a b c d f e e a b c d f a a b e f c d b b e a d f c c c d f e a b d d f c b e a f f c d a b e

{e }, {e , a, b }, {e , c }, {e , d }, {e , f }, G are subgroups of G
Roland Winkler, NIU, Argonne, and NCTU 2011

Conclusions from Group Multiplication Table

P3 e a b Symmetry w.r.t. main diagonal c group is Abelian d order n of g G : smallest n > 0 with g n = e f

e e a b c d f

a a b e f c d

b b e a d f c

c c d f e a b

d d f c b e a

f f c d a b e

{g , g 2 , . . . , g n = e } with g G is Abelian subgroup (a cyclic group) in every row / column every element appears exactly once because: Rearrangement Lemma: for any xed g G , we have G = {g g : g G} = {gg : g G} i.e., the latter sets consist of the elements in G rearranged in order.
proof: g1 = g2 g g1 = g g2 g1 , g2 , g G

Roland Winkler, NIU, Argonne, and NCTU 2011

Mappings of Groups
Let G and G be two groups. A mapping : G G assigns to each g G an element g = (g ) G , with every g G being the image of at least one g G . If (g1 ) (g2 ) = (g1 g2 ) g1 , g2 G , then is a homomorphic mapping of G on G . If the mapping is also one-to-one, then it is an isomomorphic mapping of G on G .
Isomorphic groups have the same structure.

Examples:
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

Goal: Classify elements in a group

(1) Conjugate Elements and Classes

Let a, x G . Then b = x ax 1 is called conjugate to a. Short: a b For xed a, the set of all conjugate elements is called a class. C = {x ax 1 : x G}

Remarks
ab ac bc ba
a = xcx c = x ax b = ycy 1 = (xy 1 )1 a(xy 1 )
1 1

Example: P3 x e a b c d f e e e e e e e a a a a b b b b b b b a a a c c f d c f d d d c f f d c f f d c d c f

ab

identity e is its own class: x e x 1 = e

x ax 1 = ax x 1 = a

x G

Abelian groups: each element is its own class classes are disjoint (no common elements) decompose G into classes in broad terms: similar elements form a class

Roland Winkler, NIU, Argonne, and NCTU 2011

Goal: Classify elements in a group

(2) Subgroups and Cosets

Let U G be a subgroup of G and x G . The set x U {x u : u U} (the set U x ) is called the left coset (right coset) of U . 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

If x x U , then x U = x U any x x U can be used to dene coset x U If U contains s elements, then each coset also contains s elements
(due to rearrangement lemma).

Two left (right) cosets for a subgroup U are either equal or disjoint
(due to rearrangement lemma).

Thus: decompose G into cosets G = U x U y U . . . x, y, . . . /U h Thus Theorem 1: Let h order of G N Let s order of U G s Corollary: The order of a nite group is an integer multiple of the orders of its subgroups. Corollary: If h prime number {e }, G are the only subgroups G is isomorphic to cyclic group
Roland Winkler, NIU, Argonne, and NCTU 2011

Goal: Classify elements in a group

(3) Invariant Subgroups and Factor Groups

connection: classes and cosets A subgroup U G containing only complete classes of G is called invariant subgroup (aka normal subgroup). Let U be an invariant subgroup of G and x G
x U x 1 = U

x U = Ux

(left coset = right coset)

Multiplication of cosets of an invariant subgroup U G : x, y G : (x U ) (y U ) = xy U = z U

where

z = xy

well-dened: (x U ) (y U ) = (U x ) (y U ) = U xy U = z U U = z U

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

Every factor group F = G /U is homomorphic to G .

Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Permutation Group P3

e a b c d f e e a b c d f a a b e f c d b b e a d f c c c d f e a b d d f c b e a f f c d a b e invariant subgroup U = {e , a, b } one coset c U = d U = f U = {c , d , f } factor group P3 /U = {U , c U} U cU U U cU cU cU U

We can think of factor groups G /U as coarse-grained versions of the group G . Often, factor groups G /U are a helpful intermediate step when working out the structure of more complicated groups G . Invariant subgroups are thus more useful subgroups than other subgroups.
Roland Winkler, NIU, Argonne, and NCTU 2011

Matrix Representations of a Group

Motivation
Consider symmetry group Ci = {e , i } e = identity i = inversion two types of basis functions: even and odd more abstract: reducible and irreducible representations matrix representation (based on 1 1 and 2 2 matrices) 1 = {De = 1, Di = 1} compatible with group 2 = {De = 1, Di = 1} multiplication table 10 1 0 3 = De = 0 1 , Di = 0 1 where 1 : even function fe (x ) = fe (x ) 2 : odd functions fo (x ) = fo (x ) irreducible representations
Ci e i e e i i i e

3 : reducible representation: decompose any f (x ) into even and odd parts 1 f (x ) + f (x ) fe (x ) = 2 f (x ) = fe (x )+fo (x ) with 1 fo (x ) = 2 f (x ) f (x ) How to generalize these ideas for arbitrary groups?
Roland Winkler, NIU, Argonne, and NCTU 2011

Matrix Representations of a Group

Let group G = {gi : i = 1, . . . , h} Associate with each gi G a quadratic, nonsingular 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)
= D1 (gi )

D(gi1 )

dimension of representation = dimension of representation matrices Example: G = C = rotations around a xed axis (angle ) C is isomorphic to group of orthogonal 2 2 matrices SO (2) cos sin D2 () = two-dimensional (2D) representation sin cos C is homomorphic to group {D1 () = 1} trivial 1D representation C is isomorphic to group

0 0 D2 ()

higher-dimensional representation

Generally: given matrix representations of dimensions n1 and n2 , we can construct (n1 + n2 ) dimensional representations
Roland Winkler, NIU, Argonne, and NCTU 2011

Goal: Identify and Classify Representations

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 .

Denition: Equivalent Representations Let {D(gi )} be a matrix representation for G with dimension n. Let X be a n-dimensional nonsingular matrix. The set {D (gi ) = X D(gi ) X 1 } forms a matrix representation called equivalent to {D(gi )}.
Convince yourself: {D (gi )} is, indeed, another matrix representation. Matrix representations are most convenient if matrices {D} are unitary. Thus

Theorem 3: Every matrix representation {D(gi )} is equivalent to a unitary representation {D (gi )} where D (gi ) = D 1 (gi ) In the following, it is always assumed that matrix representations are unitary.
Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Theorem 3 (cf. Falicov)

Challenge: Matrix X has to be choosen such that it makes all matrices D (gi ) unitary simultaneously. Let {D(gi ) Di : i = 1, . . . , h} be a matrix representation for G (dimension h).
h

Dene H =
i =1

Di Di (Hermitean) U 1 Di Di1 U = U 1 Di U U 1 Di1 U

i =D =D i

Thus H can be diagonalized by means of a unitary matrix U . d U 1 HU =

i i

=
i

i D D i

with d = d diagonal

Diagonal entries d are positive: d =

i

i ) (D ) = (D i

i ) (D ) = (D i
i i

i ) |2 > 0 |(D

1/2 with elements (d ) d Take diagonal matrix d

Thus

d d = d D (identity matrix) i D d 1=d i

i
Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Theorem 3 (contd)

i d + = d D U 1 Di U d + are unitary matrices Assertion: Di = d equivalent to Di
U 1 equivalent by construction: X = d
unitarity:
=1

Di Di

D + (d i d = d
k

k D d D k ) d+ Di d

= d
k

i D k D D d D i k j = D =D j (rearrangement lemma)

=d qed

Roland Winkler, NIU, Argonne, and NCTU 2011

Reducible and Irreducible Representations

(RRs and IRs)
If for a given representation {D(gi ) : i = 1, . . . , h}, an equivalent representation {D (gi ) : i = 1, . . . , h} can be found that is block diagonal D1 (gi ) 0 gi G D (gi ) = 0 D2 (gi ) then {D(gi ) : i = 1, . . . , h} is called reducible, otherwise irreducible. It is crucial that the same block diagonal form is obtained for all representation matrices D(gi ) simultaneously. Block-diagonal matrices do not mix, i.e., if D (g1 ) and D (g2 ) are block diagonal, then D (g3 ) = D (g1 ) D (g2 ) is likewise block diagonal. Decomposition of RRs into IRs allows one to decompose the problem into the smallest subproblems possible. Goal of Representation Theory Identify and characterize the IRs of a group. We will show The number of inequivalent IRs equals the number of classes.
Roland Winkler, NIU, Argonne, and NCTU 2011

Schurs First Lemma

Schurs 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 c C. () then M is a multiple of the identity matrix M = c 1, Corollaries If () holds with M = c 1, c C, then {D(gi )} is reducible. All IRs of Abelian groups are one-dimensional
Proof: Take gj G arbitrary, but xed. 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

Proof of Schurs First Lemma (cf. Bir & Pikus)

Take Hermitean conjugate of (): M D (gi ) = D (gi ) M Multiply with D (gi ) = D1 (gi ): D(gi ) M = M D(gi )

Thus: () holds for M and M , and also the Hermitean matrices

1 M =2 (M + M ) i M =2 (M M )

It exists a unitary matrix U that diagonalizes M (similar for M ) d = U 1 M U with d = d Thus () implies D (gi ) d = d D (gi ), where D (gi ) = U 1 D(gi ) U more explicitly: D (gi ) (d d ) = 0 i , ,

Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Schurs First Lemma (contd)

Two possibilities: All d are equal, i.e, d = c 1. So M = UdU 1 and M are likewise proportional to 1, and so is M = M iM . Some d are dierent: Say {d : = 1, . . . , r } are dierent from {d : = r + 1, . . . , h}. Thus: D (gi ) = 0
= 1 , . . . , r ; = r + 1 , . . . , h

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

Schurs Second Lemma

Schurs 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 = 0 at least {D1 (gi )} or {D2 (gi )} are reducible. (3) If {D1 (gi )} and {D2 (gi )} are equivalent, then D2 (gi ) = X 1 D1 (gi )X with nonsingular X , so that M = cX with c C.

Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Schurs Second Lemma (cf. Bir & Pikus)

Take Hermitean conjugate of (); use D (gi ) = D1 (gi ) = D(gi1 ), so M D1 (gi1 ) = D2 (gi1 )M Multiply by M on the left; Eq. () implies M D2 (gi1 ) = D1 (gi1 ) M , so MM D1 (gi1 ) = D1 (gi1 )MM gi1 G Schurs rst lemma implies that MM is square matrix with with c C MM = c 1 Case 1: n1 = n2
If c = 0 then det M = 0 because of (*), i.e., M is invertible.

(*)

So () implies

M 1 D1 (gi ) M = D2 (gi ) contrary to assumption.

gi G

So c = 0 and MM = 0, i.e.,
M M = = M M

|M |2 = 0

so that M = 0.
Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Schurs Second Lemma (contd)

Case 2: n1 = n2 (n1 < n2 to be specic)
with det M = 0. Fill up M with n2 n1 rows to get matrix M M = MM , so that However M M ) = (det M ) (det M ) = 0 det(MM ) = det(M
So c = 0, i.e., MM = 0, and as before M = 0.

qed

Roland Winkler, NIU, Argonne, and NCTU 2011

Orthogonality Relations for IRs

Notation: Irreducible Representations (IR): I = {DI (gi ) : gi G} nI = dimensionality of IR I h = order of group G

Theorem 4: Orthogonality Relations for Irreducible Representations (1) two inequivalent IRs I = J
h i =1

DI (gi ) DJ (gi ) = 0

, = 1, . . . , nI , = 1, . . . , nJ

(2) representation matrices of one IR I nI h DI (gi ) DI (gi ) = h i =1

, , , = 1, . . . , nI

Remarks [DI (gi ) : i = 1, . . . , h] form vectors in a h-dim. vector space vectors for dierent I , are orthogonal in total, we have
I

nI2 such vectors; therefore

nI2 h

Corollary: For nite groups the number of inequivalent IRs is nite.

Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Theorem 4: Orthogonality Relations for IRs

(1) two inequivalent IRs I = J Take arbitrary nJ nI matrix X = 0 Let M
i

(i.e., at least one X = 0)

=1

DJ (gi ) X

DI (gi1 )
=M

DJ (gk ) M =
i

DJ (gk ) DJ (gi ) X DJ (gk gi )

i =gj

DI (gi1 )

DI1 (gk ) DI (gk ) DI (gk )

= =
j

DI1 (gk gi )
=gj

DJ (gj ) X

DI (gj1 )

DI (gk ) DI (gk )

= M (Schurs Second Lemma) 0 = M , =

i ,

DJ (gi ) X DI (gi1 )

in particular correct for

X =

=
i

DJ (gi ) DI (gi1 ) DI (gi ) DJ (gi ) qed

Roland Winkler, NIU, Argonne, and NCTU 2011

=
i

Proof of Theorem 4: Orthogonality Relations for IRs (contd)

(2) representation matrices of one IR I First steps similar to case (1): Let M
i

DI (gi ) X DI (gi1 ) with nI nI matrix X = 0

DI (gk ) M = M DI (gk ) (Schurs First Lemma): M = c 1, Thus c =

i

c C choose X =

DI (gi ) X DI (gi1 ) DI (gi ) DI (gi1 ) 1 nI

=
i

c=

1 nI

M =

DI (gi ) DI (gi1 ) DI (e ) =

h nI

qed

Roland Winkler, NIU, Argonne, and NCTU 2011

Goal: Characterize dierent irreducible representations of a group

Characters
Equivalent IRs are related via a similarity transformation D (gi ) = X D(gi )X 1 with X nonsingular This transformation leaves the trace invariant: tr D (gi ) = tr D(gi ) The traces of the representation matrices are called characters (gi ) tr D(gi ) =
i

D(gi )

Equivalent representations have the same characters.

Theorem 5: If gi , gj G belong to the same class Ck of G , then for every representation I of G we have I (gi ) = I (gj )
Proof: gi , gj C x G with gi = x gj x 1
(trace invariant under cyclic permutation)

Thus DI (gi ) = DI (x ) DI (gj ) DI (x 1 ) I (gi ) = tr DI (x ) DI (gj ) DI (x 1 )

= tr DI (x 1 ) DI (x ) DI (gj ) = I (gk )
=1
Roland Winkler, NIU, Argonne, and NCTU 2011

Characters (contd)
Notation I (Ck ) denotes the character of group elements in class Ck The array [I (Ck )] with I = 1, . . . , N k = 1, . . . , N is called character table. (N = number of IRs) = number of classes) (N

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 the number of classes. Then elements in class Ck and N N hk Proof: Use orthogonality I (Ck ) J (Ck ) = IJ I , J = 1, . . . , N relation for IRs h k =1 Interpretation:
I = I (C1 ) . . . I (CN ) I = 1, . . . , N

-dimensional vector space are like N orthonormal vectors in a N . Thus N N

Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Symmetry group C3v of equilateral triangle

(isomorphic to permutation group P3 )
rotoreection = 120
y x

identity

P3 Koster 1 2 3

e E (1) (1)
1 0 0 1

a C3 (1) (1)
1 2
3 2 1 2

b = a2
2 C3

c = ec v (1) (1)

d = ac v (1) (1)
1 2 3 2 2 3

f = bc v (1) (1)
1 2 23

(1) (1)
1 2
2 3

23

23 1 2

1 0

0 1

1 2

P3 e multipli- a cation b table c d f

e e a b c d f

a a b e f c d

b b e a d f c

c c d f e a b

d d f c b e a

f f c d a b e

Character table P3 e a, b c , d , f C3v E 2C3 3v 1 1 1 1 2 1 1 1 3 2 1 0

Roland Winkler, NIU, Argonne, and NCTU 2011

rotoreection = 240
23 1 2

rotation = 240

rotation = 120

reection x x

Notation: Character Tables

A character table is the uniquely dened signature of a group and its IRs I [independent of, e.g., phase conventions for representation matrices DI (gi ) that are quite arbitrary]. Isomorphic groups have the same character tables. Yet: the labeling of IRs I is a matter of convention. Customary:
1 = identity representation: all characters are 1 IRs are often numbered such that low-dimensional IRs come rst;

higher-dimensional IRs come later

If G contains the inversion, a superscript is added to I indicating

the behavior of I under inversion (even or odd)

other labeling schemes are inspired by compatibility relations

(more later)

Dierent authors use dierent conventions to label IRs. To compare such notations we need to compare the uniquely dened 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

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
D (g ) 1 i .. . D1 (gi ) .. D (gi ) = . DN (gi ) ..

a1 . . . aN

D(gi )

. DN (gi )

Theorem 7: Let aI be the multiplicity, with which the IR I {DI (gi )} is contained in the representation {D(gi )}. Then
N

(1) (gi ) = (2) aI =

I =1 1 h

aI I (gi ) I (gi ) (gi ) = hk I (Ck ) (Ck ) k =1 h

Roland Winkler, NIU, Argonne, and NCTU 2011

h i =1

We say: {D(gi )} contains the IR I aI times.

Proof: Theorem 7 (1) due to invariance of trace under similarity transformations

N

(2) we have
J =1 N

aJ J (gi ) = (gi )
h

1 h
h

I (gi )
i =1

J =1

1 aJ h

i =1

1 I (gi ) J (gi ) = h
=IJ

I (gi ) (gi )
i =1

qed

Applications of Theorem 7: Corollary: The IR {D(gi )} is irreducible if and only if

h

|(gi )|2 = h
i =1

Proof: Use Theorem 7 with aI =

1 0

for one I otherwise

Decomposition of Product Representations (see later)

Roland Winkler, NIU, Argonne, and NCTU 2011

Where Are We?

We have discussed the orthogonality relations for irreducible representations 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

The Regular Representation

Finding the IRs of a group can be tricky. Yet for nite groups we can derive the regular representation from the group multiplication table. This representation contains all IRs of the group. g1 =e g2 g3 . . . e e ... Write the group multiplication table 1 g2 e as follows . 1 . g3 e . . . The regular representation is dened as e .

DR (gi ) =

1 0

1 if g g = gi otherwise

Thus for each group element gi we have DR (gi ) = 1, if the entry (, ) in the group multiplication table equals gi , otherwise we have DR (gi ) = 0.

Properties of {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 gi = e (3) R (gi ) = 0 otherwise
Roland Winkler, NIU, Argonne, and NCTU 2011

Proof: Properties of Regular Representation

(1) Matrices {DR (gi )} are nonsingular, as every row / column contains 1 exactly once. Show: if gi gj = gk , then DR (gi )DR (gj ) = DR (gk ) Take i , j , , arbitrary, but xed DR (gi ) = 1 DR (gj ) = 1
1 g 1 only for g g = gi 1 only for g g = gj

g = g gi g = g gj1

DR (gi ) DR (gj ) = 1 g = gi gj = gk

only for g gi = g gj1

[denition of DR (gk ) ]

(2) immediate consequence of denition of DR (gi ) (3) DR (gi ) = R (gi ) =

1 0

1 if gi = g g = e otherwise

DR (gi ) =

h 0

if gi = e otherwise
Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Regular Representation for P3

e a b c d f e e a b c d f a a b e f c d b b e a d f c c c d f e a b d d f c b e a f f c d a b e e 1 = e a1 = b b 1 = a c 1 = c d 1 = d f 1 = f e e b a c d f a a e b f c d b b a e d f c c c f d e a b d d c f b e a f f d c a b e 0 0 1 1 0 0 0 1 0 b= 0 1 0 0 0 1 1 0 0 0 0 1 1 0 0 0 1 0 f = 0 1 0 0 0 1 1 0 0

1 0 0 0 1 0 0 0 1 e= 1 0 0 0 1 0 0 0 1 1 0 0 0 1 0 0 0 1 c= 1 0 0 0 1 0 0 0 1

0 1 0 0 0 1 1 0 0 a= 0 0 1 1 0 0 0 1 0 0 1 0 0 0 1 1 0 0 d = 0 0 1 1 0 0 0 1 0

Roland Winkler, NIU, Argonne, and NCTU 2011

Completeness of Irreducible Representations

Lemma: The regular representation contains every IR nI times, where nI = dimensionality of IR I . Proof: Use Theorem 7: R (gi ) = aI =
1 h

aI I (gi ) where R (e ) = nI
=h i

I (gi ) R (gi )
i

1 h

I (e )
=nI

Corollary: For a group G of order h, the dimensionalities nI of the IRs I obey nI2 = h
I

Theorem 8 (Burnsides Theorem): The representation matrices DI (gi ) of a group G of order h obey the completeness relation nI D (gi ) DI (gj ) = ij i , j = 1, . . . , h (*) h I I , Proof: Theorem 4: [DI (gi ) : i = 1, . . . , h] are orthonormal vectors in a h-dim. vector space completeness relation is (*) Corollary: we have h such vectors
Roland Winkler, NIU, Argonne, and NCTU 2011

Completeness Relation for Characters

Theorem 9: Completeness Relation for Characters If I (Ck ) is the character for class Ck and irreducible representation I , then hk k , k = 1, . . . , N I (Ck )I (Ck ) = kk h
I

Interpretation:

1 (Ck ) . . I = . N (Ck )

k = 1, . . . , N

orthonormal vectors in a N -dimensional vector space are like N N. Thus N Also N N

(from completeness) (from orthogonality)

Number N of irreducible representations of classes = Number N

Character table
square table rows and column form orthogonal vectors
Roland Winkler, NIU, Argonne, and NCTU 2011

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 hk I (Ck ) 1 DI (gi ) = nI
i Ck

The sum over all representation matrices in a class of an IR is proportional to the identity matrix.

Proof of Lemma: For arbitrary gj G DI (gj ) DI (gi ) DI (gj1 ) =

i Ck

i Ck

DI (gj ) DI (gi ) DI (gj1 ) =

=DI (gi ) with i Ck

DI (gi )

i Ck

because gj maps gi1 = gi2 onto gi1 = gi2

(Schurs First Lemma):

i Ck

DI (gi ) = ck 1 qed

ck =

1 tr nI

DI (gi ) =
i Ck

hk I (Ck ) nI

Roland Winkler, NIU, Argonne, and NCTU 2011

Proof of Theorem 9: Completeness Relation for Characters

Use Theorem 8 (Completeness Relations for Irreducible Representations)
N I =1 ,

nI D (gi ) DI (gj ) = ij h I DI (gi )

i Ck j Ck

nI h

DI (gj )
j Ck

= hk kk

i Ck

hk (Ck ) nI I

hk I (Ck ) nI
, = nI

(Lemma)

hk hk I (Ck ) I (Ck ) nI2

qed

Roland Winkler, NIU, Argonne, and NCTU 2011

Summary: Orthogonality and Completeness Relations

Theorem 4: Orthogonality Relations for Irreducible Representations h I , J = 1, . . . , N nI , = 1 , . . . , nI DI (gi ) DJ (gi ) = IJ h , = 1, . . . , nJ
i =1

Theorem 8: Completeness Relations for Irreducible Representations

N

I =1 ,

nI D (gi ) DI (gj ) = ij h I

i , j = 1, . . . , h

Theorem 6: Orthogonality Relations for Characters

N

k =1

hk (Ck ) J (Ck ) = IJ h I

I , J = 1, . . . , N

Theorem 9: Completeness Relation for Characters hk h

N

I (Ck )I (Ck ) = kk
I =1

k , k = 1, . . . , N
Roland Winkler, NIU, Argonne, and NCTU 2011

Unreducible

More on Irreducible h h Problems

hhh hhhh

Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Quantum Mechanics

Topics:
behavior of quantum mechanical states and operators under symmetry operations relation between irreducible representations and invariant subspaces of the Hilbert space connection between eigenvalue spectrum of quantum mechanical operators and irreducible representations selection rules: symmetry-induced vanishing of matrix elements and Wigner-Eckart theorem

Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry Operations in Quantum Mechanics (QM)

Let G = {gi } be a group of symmetry operations of a qm system
e.g., translations, rotations, permutation of particles

Translated into the language of group theory: In the Hilbert space of the qm system we have a group of unitary (gi )} such that G is isomorphic to G . operators G = {P

Examples
translations Ta (Ta ) = exp (i p a/ ) unitary operator P ( p = momentum) 2 1 P (Ta ) (r) = 1 + a + ( a) + . . . (r) = (r + a)
2

rotations R (n, ) = exp i L n/ unitary operator P

= angular momentum (L = angle of rotation n = axis of rotation)

Roland Winkler, NIU, Argonne, and NCTU 2011

Transformation of QM States
Let {| } be an orthonormal basis (gi ) be the symmetry operator for the symmetry transformation gi Let P with symmetry group G = {gi }. (gi ) | = Then P

(gi ) | | | P
D (gi )

1=

| |

So

(gi ) | = P

matrix of a unitary

D(gi ) |

where D(gi ) = representation of G

(gi ) unitary because P

Note: bras and kets transform according to complex conjugate representations (gi ) = |P | D(gi )

Let gi , gj G with gi gj = gk G . Then (gi ) P (gj ) | = P

D (gi ) D (gi )
Roland Winkler, NIU, Argonne, and NCTU 2011

(gi ) | | P (gi ) | = | | P

D(gk ) |

Transformation of QM States (contd)

in general: representation {D(gi )} is reducible We have D (gi ) = U 1 D(gi ) U More explicitly: D (gi ) =
1 U D (gi ) U (gi ) | | P

1 |U P (gi ) U |

(gi ) | = |P with | =

U |

Thus: block diagonalization {D(gi )} {D (gi ) = U 1 D(gi ) U } corresponds to change of basis {| } {| =

Roland Winkler, NIU, Argonne, and NCTU 2011

U | }

Relevance of Irreducible Representations

Invariant Subspaces
Denition: Let G = {gi } be a group of symmetry transformations. Let H = {| } be a Hilbert space with states | . A subspace S H is called invariant subspace (with respect to G ) if (gi ) | S P gi G , | S 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: Suppose {D(gi )} is reducible. unitary transformation U with {D (gi ) = U 1 D(gi ) U } block diagonal For {D (gi )} we have the basis {| = U | } The block diagonal form of {D (gi )} implies that {| is reducible
Roland Winkler, NIU, Argonne, and NCTU 2011

Invariant Subspaces (contd)

Corollary: Every Hilbert space H can be decomposed into irreducible invariant subspaces SI transforming according to the IR I Remark: Given a Hilbert space H we can generally have multiple (possibly orthogonal) irreducible invariant subspaces SI SI = |I : = 1, . . . , nI transforming according to the same IR I (gi ) |I = P DI (gi ) |I

Theorem 11: (1) States transforming according to dierent IRs are orthogonal (2) For states |I and |I transforming according to the same IR I we have I | I = I || I where the reduced matrix element I || I is independent of , . Remark: This theorem lets us anticipate the Wigner-Eckart theorem.
Roland Winkler, NIU, Argonne, and NCTU 2011

Invariant Subspaces (contd)

Proof of Theorem 11 (gj ): Use unitarity of P Then I |J =
1 h

(gj ) = (gj ) P 1=P

1 h

(gi ) (gi ) P P
i

(gi ) |J (gi ) P I | P
i

I |DI (gi )

DJ (gi )

|J

I |J
1 IJ n I

1 h

DI (gi ) DJ (gi )
(1/nI ) IJ

I |I
I || I

qed

Roland Winkler, NIU, Argonne, and NCTU 2011

Decomposition into Invariant Subspaces

Goal: Decompose general state | H into components from invariant subspaces SI I = Projection operator
nI h i

DI (gi ) P (gi )
= I (gi )

Let | =
J

cJ |J
nI h

with coecients cJ (gi ) |J cJ P

J

I | =

DI (gi )
i nI h

DJ (gi )

|J

=
J

cJ

DI (gi ) DJ (gi ) |J
i IJ

cI |I

I projects a general state | on the invariant subspace SI Thus: transforming according to the IR I = {DI (gi )}.
Roland Winkler, NIU, Argonne, and NCTU 2011

Decomposition into Invariant Subspaces (contd)

Example:
Group Ci = {e , i } e = identity i = inversion
Ci e i 1 1 1 2 1 1 Ci e i e e i i i e

character table (e ) (x ) = (x ), P

(i ) (x ) = (x ) P
nI h

I = Projection operator 1 = 2 =
1 2 1 2

I (gi ) P (gi )
i 1 2 1 2

with nI = 1, h = 2

(e ) + P (i )] [P (e ) P (i )] [P

1 (x ) = 2 (x ) =

[ (x ) + (x )] even part [ (x ) (x )] odd part

Roland Winkler, NIU, Argonne, and NCTU 2011

Product Representations
Let {I (x ) : = 1, . . . nI } and {J (x ) : = 1, . . . nJ } denote basis functions for the invariant subspaces SI and SJ . Consider the product functions {I (x ) J (x ) : = 1, . . . nI , = 1, . . . nJ }. How do these functions transform under G ? Denition: 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 ( ) and column ( ) are given by [DI (g ) DJ (g )] , = DI (g ) DJ (g ) Example: Let DI (g ) = DI (g ) DJ (g ) =
x11 x12 x21 x22 , = 1, . . . , nI , = 1, . . . , nJ y11 y12 y21 y22 x11 y12 x11 y22 x12 y12 x12 y22

and DJ (g ) =
x11 y11 x11 y21 = x12 y11 x12 y21

x11 DJ (g ) x12 DJ (g ) x21 DJ (g ) x22 DJ (g )

x12 y11 x12 y21 x22 y11 x22 y21

x12 y12 x12 y22 x22 y12 x22 y22

Details of the arrangement in the following not relevant

Roland Winkler, NIU, Argonne, and NCTU 2011

Dimension of product matrix dim [DI (g ) DJ (g )] = dim DI (g ) dim DJ (g )

Product Representations
Let I = {DI (gi )} and J = {DJ (gi )} be representations of G with characters I (gi ) and J (gi ). Then I J {DI (g ) DJ (g )} is a representation of G called product representation.
I J is, indeed, a representation: Let DI (gi ) DI (gj ) = DI (gk ) and DJ (gi ) DJ (gj ) = DJ (gk ) [DI (gi ) DJ (gi )] [DI (gj ) DJ (gj )] =
,

DI (gi ) DJ (gi ) DI (gj ) DJ (gj )

DI (gk ) = [DI (gk ) DJ (gk )]

DJ (gk )
,

The characters of the product representation are I J (gi ) = I (gi ) J (gi )

Roland Winkler, NIU, Argonne, and NCTU 2011

Decomposing Product Representations

Let I = {DI (gi )} and J = {DJ (gi )} be irreducible representations of G The product representation I J = {DI J (gi )} is generally reducible According to Theorem 7, we have I J =
K IJ aK K IJ where aK = N

hk K (Ck ) I J (Ck ) k =1 h
=I (Ck ) J (Ck )

The multiplication table for the irreducible representations I of G IJ lists k aK K Example: Permutation group P3 1 2 3 e 1 1 2 a, b c , d , f 1 1 1 1 1 0 I J 1 2 3

I J 1 2 . . . 1 1 2 2 1

...

3 3 3 1 + 2 + 3

Roland Winkler, NIU, Argonne, and NCTU 2011

Discussion
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. u = (u1 , . . . , un ) V : u u V
D (gi )

with u =

D(gi ) u

Irreducible Representation (IR) Invariant Subspace The decomposition of a reducible representation into IRs I corresponds to a decomposition of the vector space V into invariant subspaces VI such that VI VI
DI (gi )

gi G

(i.e., no mixing)

This decomposition of V lets us break down a big physical problem into smaller, more tractable problems

Product Representation Product Space A product representation I J describes a linear mapping of the product space VI VJ onto itself VI VJ VI J
DI J (gi )

gi G

IJ The block diagonalization I J = K aK K corresponds to a decomposition of VI VJ into invariant subspaces VK

Roland Winkler, NIU, Argonne, and NCTU 2011

Discussion (contd)
Clebsch-Gordan Coecients
IJ The block diagonalization I J = K aK K corresponds to a decomposition of VI VJ into invariant subspaces VK

Change of Basis: unitary transformation

IJ aK

VI VJ
J old basis {eI e }

K =1

VK

new basis {eK }

J eI e

Thus

eK =

IJ

I J K

where

I J K

= Clebsch-Gordan coecients (CGC)

Clebsch-Gordan coecients describe the unitary transformation for the decomposition of the product space VI VJ into invariant subspaces VK
Roland Winkler, NIU, Argonne, and NCTU 2011

Clebsch-Gordan Coecients (CGC)

Remarks
CGC are independent of the group elements gi CGC are tabulated for all important groups (e.g., Koster, Edmonds) Note: Tabulated CGC refer to a particular denition (phase convention) for the basis vectors {eI } and representation matrices {DI (gi )} Clebsch-Gordan coecients C describe a unitary basis transformation C C = C C = 1 Thus Theorem 12: Orthogonality and completeness of CGC I J K I J K
K
Roland Winkler, NIU, Argonne, and NCTU 2011

I J K I J K

= KK =

Clebsch-Gordan Coecients (contd)

Clebsch-Gordan coecients diagonalize the representation matrices (unitary transformation) (1)
11111 00000 00000 11111 00000 11111 00000 11111 00000 11111 IxJ 00000 11111 00000 11111 00000 11111 00000 11111 111 000 000 111 000 111 00 11 000 111 00 11 00 11

=C = C

111 000 000 111 000 111 00 11 000 111 00 11 00 11

00 11 11 00 00 11

C C

(2) More explicitly:

11 00 00 11 00 11

11111 00000 00000 11111 00000 11111 00000 11111 00000 11111 IxJ 00000 11111 00000 11111 00000 11111 00000 11111

Theorem 13: Reduction of Product Representation I J (1) DI (gi ) DJ (gi ) =

K

I J K

DK (gi )

I J K

(2)

DK (gi ) KK =

I J K

DI (gi ) DJ (gi )

I J K

Remark: Given the representation matrices {DI (gi )}, these formulas can be used to evaluate successively the Clebsch-Gordan coecients
Roland Winkler, NIU, Argonne, and NCTU 2011

Example: CGC for permutation group P3

IJ Here we always have aK 1, so that we may drop the index .

11 1 11 1 13 3 1 33 3 1

12 2 11 1 1 0 0 1

22 1 11 1

=1 0 1 1 0 =

23 3 1

0 1/ 2 0 1/ 2

33 3 2

0 1/ 2 1/ 2 0

Roland Winkler, NIU, Argonne, and NCTU 2011

Comparison: Rotation Group

Angular momentum j = 0, 1/2, 1, 3/2, . . . corresponds to the irreducible representations of the rotation group For each j , these IRs are (2j + 1)-dimensional, i.e., the z component of angular momentum labels the basis states for the IR j . j = 0 is the identity representation 1 of the rotation group The product representation I J corresponds to the addition of angular momenta j1 and j2 ; j1 j2 = |j1 j2 | + . . . + j1 +j2
1 j2 are one. Here all multiplicities ajj3

In our lecture, Clebsch-Gordan coecients have the same meaning as in the context of the rotation group: They describe the unitary transformation from the reducible product space to irreducible invariant subspaces. This unitary transformation depends only on (the IRs of) the symmetry group of the problem so that the CGC can be tabulated.
Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry of Observables
. Consider Hermitian operator (observable) O Let G = {gi } be a group of symmetry transformations (gi )} the group of unitary operators isomorphic to G . with {P (gi ) O (gi )1 = O (gi ), O P [P ] = 0 If P gi G . we call G the symmetry group of O
Of course, we want the largest G possible.

, i.e., O |n = n |n , Lemma: If |n is an eigenstate of O (gi ), O (gi ) |n is likewise an eigenstate of O ] = 0, then P and [P for the same eigenvalue n .
(gi ) |n need not be orthogonal to |n . As always P (gi ) |n ] = P (gi ) O (gi ) |n ] [P |n = n [P Proof: O

Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry of Observables (contd)

Theorem 14: (gi )} be the symmetry group of the observable O . Let G = {P Then the eigenstates of a d -fold degenerate eigenvalue n of O form a d -dimensional invariant subspace Sn .
The proof follows immediately from the preceding lemma.

Most often: Sn is irreducible Identify d -fold degeneracy of n with d -dimensional IR of G . Under which cirumstances can Sn be reducible?
G does not include all symmetries realized in the system, i.e., G

G (hidden symmetry). Then Sn is an irreducible invariant subspace of G . Examples: hydrogen atom, m-dimensional harmonic oscillator (m > 1).

A variant of the preceding case: The extra degeneracy is caused by the

antiunitary time reversal symmetry (more later).

The degeneracy cannot be explained by symmetry: rare! (Usually such accidental degeneracies correspond to singular points in the parameter space of a system.)
Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry of Observables (contd)

Remarks:
= Hamiltonian =H Let O

according to the IRs I Classify the eigenvalues and eigenstates of H . of the symmetry group G of H |I , = EI |I , Notation: H = 1, . . . , nI
: distinguish dierent levels transforming according to same I

If I is nI -dimensional, then eigenvalues EI are nI -fold degenerate.

Note: In general, the quantum number I cannot be associated directly with an observable.

For given EI , it suces to calculate one eigenstate |I 0 , . Then (gi ) |I 0 , : gi G} {|I , : = 1, . . . , nI } = {P

(i.e., both sets span the same subspace of H)

Expand eigenstates |I , in a symmetry-adapted basis |I , =

J , =IJ I || I

|J , :

J = 1, . . . , N ; = 1 , . . . , nJ ; = 1, 2, . . .

J , |I , |J , =

I || I |I ,

see Theorem 11

independent of specic details partial diagonalization of H

Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry of Observables (contd)

Eect of Perturbations
=H 0 + H 1, Let H 0 = unperturbed Hamiltonian, H 1 = perturbation H 0 H H usually G G0

Let G0 = symmetry group of G = symmetry group of

The unperturbed eigenkets {|n } transform according to IRs 0 I of G0 {0 I } are also representations of G , yet then reducible Every IR 0 I of G0 breaks down into (usually multiple) IRs {J } of G 0 I =
J

aJ J

(see Theorem 7)

compatibility relations for irreducible representations

Compatibility relations tell us how a degenerate level transforming according to the IR 0 I of G0 splits into multiple levels transforming according to certain IRs 1 reduces the symmetry from G0 to G G0 . {J } of G when the perturbation H
Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Compatibility Relations for C3v

Character table C3v P3 C3v P3 1 2 3 E e 1 1 2 2C3 3v a, b c , d , f 1 1 1 1 1 0

P3

C3v

1 2 = C3 } G1 = {e , a, b } P3 has two subgroups C3 = {E , C3 , C3 Cs = {E , v } G2 = {e , c } = {e , d } = {e , f }

Both are Abelian, so they have only one-dimensional IRs C3 E C3 2 C3 E E C3 2 C3 C3 C3 2 C3 E

2 C3 2 C3 E C3

C3 1 2 3

2 E C3 C3 1 1 1 1 1 e 2i /3

Cs E i E E i i i E
C3 2 3 1 1

Cs E i 1 1 1 2 1 1
C 3v Cs 1 2 2 1

compatibility relations C3v P3 1 2 3 C3 G1 1 1 2 + 3 Cs G2 1 2 1 + 2

2 2fold degen. 2 nondegen. 2 nondegen.

Roland Winkler, NIU, Argonne, and NCTU 2011

Discussion: Compatibility Relations

Compatibility relations tell us how a degenerate level transforming according to the IR 0 I of G0 splits into multiple levels transforming 1 reduces according to certain IRs {J } of G when the perturbation H the symmetry from G0 to G G0 .

Thus qualitative statements:

1? Which degenerate levels split because of H 1? Which degeneracies remain unaected by H These statements do not require any perturbation theory in the conventional sense. (For every pair G0 and G , they can be tabulated once and forever!) 1. These statements do not require some kind of smallness of H But no statement whether (or how much) a level will be raised 1. or lowered by H
Roland Winkler, NIU, Argonne, and NCTU 2011

(Ir)Reducible Operators
requires Up to now: symmetry group of operator O (gi ) O (gi )1 = O P P gi G : = 1, . . . , n} with More general: A set of operators {Q (gi ) Q (gi )1 = P P
n

D(gi ) Q

=1

= 1, . . . , n gi G

is called reducible (irreducible), if = {D(gi ) : gi G} is a reducible (irreducible) representation of G .

(gi ) Q (gi )1 P P Often a shorthand notation is used: gi Q

} transform according to . We say: The operators {Q } will not transform Note: In general, the eigenstates of {Q according to .
Roland Winkler, NIU, Argonne, and NCTU 2011

(Ir)Reducible Operators (contd)

Examples:
1 = identity representation; D(gi ) = 1 gi G ; (gi ) Q (gi )1 = Q P P gi G nI = 1

is a scalar operator or invariant. We say: Q most important scalar operator: the Hamiltonian H i.e., H always transforms according to 1 is the largest symmetry group The symmetry group of H that leaves H invariant. momentum operator p = i x
= 1, 2, 3

{p : = 1, 2, 3} transforms according to a three-dimensional representation (possibly reducible!)

composite operators (= tensor operators) e.g., angular momentum i x

,

x x x

= 1, 2, 3

Roland Winkler, NIU, Argonne, and NCTU 2011

Tensor Operators
I Let {Q : = 1, . . . , nI } J {Q : = 1, . . . , nJ } I J = 1 , . . . , nI Then Q Q : = 1, . . . , nJ representation I J transform according to transform according to I = {DI (gi )} J = {DJ (gi )}

transforms according to the product

I J is, in general, reducible J I The set of tensor operators Q Q is likewise reducible

A unitary transformation brings I J = {DI (gi ) DJ (gi )} into block-diagonal form J I The same transformation decomposes Q Q into irreducible tensor operators (use CGC)

Roland Winkler, NIU, Argonne, and NCTU 2011

Where Are We?

We have discussed the transformational properties of states the transformational properties of operators Now: the transformational properties of matrix elements Wigner-Eckart Theorem

Roland Winkler, NIU, Argonne, and NCTU 2011

Wigner-Eckart Theorem
Let {|I , : = 1, . . . , nI } {|I , : = 1, . . . , nI } J {Q : = 1, . . . , nJ } Then J I , |Q | I , = transform according to I = {DI (gi )} transform according to I = {DI (gi )} transform according to J = {DJ (gi )} J I I J || I I || Q is independent

where the reduced matrix element of , and .

Proof:
= 1 , . . . , nI J Q |I , : = 1, . . . , nJ

J || I I || Q

transforms according to I J J I K |K ,
where = , I , J , ) (Q

J Thus CGC expansion Q |I , =

K ,

J I , |Q | I , =
K ,

J I K

I , | K , I
K

Theorem 11

J || I I || Q

Roland Winkler, NIU, Argonne, and NCTU 2011

Wigner-Eckart Theorem (contd)

Discussion
Matrix elements factorize into two terms
the reduced matrix element independent of , and CGC indexing the elements , and of I , I and J .

) (CGC are tabulated, independent of Q

Thus:

reduced matrix element = physics CGC = geometry

Matrix elements for dierent values of , and have a xed ratio independent of Q If I is not contained in I J
J I I =0 , , , ,
Equivalent to: If J I does not I contain the identity representation

J I , |Q | I , = 0

Many important selection rules are some variation of this result.

Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Optical Selection Rules

Example: Optical transitions for a system with symmetry group C3v (e.g., NH3 molecule) Optical matrix elements where iI |e r|fJ (dipole approximation)

|iI = initial state (with IR I ); |fJ = nal state (IR J ) e = (ex , ey , ez ) = polarization vector r = ( x, y , z ) = dipole operator ( position operator)

x , y transform according to 3 z transforms according to 1 for example xy polarized: i1 |ex x + ey y |f3

transition allowed because 3 3 = 1 + 2 + 3 in total 4 dierent matrix elements, but only one reduced matrix

element

z polarized: i1 |ez z |f3

transition forbidden because 1 3 = 3

These results are independent of any microscopic models for the NH3 molecule!
Roland Winkler, NIU, Argonne, and NCTU 2011

Goal: Spin 1/2 Systems and Double Groups

Rotations and Euler Angles

So far: transformation of functions and operators dependent on position Now: systems with spin degree of freedom wave functions are two-component Pauli spinors (r) (r) How do Pauli spinors transform under symmetry operations? Parameterize rotations via Euler angles , , (r) = (r) | + (r) |
z z y y z z z y y y x x x x x x z z z z y y y y x x x

axis z, angle

axis y, angle

axis z, angle

Thus general rotation R (, , ) = Rz ( ) Ry ( ) Rz ()

Roland Winkler, NIU, Argonne, and NCTU 2011

Rotations and Euler Angles (contd)

General rotation R (, , ) = Rz ( ) Ry ( ) Rz () Diculty: axes y and z refer to rotated body axes (not xed in space)
1 Use Ry ( ) = Rz () Ry ( ) Rz ( ) 1 Rz ( ) = Ry ( ) Rz ( ) Ry ( )

preceding rotations are temporarily undone

1 Thus R (, , ) = Ry ( ) Rz ( ) Ry ( ) Ry ( ) Rz ()
1 Rz () Ry ( ) Rz ()

=1

rotations about z axis commute

Thus R (, , ) = Rz () Ry ( ) Rz ( )

rotations about space-xed axes!

More explicitly: rotations of (x , y , z ) R3 vectors r =

cos sin 0 Rz () = sin cos 0 , 0 0 1

cos 0 sin 1 0 R y ( ) = 0 sin 0 cos

etc.

SO (3) = set of all rotation matrices R (, , ) = set of all orthogonal 3 3 matrices R with det R = +1. R (2, 0, 0) = R (0, 2, 0) = R (0, 0, 2 ) = 1 e
Roland Winkler, NIU, Argonne, and NCTU 2011

Rotations: Spin 1/2 Systems

Rotation matrices for spin-1/2 spinors Rn () = exp
i 2

(axis n)

n = 1 cos(/2) i n sin(/2)

R(, , ) = Rz () Ry ( ) Rz ( ) = e i (+ )/2 cos(/2) e i ( )/2 sin(/2) e i ( )/2 sin(/2) e i (+ )/2 cos(/2) transformation matrix for spin 1/2 states SU (2) = set of all matrices R(, , ) = set of all unitary 2 2 matrices R with det R = +1. R(2, 0, 0) = R(0, 2, 0) = R(0, 0, 2 ) = 1 e R(4, 0, 0) = R(0, 4, 0) = R(0, 0, 4 ) = 1 = e
rotation by 2 is not identity rotation by 4 is identity

Every SO (3) matrix R (, , ) corresponds to two SU (2) matrices R(, , ) and R( + 2, , ) = R(, + 2, ) = R(, , + 2 ) =e R(, , ) = R(, , ) e SU (2) is called double group for SO (3)
Roland Winkler, NIU, Argonne, and NCTU 2011

Double Groups
Deniton: Double Group Let the group of spatial symmetry transformations of a system be G = {gi = R (i , i , i ) : i = 1, . . . , h} SO (3) Then the corresponding double group is Gd = {gi = R(i , i , i ) : i = 1, . . . , h} {gi = R(i + 2, i , i ) : i = 1, . . . , h} SU (2)

Thus with every element gi G we associate two elements gi and g i e gi = gi e Gd If the order of G is h, then the order of Gd is 2h. Note: G is not a subgroup of Gd because the elements of G are not a closed subset of Gd . Example: Let g = rotation by
in G : g 2 = e in Gd : g 2 = e

the same group element g is thus interpreted dierently in G and Gd

Yet: {e , e } is invariant subgroup of Gd and the factor group Gd /{e , e } is isomorphic to G . The IRs of G are also IRs of Gd
Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Double Group C3v

C3v 1 2 3 4 5 6 E 1 1 2 2 1 1 E 1 1 2 2 1 1 2C3 1 1 1 1 1 1 3 2C 1 1 1 1 1 1 3v 1 1 0 0 i i 3 v 1 1 0 0 i i

For 1 , 2 , and 3 the barred group elements have the same characters as the unbarred elements. Here the double group gives us the same IRs as the single group For other groups a class may contain both barred and unbarred group elements. the number of classes and IRs in the double group need not be twice the number of classes and IRs of the single group
Roland Winkler, NIU, Argonne, and NCTU 2011

Time Reversal (Reversal of Motion)

: t t Time reversal operator : Action of 1 = r r 1 p = p 1 L = L 1 S = S independent of t linear in t

t / ( t ) = 1 i H Consider time evolution: U | = U ( t ) ( t ) | U | = iH | i H Need = UK

but need also

H ] = 0 [,

U = unitary operator K = complex conjugation

= UK : Properties of K c 1 | + c 2 | = c 1 | + c2 | | and | = | Let | = | = | = |

(antilinear)

= UK is antiunitary operator

Roland Winkler, NIU, Argonne, and NCTU 2011

Time Reversal (contd)

depends on the representation The explicit form of 1 = position representation: r r p 1 = p momentum representation: (r) = (r) (p) = (p)

spin 1/2 systems: = i y K 2 = 1 are at least two-fold degenerate all eigenstates |n of H (Kramers degeneracy)

Roland Winkler, NIU, Argonne, and NCTU 2011

Time Reversal and Group Theory

. Consider a system with Hamiltonian H Let G = {gi } be the symmetry group of spatial symmetries of H (gi ), H ] = 0 gi G [P Let {|I : = 1, . . . , nI } be an nI -fold degenerate eigenspace of H which transforms according to IR I = {DI (gi )} |I = EI |I H (gi ) |I P =

DI (gi ) |I

be time-reversal invariant: [H , ] = 0 Let H (gi )}) with is additional symmetry operator (beyond {P P (gi )] = 0 [, (gi ) |I = P (gi )|I = P

DI (gi ) |I =

|I DI (gi )

|I } transform according to Thus: time-reversed states { complex conjugate IR I = {DI (gi )}

Roland Winkler, NIU, Argonne, and NCTU 2011

Time Reversal and Group Theory (contd)

|I } transform according to time-reversed states { complex conjugate IR I = {DI (gi )} Three possiblities (known as cases a, b, and c) |I } are linear dependent (a) {|I } and { |I } are linear independent (b) {|I } and { The IRs I and I are distinct, i.e., I (gi ) = I (gi ) |I } are linear independent (c) {|I } and { I = I , i.e., I (gi ) = I (gi ) gi Discussion Case (a): time reversal is additional constraint for {|I } e.g., nI = 1 | reell Cases (b) and (c): time reversal results in additional degeneracies
Our denition of cases (a)(c) follows Bir & Pikus. Often (e.g., Koster) a dierent classication is used which agrees with Bir & Pikus for spinless systems. But cases (a) and (c) are reversed for spin-1/2 systems.
Roland Winkler, NIU, Argonne, and NCTU 2011

Time Reversal and Group Theory (contd)

When do we have case (a), (b), or (c)? Criterion by Frobenius & Schur case (a) 1 2 0 case (b) I (gi ) = h i case (c) where = +1 systems with integer spin 1 systems with half-integer spin

Proof: see, e.g., Bir & Pikus

Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Cyclic Group C3

C3 is Abelian group with 3 elements: C3 = {q , q 2 , q 3 e } Multiplication table C3 e q q e e q q2 q q q2 e q2 q2 e q
2

Character table C3 e q q2
time reversal

1 1 1 1 2 1 3 1
e 2i /3

a b b

IR 1 : no additional degeneracies because of time reversal IRs 2 and 3 : these complex IRs need to be combined two-fold degeneracy because of time reversal symmetry (though here no spin!)

Roland Winkler, NIU, Argonne, and NCTU 2011

Group Theory in Solid State Physics

First: Some terminology Lattice: periodic array of atoms (or groups of atoms) Bravais lattice: Rn = n1 a1 + n2 a2 + n3 a3
n = (n1 , n2 , n3 ) Z3 ai linearly independent

Every lattice site Rn is occupied with one atom Example: 2D honeycomb lattice is not a Bravais lattice Lattice with basis:
Every lattice site Rn is occupied with z atoms Position of atoms relative to Rn :

i,

i = 1, . . . , z

These z atoms with relative positions i form a basis. Example: two neighboring atoms in 2D honeycomb lattice

Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry Operations of Lattice

Translation t (not necessarily by lattice vectors Rn ) 3 3 matrices Rotation, inversion

Combinations of translation, rotation, and inversion general transformation for position vector r R3 : r = r + t {|t} r Notation {|t} includes also
Mirror reection = rotation by about axis perpendicular

to mirror plane followed by inversion

Glide reection = translation followed by reection Screw axis = translation followed by rotation

Symmetry operations {|t} form a group Multiplication { |t } {|t} r = r + t = r + t + t = { | t + t } r r =r+t

because {|t}1 {|t} = {1 |1 t 1 t} = {1|0}
Roland Winkler, NIU, Argonne, and NCTU 2011

Inverse Element {|t}1 = {1 | 1 t}

Classication

Symmetry Groups of Crystals

to be added . . .

Roland Winkler, NIU, Argonne, and NCTU 2011

Symmetry Groups of Crystals

Translation Group
Translation group = set of operations {1|Rn } {1|Rn } {1|Rn } = {1|Rn + 1 Rn } = {1|Rn +n } Abelian group associativity (trivial) identity element {1|0} = {1|R0 } inverse element {1|Rn }1 = {1| Rn } Translation group Abelian only one-dimensional IRs

Roland Winkler, NIU, Argonne, and NCTU 2011

Irreducible Representations of Translation Group

(for clarity in one spatial dimension)

Consider translations {1|a} a = T {1|a} is unitary operator Translation operator T eigenvalues have modulus 1 a | = e i | < eigenvalue equation T in na | = e more generally T | nZ representations D({1|Rna }) = e in
r a |

<

Physical Interpretation of a | Consider r | T

e i |

r a| = e i r |

Thus: Bloch Theorem (for = ka) The wave vector k (or = ka) labels the IRs of the translation group The wave functions transforming according to the IR k are Bloch functions r | = e ikr uk (r ) with e ik (r a) uk (r a) = e ikr uk (r )e ika or uk (r a) = uk (r )
Roland Winkler, NIU, Argonne, and NCTU 2011

Irreducible Representations of the Space Group

to be added . . .

Roland Winkler, NIU, Argonne, and NCTU 2011

Theory of Invariants
Idea:

Luttinger (1956) Bir & Pikus

Hamiltonian must be invariant under all symmetry transformations of the system Example: free particle 2 = Ekin = p H 2m
scalar

+ X + c4 p + H c1H p c2 X r X p 4 + . . .
not inversion symmetric breaks time reversal scalar

Crystalline solids: Ekin = E (k) = kinetic energy of Bloch electron with crystal momentum p = k dispersion E (k) must reect crystal symmetry E (k) = a0 + a1 k +a2 k 2 + a3 k 3 + . . .
only in crystals without inversion symmetry
Roland Winkler, NIU, Argonne, and NCTU 2011

Theory of Invariants (contd)

More generally:
Bands E (k) at expansion point k0 n-fold degenerate Bands split for k = k0 Example: GaAs (k0 = 0) Band structure E (k) for small k via diagonalization of n n matrix Hamiltonian H(k).
k E conduction band valence band HH LH SO

Goal: Set up matrix Hamiltonian H(k) by exploiting the symmetry at expansion point k0 Incorporate also perturbations such as

spin-orbit coupling (spin S) electric eld E , magnetic eld B strain etc.

Roland Winkler, NIU, Argonne, and NCTU 2011

Invariance Condition
Consider n n matrix Hamiltonian H(K) K = K (k, S, E , B, , . . .) = general tensor operator where k = wave vector E = electric eld S = spin B = magnetic eld = strain eld etc.

Basis functions { (r) : = 1, . . . , n} transform according to representation = {D(gi )} of group G . ( need not be irreducible) Symmetry transformation gi G applied to tensor K (gi ) K P (gi )1 K gi K P H(K) H(gi K) D(gi1 ) (r) Equivalent to inverse transformation gi1 applied to (r): (r) Thus H(K) (gi r) =

D(gi ) H(K) D(gi1 ) gi G

Roland Winkler, NIU, Argonne, and NCTU 2011

D(gi1 ) H(gi K) D(gi ) = H(K)

really n2 equations!

Invariance Condition (contd)

Remarks
If the expansion point k0 is in the same star as k0 , additional constraints arise from time reversal symmetry. If is reducible, the invariance condition can be applied to each irreducible block of H(K) (see below).

Roland Winkler, NIU, Argonne, and NCTU 2011

Invariant Expansion
Expand H(K) in terms of irreducible tensor operators and basis matrices Decompose tensors K into irreducible tensors KI transforming according to IR I of G
I I K gi K I DI (gi ) K

n2 linearly independent basis matrices {Xq : q = 1, . . . , n2 } transforming as Xq D(gi1 ) Xq D(gi ) = DX (gi )pq Xp
p

with expansion coecents DX (gi )pq Representation {DX (gi )} is reducible.

We have IR for the ket basis functions of H and IR (i.e., the complex conjugate IR) for the bras
I from {Xq : q = 1, . . . , n2 } form linear combinations X transforming according to the IRs occuring in I I I X D(gi1 ) X D(gi ) =

I DI (gi ) X

Then H(K) =
I

aI

I I X K

aI = expansion coecients
Roland Winkler, NIU, Argonne, and NCTU 2011

Proof: Invariance of invariant expansion

Consider most general expansion H(K) =
IJ IJ I J b X K

Transformations gi G :
I I X D(gi1 ) X D(gi ) = J K

I DI (gi ) X

DJ (gi )

J K

Use invariance condition

IJ I J b X K =

(must hold gi G !)
1 h

IJ b

DI (gi ) DJ (gi )
i IJ

I J X K

= IJ

II b aI

I X

I K

qed
Roland Winkler, NIU, Argonne, and NCTU 2011

Irreducible Tensor Operators

Construction of irreducible tensor operators K = K(k, S, E , B, ) Components ki , Si , Ei , Bi , ij transform according to some IRs I of G . elementary irreducible tensor operators KI Construct higher-order irreducible tensor operators with CGC:
K K =

I J K

I J K K

IJ If we have multiplicities aK >1 we get dierent tensor operators for each value

Irreducible tensor operators KI are universally valid for any matrix Hamiltonian transforming according to G Yet: if for a particular matrix Hamiltonian H(K) with basis functions transforming according to an IR I does not appear in , then the tensor operators KI may not appear in H(K).

Roland Winkler, NIU, Argonne, and NCTU 2011

Basis Matrices
In general, the basis functions { (r) : = 1, . . . , n} include several irreducible representations J Decompose H(K) into nJ nJ blocks HJJ (K), such that
rows transform according to IR J (dimension nJ ) columns transform according to IR J (dimension nJ )

c v v
k

Hcc Hcv Hcv Hvv Hvv H(K) = Hcv Hcv Hvv Hv v

I Choose basis matrices X transforming according to the IRs I in J J I X

I DJ (gi1 ) X DJ (gi ) =

I DI (gi ) X

More explicitly:

I (X ) =

I J J

I |J which reects the transformation rules for matrix elements J |K

For each block we get

HJJ (K) =
I

aI

I I X K
Roland Winkler, NIU, Argonne, and NCTU 2011

Example: Graphene
Electron states at K point: point group C3v Dirac cone: IR 3 of C3v We have 3 3 = 1 + 2 + 3
1 basis matrices X1 = 1; 2 X1 = y ;

strictly speaking D3h = C3v + inversion It must be 3 because this is the only 2D IR of C3v (with 3 = 3 )

3 3 X1 = z , X2 = x

Irreducible tensor operators K up to second order in k : 2 2 2 2 1 : kx + ky 3 : kx , ky ; ky kx , 2kx ky Hamiltonian

(here: basis functions |x and |y )

no tensor with 2

2 2 2 2 H(K) = a31 (kx z ky x ) + a11 (kx + ky )1 + a32 [(ky kx )z 2kx ky x ]

More common: basis functions |x iy and |x + iy

1 basis matrices X1 = 1; 2 X1 = z ; 3 3 X1 = x , X2 = y 2 2 2 2 H(K) = a31 (kx x + ky y ) + a11 (kx + ky )1 + a32 [(ky kx )x + 2kx ky y ]

additional constraints for H(K) from time reversal symmetry

[see Winkler and Z ulicke, PRB 82, 245313 (2010)]
Roland Winkler, NIU, Argonne, and NCTU 2011

Graphene: Basis Matrices (D3h )

Symmetrized matrices for the invariant expansion of the blocks H for the point group D3h . Block H55 Representations 5 5 = 1 + 2 + 6 7 7 = 1 + 2 + 5 9 9 = 1 + 2 + 3 + 4 Symmetrized matrices 1 : 1 2 : z no spin 6 : x , y 1 : 1 2 : z 5 : x , y 1 : 1 2 : z with spin 3 : x 4 : y 5 : 1, i z 6 : x , y
Roland Winkler, NIU, Argonne, and NCTU 2011

H77

H99

H79

7 9 = 5 + 6

Graphene: Irreducible Tensor Operators (D3h )

K (K) allowed by Terms printed in bold give rise to invariants in H55 1 time-reversal invariance. Notation: {A, B } 2 (AB + BA).

3 4 5

2 2 2 1; k2 x + ky ; {kx , 3ky kx }; kx Ex + ky Ey ; xx + yy ; ( yy xx )kx + 2 xy ky ; ( yy xx )Ex + 2 xy Ey ; sx Bx + sy By ; sz Bz ; (sx ky sy kx )E z ; sz (kx E y ky E x ); 2 2 {ky , 3kx ky }; Bz ; kx E y ky E x ; ( xx yy )ky + 2 xy ky ; ( xx + yy )Bz ; ( xx yy )E y + 2 xy E x ; sz ; sx By sy Bx ; (sx kx + sy ky )Ez ; sz ( xx + yy ); Bx kx + By ky ; Ex Bx + Ey By ; Ez Bz ; ( yy xx )Bx + 2 xy By ; sx kx + sy ky ; sx Ex + sy Ey ; sz Ez ; sx ( yy xx ) + 2sy xy Bx ky By kx ; Ez ; Ex By Ey Bx ; ( xx yy )By + 2 xy Bx ; ( xx + yy )Ez ; sx ky sy kx ; sx Ey sy Ex ; sy ( xx yy ) + 2sx xy Bx , By ; By ky Bx kx , Bx ky + By kx ; ky Ez , kx Ez ; Ey By Ex Bx , Ey Bx + Ex By ; ( xx + yy )(Bx , By ); ( xx yy )Bx + 2 xy By , ( yy xx )By + 2 xy Bx ; 2 xy Ez , ( xx yy )Ez ; sx , sy ; sy ky sx kx , sx ky + sy kx ; sy Bz , sx Bz ; sz By , sz Bx ; sy Ey sx Ex , sx Ey + sy Ex ; (sx , sy )( xx + yy ); sx ( xx yy ) 2sy xy , sy ( yy xx ) 2sxRoland xy Winkler, NIU, Argonne, and NCTU 2011

Graphene: Irreducible Tensor Operators (contd)

6 kx , ky ; {ky + kx , ky kx }, 2{kx , ky }; 2 2 2 {kx , k2 x + ky }, {ky , kx + ky }; Bz ky , Bz kx ; Ex , Ey ; ky Ey kx Ex , kx Ey + ky Ex ; E y Bz , E x Bz ; E z By , E z Bx ; yy xx , 2 xy ; ( xx + yy )(kx , ky ); ( xx yy )kx + 2 xy ky , ( yy xx )ky + 2 xy kx ; 2 xy Bz , ( xx yy )Bz ; ( xx yy )Ex + xy Ey , ( yy xx )Ey + xy Ex ; ( xx + yy )(Ex , Ey ); sz ky , sz kx ; sy By sx Bx , sx By + sy Bx ; sz E y , sz E x ; sy E z , sx E z ; sz (kx E y + ky E x ), sz (kx E x ky E y ); (sx ky + sy kx )E z , (sx kx sy ky )E z ; 2sz xy , sz ( xx yy );

Roland Winkler, NIU, Argonne, and NCTU 2011

Graphene: Full Hamiltonian

H(K) = a61 (kx x + ky y )
2 a12 (kx 2 ky ) 2 a62 [(ky 2 kx )x

Winkler and Z ulicke, PRB 82, 245313

Dirac term nonlinear + anisotropic + + 1+ + 2kx ky y ] corrections + a22 (kx Ey ky Ex )z orbital Rashba term +a21 Bz z orbital Zeeman term + a14 ( xx + yy )1 + a66 [( yy xx )x + 2 xy y ] strain-induced terms +a15 [( yy xx )kx + 2 xy ky ]1 isotropic velocity +a67 ( xx + yy )(kx x + ky y ) renormalization a68 {[( xx yy )kx + 2 xy ky ]x anisotropic velocity renormalization +[( yy xx )ky + 2 xy kx ]y } strain - orbital Zeeman +a23 ( xx + yy )Bz z strain - orbital Rashba +a24 [( xx yy )Ey + 2 xy Ex ]z + a21 sz z intrinsic SO coupling + a61 sz (Ey x Ex y ) + a62 Ez (sy x sx y ) Rashba SO coupling +a63 sz [(kx Ey + ky Ex )x + (kx Ex ky Ey )y ] +a64 Ez [(sx ky + sy kx )x + (sx kx sy ky )y ] +a11 (sx ky sy kx )Ez + a12 sz (kx Ey ky Ex ) + a26 ( xx + yy )sz z strain-mediated SO coupling
Roland Winkler, NIU, Argonne, and NCTU 2011

G U a, b , c , . . . i, j, k, . . . e h Ck hk N D(gi ) = {D(gi )} I , J, K , . . . N , , , . . . nI I (gi ) IJ aK H SI , (gi ) P IR

group subgroup group elements indeces labeling group elements unit element (= identity element) of a group order of a group (= number of group elements) classes of a group number of group elements in class Ck number of classes matrix representation for group element gi (irreducible) representation indeces labeling irreducible representations number of irreducible representations indeces labeling the elements of representation matrices D dimensionality of irreducible representation I character of representation matrix for group element gi multiplicity with which K is contained in I J Hilbert space invariant subspace (IR I ) we may have multiple irreducible invariant subspaces SI for one IR I unitary operator that realizes the symmetry element gi in the Hilbert space irreducible representation
Roland Winkler, NIU, Argonne, and NCTU 2011

Symbols