Professional Documents
Culture Documents
Group Theory
Group Theory
6, () = 6(g""19). (2.40)22 In this rather trivial way, W becomes a Lie trqansfounation group on the differentiable manifold M = The general definition of a Lie transformation group is such that to each pair (p,g'), pM (a differentiable manifold), g'© G (a Lie group), there is assgciated an element pgi© M (which may be denoted and mean pg’ or g’~! according to convenience), such that a) de map is Ad b) Bee pe or (ey 5 °) tp 9 4189° Play a) or gy (9, pl = (9499) pe These axioms are the result of the group axioms when M is G itself, as above. But they also hold when M is a coset 4, if HCG, we consider the points of M to be the sets pl-Hg or pk =gH. There is a standard argument to show that any two of these sets are either disjoint or they coincide, and that they partition G. They are called M'= H\G (left cosets) or MR= G/H (right cosets) respectively. Legt cosets map into each other under the right action of the group, i. e. pls Hg 2-8) yigigte Hlga')= pli Right cosets map under the ce action of the group p= gh 2——> gi“! (gH) =(g"~Ig]H= r Consider our example G-W and Hg=(9(%,0,0) } ey Since G(X, 4,2 ~ pul = glx,0,0) gld,y,z), (2.5a) we may partition W into cosets by Hy letting g(x,0,0) range over Hg (i. e.x over R}: 5 Wr2)=(91%0,0) ey 91042) = (9lxet2 ~7 XM Ieq, (2.56) The manifold M = H)\W is then isomorphic to R2 and its representa- tive elenents may be labelled by gl0,y,z). The space of left cosets Ne may be subject to transformations through the right action of “W glx',y",z") (R) c5ly2) — ee lyyzlalx'y', 24) = (91X00) } ey 910,4,2) g(x',y',2") = (91,0, 0) eg IXY t yt, ZH ZT AY) — (2.50) £G1%,0,0) yey 91", 0,0) 910, yry?, ez" 45x" y! 4x’ y) = “5 (yty', zz + 7 xty'+ x'y).23 2.4 Transitive and effective action, The action of W on the right on M = H,\W given by (2.5c) is transitive and effective. We recall these “concepts. A group @ acts transitively on a manifold M when for any_two points p,pEM there exists ag ©G which maps p on p, i. e./p * pg. In our example above, given two cosets _ oply,2} and, cha, 9 the g EW which does the job is any g(x’, y-y™, Z-. xt Gayl), xt ER. A differentiable manifold where a Lie“group acts transitively is a homogeneous space for the group. Every coset space of a Lie group is homogeneous for it, and in fact, every homogeneous space is a coset space for these groups (19, Chapter I, Theorem 3.2). The isotopy group (also called stabélity or Little group) of any fixed point pM is the set of elements h © 1, CG such that Ph = P+ In the preyious paragraph, setting ¥ ‘and 2 = z, the isotropy group of og (y,z) © Hg\W is To(y,z) = t9(X',0,x' WI ER, The isotorpy group, quite clearly, may depend on the coordinates of p. In fact, if G is transitive over M, the leu groups of any two points in M are isomorphic and conjugate: The action of Gon M is said to be ogee if no ee tion in G are the identity leaves all of the points in M fixed. That is 9 ts {e} and is the case in the example(2.5c),. It would not havePbeett the case had we chosen.H7= (9(0,0,21}, @ p to define the coset space H,\W= W/Hy with cosets cz(x,y). There the non-abel- ianity of W would have béen completely fost as Hz would be the isotropy group for every coset ¢z(x,y). It is straightforward to show that a group G acts effectively on a coset space H\G if and on- ly if H does not contain a normal subgroup N of G. (A nogmal sub - group Ngp. The physicists’ expertence with combagt groups, where the V are finite-dimensional, leads one to search for matrix representations which may be of infinite dimension, or for their generalizatigns as integral kernels. A finite-dimensional representation on V-R® is already provided by (2.1), although it is non-unitary. An appropiate Hilbert space may be constructed for functions of three, two or one variable on which we defined the action of W. Infinite-dimensional matrix representations may be obtained if we give a complete denumerable basis for the separable Hilbert space, tH ly) dng? and build the matrix o(g)= ||D.(g) || with gle ments Dani(g) "29(¥,¥ 14), the inner product being that of L”(R) and the group action givan by (2.7). Similarly, integral-kernel representations may be obtained through working with a Dirac-ortho- normal basis for the space, fxy(y)}, € where 4 is some index in- terval. There, the integral kernel5 are’o(g)= HOt) I with Dyrlgl= (xyoxyegl- The one-vafiable L2(R) space the is manifestly mapped onto26 itself by W, Its chgice is dictated oyer any other L?(a,6) by the fact that the group action (2.7) includes translations in the argu- ment Y, so that only spaces of periodic functions of period b-a may be contemplated, but the exponential factor e¢X'Y does not res- pect this periodicity over any finite interval. The Q-subgroup bas Let us start first with the integral-kernel representations afforded by the normalized Dirac eigenbasis of the ‘position’ opera- tor Q) , a) ei q xy) 1 tA en (2.10a) xyly) = ill? sta ay). (2.100) Then, the representation of W labelled by \ ©R in the @ -eigen- basis is a (9) rytz)) = (yr, Dag? iglxtay'sz')) (xq Xqrg! Rea ’ Pafetak gtgte xt dy xqly) xgr lyty lexplaalz'+y x'y!# x'y)] (2.11) 6 (q-q' + dy") exp <[Alz'+ + x'y')+ x'q]. The representation property holds: i da! OA9) 195) 0218) tap) = OS (9,99), (2.12) and 0*'2) (e)* 8(q-q') is the unit operator. The strong continuity of % (2.11), meaning the strong continuity of (2.7) is easily ascertained. The subgroup g(x',0,0) is represented by a ‘diagonal’ integral kernel (i. e. one with a factor of 6(q-q']). This represen- tation diagonalizes the subgroup generated by Q. A representation of the operators in the Lie algebra w and W through integral kernels may be obtained subjecting the integral kernel (2.11) to the limiting procedure in (2.8): Qe = 43,0 OND fale ys geg 29 Slang), (2.138) PE = day OND) (glx! y's geg 8" laa"), (2.136)27 tee 74d, on ot. = 6lq-q'). — (2.13c) These integral kernels represent the differential operators of the Sghr&dinger realization on the space of differentiable functions in L2(R], a space dense in the latter. 2.9 The P -subgroup ba A second generalized Dirac basis of 12(R) is provided by the generalized eigenfuctions of [P'"!in (2.9c): PMP ty) =p xh Wy we AER, (2.14a) M ty) = (2n)1/? eAPy, (2.14) The representation thus obtained is 2 lalx'sy",21) = (25 Rag) = 6 (p-p-ax') expéla(z'+ t xty') + pty") (2.15) i. sus kl a f. wf dq Wy XQ! Dag’ '9) Xr Xp It satisfies the representation property (2.12), and diagonalizes the subgroup g(0,y,0| pgenerated by P. The last ligegjn Eq. (2.15) shows that the D (g) are equivalent to the DO ™'(g) in (2.1 since they are bSGically the double Fourier tran{forms of the 0’ in (2.11), as (xt, G4) = (anal) 7!/2 e€P@/* is the Fourier trans- form kernel. The overlap coefficients between the two eigenbases (2.10) and (2.14) yield the tranformation kernel between the two re- presentations (2.11) and (2.15). We shall return to this point below when we comment on the unitarity and completeness of the represen- tation set. Meanwhile, let us produce further representations equive lent to the above ones. The Lie algebra representation obtained from (2.15) may be written as 10) aq (2) Ol te) 1) alPl be 4&5" (p-p'), Ppl po (p-p'), Tops which, again, is only the Fourier transform of the former. = d6(p-p') (2.16)28 2.10 The quantum free-fall nansubgroup ba Further generalized bases of wR), unrelated to subgroups of W may be produced as eigenbases of self-adjoint operators in the envel oping algebra of w. Quantum Mechanics has a large supply of such operators. Let us start with the free-fall Schrédinger Hamiltonian and its Dirac-orthonormal eigenfunctions: Hoarty) 04 pelt, gly aX(y) = ath (y) , a ER (2.178) 1/2) 1/3 1/3 ty) = (2hal ai(t2a)’!” (y a/2)), (2.176) where Ai(z) is the Airy function of the first kind (14, subsect. 9.5.3). Then, a calculation aided by the Fourier transform shows that on (g(x',y',z") iF amet F/2 jay x (2.18) x exp Drzt+ Fx! (ata!) + For - ped. Rather tediously, we can verify that the analogue of the representa- tion property (2.12) holds. Using Lim, 2/4 ane)? exp d(%/ta"e) = Jalsis), (2.19) e>0 we can show that (2.18) is indeed 6{a-a') for g=e , as well as some further properties which will be seen later. (If the powers of A throughout produce any distress, the reader may check that "units'' of A are properly given if the arguments of transcendental functions are to be dimensionless, while wavefunctions have units of (@enght)-1/2 and integral kernels of (Lenght)~!. The task of finding the Lie algebra of integral kernels now takes us to find through (2.19) and its derivatives with respect to 4, as in (2.13), the integral kernel representations: oe) - 7 a8" (ana!) +a6(a-a'), (2.20a) ple . 943 6! lew'l, (2.206) on a (2.20¢) aa29 Note that, indeed, pr att a = a S{a-a') (2.204) is a diagonal numbex integral kernel, as it ought to be from (2.17a). 2.11 The quantum harmonic oscillator basis. A most convenient denumerable orthonormal basis is provided by the eigenfunctions of the quantum harmonic oscillator Hamiltonian 2 2 Bvt = Fees ay iy = aloe D ety, (2.210) n = n 7. 2 eg = (anne ar |ag 2-1/2 ell? taf /2y) (2.216) In this basis we must calculate OM tgtxtyy?y2")) | dy ty) *e (yey!) exp lz" ty! yx I, (2.22a) mise may be done through multiplying (2.19) by glnen")/2 irq) 1/2, i | summing over n and n' so as to use the known generating Brune ions for the Hermite polynomials (16, Eq. 22.9.17), integrating, and finally using the generating function (3, Eq. (2.42) explabtac-bd) = oy > n=0 n=0 in order to separate powers of 4 and zt. Setting a= /2s, b=v2t (n-n') nyt’ RM pay (ed) a’ and c=(|A|/2) a (-y'+ix') = d* we obtain the result DUN gtx yty2t))= exp aléze Epa’ y' 2) nttynty'? x pai! payreaeryy Len Lite? + yh, (2,22b)30 valid for n >n' , Similarly, for a= v2, b= YZA and c= ([a}/2)!/? (ye exl= d* we obtain prlh) DE (gtx uty z= exp alee + Fret ey ty intinrty/? x (n'-n) x fal/an/? cyrece nytt Le) Laer tey! 2), (2.22) valid for n o O,y>o0 4, * { yy { (2,24¢) 0,y<0 -y, 9 <0. The representation kernels are now 2 x 2 matrices (rows and cok ums labelled by o = ) with integral kernel elements. We must consi- der the four pieces (o =t,0'= +)for y’ 0 and y’ <0 separately. In constructing the kernels through inner products of (2.24b) we obtain integral representations for the confluent hypergeometric functions M (a,c,z) and Ula,c, z) (16, Eqs. 13.2.1 and 13.2.5): AU) ae : : Donor (2 Y'az'N) =t2a) '[ dy u(y) uy, 5, (yey) x x exp [éatzttx! y"/2 +x"y)l = (2n)Joxp Laalz'+ xty!/2] ee (xy), (2.25a) KA Unt, yh >a) = 01/2 - dphr) gitlor-Pl/d x U[T/2-€ o/A, 144 (p'-pi/a, -daxty"). (2.25b) Kae (gt >a) = = eMlPrPDIN Ax Y KH, xt, yt >), (2.25c) Ke (x1, y'< 9) = (xt, 9 > 0), (2.25d) per Pp, “8 Kesler, gt Daye = oMlPO'M% niajte gota) TN/2 - & e/a x x rll + élp!-p) AN “1 yilo'-p Ih x M(1/2- fr, 144 (0'-—) fA, -Ax'y'), (2.25e)tn maxtyt pnt K toyt 0), (2.25f) tt Rp (2.259) The 2 x 2 matrix is lower-triangular for y' > 0, and upper-trian- gular for y! <0 the first and second rows and columns are labelled by o = + and o= respectively). it should be interes- ting to verify that the group representation composition and inden- y properties hold for the integral kernel (2.25). Differentia- ting with respect to the group parameters should yield the integral x(D) pal) A(D) 2 ' one Fon a! a! and Toa, dia! * 26,41 5le-e ). These will constitute yet another form for the Heisenberg-Weyl algebra such that 1/2(QP+ PQ) js the number operator 6 _,pé(p-o'). Any self adjoint operator in L“(R) can be seen thus toRave associated to it an eigenbasis which determines a corresponding representation. The first two cases we gave had Q and P for number operators. These determine that the subgroups g(x,0,0) and g(0,y,0), respectively, be represented by purely diagonal, jptegral kernels. (¢.f. Eqs. (2.11) and (2.15), these are 4(q-q') e** 4 and é(p-p')e“P4 , and hence the corresponding representations are said to be reduced or classified according to a subgroup. The other three cases seen involve nonsubgroup bases, as the corresponding number operator was given by two Schr8dinger Hamil to- nians (2 and h cases) or by the dilatation generator ]J, In these cases, the representation is given by matrices/kernels which are not diagonal in any subgroup, as suggested by the absence of Dirac 6's in (2.18) and (2.25). They are truly integral kernel representations of the group. kernels for Q 3_Unitary irreducible representations. We now define and verify the unitarity and irreducibility of the representations obtained in (2.11) (Q), (2.14)(P), (2.18) (2) ,(2.22) (y) and (2.25) (0). In each of these cases we chose a basis for L4(R) through a self-adjoint operator and the bases are therefore orthonormal. The Heisenberg-Weyl alggbra generators are self-adjoint in their Schr&dinger realization in L*(R), and their exponentiated group operators representing (2.1) unétaty. A unitary operator (DJigl6=6, such that (J (a1 6, (Ig) h}= (4,4) in an orthonormal basis ts représented by a unitary matrix or integral kernel: r a -1))* 2 =( 0 2.26 uur (9) [ ry 19 Ny ( ) where uw ER for cases 9, P and £, uS{0,1,2,...} for case h and us (9,0), o=t, p ER for case D. For A real, this is indeed the case, manifestly, for cases P, Q and £. In the latter care has to33 be taken to define the phase of the parameters in g!(x!,y'z!] = gl-x',-y',~z'), The matrix for the h -case also satisfies (2.26) with both (2,24b) and (2,24c) pequired, The last case, D, can also be seen from (2.25) using the pairs (6,d), (c,d) and ({,g) which correspond under the adjunction of the 2x 2 matrix. In all of these matrices no block-decomposition occurs and the rgpresentat ions are in fact irreducible. This is to be expected, as L“(R) itself is irreducible under (2.7). 2.14 Equivalence of representations. | By equivalence of two representations 2!) (a) ana 0!) (g} of 3 group G, associated to given number operators H(/) and H + we mean that there exist (fixed) invertible transformations C of L“(R] such that the two representations are mapped into each other as dl4lig) 2c 1g) oo? , vge G, (2.278) f 4 cat : (2.27) al igi=t otal t,o ig volZligit, where the set of values which wu can take is the generalized spec trum of H!") in L2(R),and p in that of H!2) in L2(R): wl) gy ot (2) 6 p(t) gl o VoL Gop {ali (227) eH!) ul2) © @ then C in (2.27c) is a (possibly exterior) automorphism of the algebra, mapping one subgroup basis into another subgroup basis. An example is provided by the two subgroup represen- tations denoted by Q and P , as we shall see below. If the H) are self-adjoint and have the same spectrum, C will be a unitary transformation of L2(R), i.e. C! = Ct. this is the for the transformation connecting the Q ang P ‘representations (R} the Fourier transform- which in fact maps L unitarily onto it- self. There C*(q,p)= (2n)7!/2 explipq/a} is the transformation kernel. The Q- and L-representations werg defined through eigen bases of the operators Q in (2.10a) and H* in (2.17a), and have the same spectrum. The transformation kernel C is then 2 ae c = . wa “Karty! i. e. the Aipy transform (14, Sect. 8.5.3), which as expected is ary in L4(R). This is a point transformation, where to the op- erator representing 9 we niay add any function of its canonically conjugate P, as 9>Q+ 6(P], while leaving P invariant. Thus pir’? a taal, (2.28)35 ars JH) = afe pe Lf dg’ 2tHla,pl-vla/? q. The angle variable is (26, p. 292) -with a minus sign- w (H,T) = - (aH/aJ)T. (2.316) One can depart slightly from the usual notation: Since q, are the libration endpoints where V(q,) = H (while otherwise ~ H (q,p). The first observation is that even in classical mechanics, this transformation is generally not bijec~ tive, i. e. the classical phase space motion may be subject to a group of transformations which leave the action-angle variables in~ variant, as the projection of the Riemann surface for a multivalued function over the complex plane, under exchange of sheets. This ambiguity group A is therefore an object which appears already in classical mechanics. Quantizing the system now means replacing q and p for self-adjoint operators and in soe appropiate representation, say, the Schrédinger ee on PPR while q and p are replaced by operators @Q and Pp ... in what space? If we have constra g potentials giving rise to a closed classi- cal orbit in phase space and were to propose simply L“(R), the spec- trum of_{) would be lower-bound and discrete. In what space may we define Q such that its spectrum be R, as it is for) ? This is of interest since we desire that the quantum canonical transforma- tion be unitary. As we have also the ambiguity group, we may build a space L2(R,A) = De, L2 R} consisting of the ordinary L2(R) summed with itself ofce for Qvery element g of the ambiguity group, 0 as to obtain discrete and perhaps infinite matrices with integral kernel elements. This is the space where Q may act so as to have R for its,spectrum, and here the canonical transformation to the original L2(R) may be bijective and unitary. The space LP(B,A) is then classified in a way where linear com- binations of the L°(R) spaces are taken so as to build the unitary irreducible repres@ntations of the ambiguity group A; in this way one defines an ambiguity Spin. Some lower-bound discrete spectra allow for a dihedral ambiguity group of rotations by multiples of the period,and inversions. The corresponding ambiguity spin is a pair of numbers: a continuous one x © [0,1] and a sign oc. The space L2 is thus written as cits 2s 12. (Ri. The operators Q found of Moshinsky and Seligman (33) were such that they are a sum of a Hamiltonian operator with a lower bound and discrete spectrum, témes the sign o plus the representa~ tion index «. In all, thus, the spectrum of Q is R. Transform kernels are found which represent the above quantum canonical trans- formation. 2.18 Quantum mechanics on a continuous compact space There are other approaches to the problem of making the Hamil- tonian of a contraining quantum potential to fit into a Heisenberg- Weyl algebra (3, Refs. 138-154). One of them follows a suggestion by Weyl (36) introducing a mixed group W™ (3, Sect. VI): It is a proper subgroup of W with the composition rule (2.2), and defined through37 w = {9 (mzl © W] x= nM, ny a 2 (2.33) y =y mod L3 z= z mod L/2M }. It is a Lie group of transformations on the circle as in (2.7), where X is a discrete group coordinate for the subgroup elements Ho while the subgroup manifold of Hp is a circle. For the latter, the infinitessimal generator |P fay has, discrete eigenvalues P*MPo, Mt, © 2, pg = 2u/L. The former group Hy does not have an infinitesBimal generator. It has a gcnite genePator instead: multi- plication by e(¥ . The eigenvalues of the Hy generator are similar- ly quantized to A=, ron, © Z, Ag= 4nM/L = 2 Pg M. In this model for Quantum mechanics on compact spaces one way work with a properly defined enveloping algebra, implement quantization procedures, clas- sical limits and canonical transformations. Furthermore, as will be brought out in Chapter 3, in connection with the 2+] Lorentz group, one can define a nonlocal inner product on the circle so that the operator -4d/dy have a lower-bound spectrum (37). On the other hand, one does not have a self-adjoint 'posi:ion’ operator, and has in effect quantized on the Lorentz group level. 2.19 Left-_and right-invariant Haar measure. Having gained familiarity with the Heisenberg-Wey! algebra and group, we may state what a noncompact group is, and what its repre- sentations are like (20, Sect. 2.3). Consider a Lie group G with a finite number D of continuous parameters (Xp}P.;, XR CR, and functions §(X) over G. A pos- itive Radon measure is a positive linear form ald) Sf ssirsior f w(KAX4IK) > 0 (2.34) on the space C* (G) of continuous nonnegative functions { on G with support on a finite-radius sphere. The Left (resp. right) invariant Haar measure is a positive Radon measure which is left (resp. right) invariant under the group action (2.4): vig) = ult shy) (resp. vRi6) = u®ig8,1 1, for all g! €G. A theorem may be prove (38, Sect. IV-15): E¥ery Lie group has a unéque left (and right) invariant Haar measure, up to mu tiplication by positive constants. If we fix the multiplicative constant, we may define the vol- ume of the group G, vol G, to be the limit of the sequence of values of (2.34) when 4 is taken as a sequence of characteristic functions over any nested growing sequence of group subsets. This is (2.34) for §(g)=1 when it exists. If the volume of G is a finite number, the group is said to be compact. If it is not finite, the38 group is said to he noncompact. Let u4(§) be a given left Haar measure. We can produce a new left Haar measure through acting on the Aight of the argument with a fixed group element: ud (§) = u!(g® ). But since right and left actions commute, t.e. 9 (6 yh (gh? gig' Fog (6ts)® igh, it follows that we (gh Je ube ele wlla® ye uk (41, so ub is Go 3" Io 3" Hoe eee Io also a left-invariant measure. By the uniqueness theorem, we con- clude that “4 (6) = Blgglub(g), i.e. ut may at most differ by a go constant A(g,] from ut. 2.20 Unimodular group: A modufar function over G_ is a positive function A:_G—> r such that 4(gg’) = A(g)A(g'). This implies A(e)=1 and alg ')=1/A(g). The constantsA(g,) seen above are modular functions, as can be veri- fied acting with two elements from the right on the left-invariant measure. If A(g,)=1 for all gg&G, then the left-invariant Haar measure is also Invariant under right action, so that the left- and right-invariant measures are the same. Such groups are called uni- modular. The difference between right and left Haar measures never ap- pears in compact group theory: Every compact group is unimodular. Proof: If G is compact, §(X)=1 is in C\(G) and we may normalize uy through asking for ufl)=1. The modular function of the group is then A(g)= A(g)u(T)=u(To)=ulT)=1. Abelian groups -cofpact or noncompact- are unimodular, since their left and right actions are the same. Noncompact groups may be non-unimodular, an example of this is the two-parameter solvable group of linear transformations x ——> x! = a,x#a@, seen in (39, p- 316). Noncompact groups which ate unimodular are the followin (a) all abelian groups, (b) all semisimple groups, (c) all connected nilpotent groups -as the Heisenberg-Wey] group, (d) Lie groups for which the range of values of modular functions is compact. (e) di- rect products of unimodular groups. 2.21 The Haar measure and weight functions. When the composition functions for the parameters of a group, are known, it is not difficult to build a weight function for a left- invariant Haar measure. Right-invariant Haar measyres are very simi lar and will be given below. We require uh({)= u-(f") in (2.34). This implies 5g dué(g) $(g7'g)= i 0 0 LL u 65, du'(g.g)gigl= uilg). — (2.35a) Since g G =G, we need dut(g g)=dué(g) for all g €G. If the coordi- nates of g are X and those’of g’=g,g are X' then the weight func- tion and parameter volume element must satisfy39 aX" (g59(X)) whK@X = whiX X= al (Xd (—ae—) PK (2.350) where J(2-/3*)= J(gg,X)_is the transformation Jacobian. If the group identity is at X=Q so that g(Q)=e, and w (Q) ig a fixed num- ber, (2.35b) gives us the appropiate weight function w (X') at g! 8X" (g,9(X)) 7 X" (9,9 (X ] (2-36) ahi) = 0) [J ae] k-0 This is the left-invariant weight function for the parameter volume element at g'=g,, where the parameters are X' and which is invariant under left group translation. A similar argument for right-invariant integration, measures and Jacobian leads to R R aX" (g(X)g) yy = Ro [J 1 | (2.368) X-0 as the appropiate right-invariant weight function at g'=g"! where the parameters are X'. . The measure_(2.35) is also invariant under the inversion involu- tion g! ——* g' of the group manifold, (20, p. 69, Prop. 3). The Heisenberg-Wey! group W, being connected and nilponent, is unimodular. Hence, right and left invariant measures are the same. The Jacobian in (2.36) may be computed from the group composi functions (2.2) for g'(X')=gg(Xo)g(X) and g"(X") =g(X)9q(Xg) + Xt XY HUot yy t'=Z_ + zH(Yx - Xpy)/2 (2.37a) xt ny = Maze - - : vty z-2,- (ye,- wf? . 42-970) The Jacobians have I's on the diagonal and are triangular, therefore the determinants are unity. Normalizing w(Q))= 1, the invariant Haar measure for W is du(X) = dx dy dz. (2.38) The Helsenberg-Wey! group Is very close to the abelian R° space. The ‘twist’ which makes the third parameter compose in a non-abelian way, is relatively minor. 2.22 Completeness of a set of representations. We would like to insist on the distinction between a representa~ tion and its subgroup (or nonsubgroup) reduction. Different examples of the latter have been given in the first part of this Chapter, and40 refer to the row-column classification through the choice of basis for the homogeneous space of the group. Since we saw they are all unitarily equivalent, we can refer, to any one fixed classification through a given basis in writing D\y1(g). The (in general, collec- tive) indices % and 'are eigenvalues of a (maximal set of mutually commuting) operator(s) in the right- and left-action enveloping algebra of the group, and self-adjoint under the inner proguct defi- ned by the Haar measure over G (i.e. in a Hilbert space L“(G)). The eigenvalues may be discrete, continuous or mixed,and their range is assumed to resolve degeneracy completely. These representations compose as (2.23) in the discrete case, and as (2.12) in the contin- uous one. We shall write vee for Dees or Sd... The represen- tation index 2 (which in‘general if a collective index) is usually given as eigenvalue(s) of a (maximal set of algebraically independent) right and left invariant operator(s) in the centre of the enveloping algebra, self-adjoint in L*(G). The range of the (generally .col lgc~ tive) representation index A(A={A,,Ag,+++,Ay}) is a subset G of R. In benign cases it is a measurable space, and a Planchere£ measure and weight function do(A)=v(A)dd exists. For W , the centre of the enveloping algebras (2.9a) and (2.9b) ig WIRD e. is I Casimir operator. The eigenvalue of tion label AER. The difficulty in treating the representations of noncompact groups vis-a-vis the same task for compact groups is the same as that of Fourier integral transforms over Fourier series. The space being noncompact in the former case will allow continuous spectra for cer- tain operators whose eigenvectors are normalizable only in the Dirac sense. Casimir operators may also have mixed (continuous and dis- crete) spectra. In the W case, the difficulties will be handled through Fourier transform techniques. In general one has to invoke rigged Hilbert spaces (20, Chapter 14) and pay close attention to functional analysis arguments (40, 41). The problem for compact group are less, as there all unitary representations are finite di- mensional and representations come in discrete series. 4a/oz = - T!4), so this generator takes the place of the (3/3z provides the representa- 2.23 The orthogonality and completeness relations for the unitar reducible representation matrix elements. For unimodular groups (with some mild conditions), we can gener- ajly assert that the unitary irreducible representation matrices Di»: (9) are a generalized orthonormal set of functions over 6, under the inner product of L“(G) given by the invariant Haar integral. We denote the latter through S = fgdulgl--, in order to include the finite and compact cases in our formulae. We may write a at = » ey (ave Peigrede “Se (Phys tal” nla aC No (2.39) G yny where 6 means a collective Kronecker delta in discrete indices, (which may’ bs generalized to a Dirac 6's over a continuous pair), 6,(A,A'] should play the role of the Kronecker or Dirac 6's for Aja! EG, (G being endowed with a Plancherel measure and weight func- tion do(A)= v(A)d (A). Equation (2.39) is not hard to prove through integration by parts with the operators which determine the represen- tation and row/column labels. Most interesting is the statement that the ot (a) are, moreover, complete in the L2(G) Hilbert space with inner product (+,*)4. This means that we may define an inner pro- duct in G through an integration as Sg dp(X).., which may contain x a sum, if the Plancherel measure is a point measure in some domain We have (Dlg,), D (apg = Ss Tr (D*(a,F Dag) (9)+59) » (2,40a) where the trace of a product of matrices is Te NW = Do hall» (2.400) For integral kernels an integral over % and, 4' is used. Equation (2.40a) defines an inner product between matrix or integral kernel functions over 4 © G. The completion of this space of func- tions with finite norm defines a Hilbert space L(G). The 86(9 499) is a Kronecker or Dirac delta for discrete or continuous groups: Sgl9y25q)=0 for g, # 92, and under sum or integration, it satisfies Ss élgldgla,g") = g{9") (2.41a) for any continuous function (g). Similarly, 6,(A,A"] is such that G S Fe) 60,09 = Fa) (2.41b) ASG c for continuous matrices or integral kernel functions F of 4 ©G. The defining properties (2.41) imply a relation between weight functions42 and 6's, as wlGIXI)6gl9(KI,9IX')) = XX, (2.42a) va) 6, (A,A") = Sfa-a'). (2.42b) G In the case of compact groups the Plancherel measure is a point measure and S,-*"= ¥ dim(x)/vol G+ where dim(4) is the dimension of the A unitSr irreducible representation, and vol G is the volume of the group. The corresponding &, is thus a Kronecker 6&,(,A') = 8,1 VO! G/dim(r). G G For compact groups, the Peter-Weyl theorem (20, Sect. 7.2, 42) supports the proof of (2.39)-(2.40). For finite groups this can be found in (39, Eqs. (3.143), completeness appears only as (3.178)). The general case of unimodular noncompact groups appears and is re- ferenced to in Barut and Raczka's book (20, Chapter 14). 4 The Heisenberg-Weyl case. We return to our example W and choose for simplicity the Q -eigenbasis where D> (g(x,y,z)) is given by (2.11). The Haar mea- sure is given by (2.9438) with unit weight function and hence 6g in (2.42a) is an ordinary Dirac 6 in x, y and z. We do not yet know the unitary irreducible representation space WW, but since 4 €R, we want to determine if R is or acts as the full representation space. To this end we have avail to (2.39). If R were not WW, we would not get a Dirac 6 in this variable. We perform fof af (5 q-q'+rylexp < [Al z+xy/2)+xql }* x x{6(qt-q ta%y) exp < Ed" (ztxy/2) #xq a pr eq" ,Par a(jada' | big-a'/a sears f dz exp i[(a-r')z] x x [ dx exp [(-< x} ((a'~a} (q"- a") /2a4(q-q") 1 = (2.43) = 4n2|a]7"5(a-a") 8(q-a")6(q'-q"") = 85 (aer"16lara"lslat-a™ | - So we do, indeed, obtain a Dirac 6. From (2.42b), the Plancherel measure for W is43 dotal= ([al/4n2] dh, AER. (2.44) Completeness is verified as ae a et (Dlay), Dag) y = (4x?) [nda fda fda" 0% 44) 0%, glap) R Re = 8(x7-Xp)8(Yy-Yo) 61Z4-Zp) = Syl9y499)- (2.45) The orthogonality and completeness relations (2.43) and (2.45) are valid whatever subgroup or nonsubgroup row/column classification we choose. Hence, orthogonality and completeness relations similar to these follow not only for (2.11), but for (2.15), (2.18), (2.22) and (2.25). 2.25 Harmo Having the complete and orthonormal generalized basis Din (g) over G allows us to perform hatmonic analysis over the manifold G, expressing any function within a wide class £(g) over G , through a series or integral over these functions, as ie S © tor i, (2.46a) r1€6 an! c¢ anal The 'linear combination coefficients! Ken ttl cn be obtained through Performing the G -inner product of (2.46a) with po $ $2 : 3 with > 2 and using (2.39) so as to obtain i yin (g), exchanging 612 § 69) Ota. (2.46) ne 6 vue We can think of F(A) as a matrix- or integral-kernel-valued function on G and write (2.46) as sla) = S TrtDiglF 1 = Dio .F I. , (2.47) 1€G 9 : ‘ Fil = S (9) DMa) =D, Fig, (2.476) 9644 Finally, the Parseval relation FeHlg= S 6lal'Atal- oe (2,48) . -. oF a = SY by Ry Od = & TRtFouko eA 1B! 1G 6, holds, telling us, that the harmonic transform (2.47) is unitary be- tween L(G) and L7(6). The subject of orthogonal and complete sets of functions over G and 6 spaces extends to all coset spaces H\G or G/H together _ with right- or left-invariant measures on these, and a corresponding reduction in the row- and column-indices (43). It also extends to other, more general equivalence sets on the group called bilateral classes (44). Rather than delve on the general theory, we shall give some results for the 2+1 Lorentz group in the next Chapter.CHAPTER 3: A FURTHER EXAMPLE. In this Chapter we shall present in some detail the case of the semisimple noncompact Lie group of lowest dimension. In the former two Chapters we dealt with the Heisenberg-Wey! group, which was ‘as abelian as possible’: Only one of the three commutators in the alge- bra was nonzero. Now we work on the ‘least abelian’ of three-para- meter groups: The 21 Lorentz group $0(2,1) and its covering group SL R) . The 2#1 Lorentz group has many resemblances as well as definite differences with the three-dimensional rotation group $0(3), which is probably most familiar for physicists who have worked with quantum mechanical systems such as atoms and nuclei. We shall find all irre- ducible unitary representations of the algebra and covering group. We then realize both the algebra and the group on a coset space rela~ ted to the Iwasawa decomposition: the circle, and functions and dif- ferential cperators thereupon. As a final development, we find the in general nonlocal measure defining Hilbert spaces of functions on the circle, for the various representation series. The S0(2,1) group and algebra. Consider a three-dingns ignal gpace,? with metric (4,-,-), where the distance is d= xp - x; - x5. This is invariant under the '2+1' Lorentz transformations. These transformations are repre- sented by (pseudo-) rotations around each of the three axes, co. 0 ee exp liv Jf lef 0 chy shy} @ Jo =| 0 0 -<], (3.12) 0 shy chy @ -t 8. chy 0 Shy © @~< exp ix Jp l={ 0 1 0 ~ 3 =( 0 0 of, (3.16) shy 0 chy -i 0 os -bind 0 ¢ ai ¢ exp (4b Jy 14 5énb coad 0 J -[¢ @ ¢) BW 0 e _ ee We have expressed the rotation around the k*" axis as explé a J?) in order to produce a 3x3 representation of the associated Lie algebra. The distance is also invariant under the inversions Xp -X9 and X1 & -X,3 these compound (in semidirect product) with (3.1). We shall not consider them in what follows, as we are interested primarily in46 reaching every element with a Lie generator, and they lie in group components not connected to the identity. From (3.1) we define the 3x3 faithful group representation of the 2+1 Lorentz group as 6° (a, 8,1) = exp (éaJ9) exp (489) exp lay9) , (3.2a) a =a mod 2x, y =ymod 21, BER, (3.2b) thereby parametrizing it through the Euler angles a,8,y adapted for the nonpositive metric. The matrices have unit determinant while the generatorsJ/ are all traceless. The abstract Lorentz group composition law is obtained from the composition law of the matrix representation (3.2) in the same way as done for the Heisenberg-Weyl group in (2.1)-(2.2). The matrices (3.2) are '2+1' pseudo-orthogonal, e. orthogonal with respect to the Lorentz metric L as “10 0 @Le™ = L : L-[ 0-1 0 be (3.3a) We call this group, Special (i.e. of unit determinant) Orthogonal group in 2+1 real dimensions: S0(2,1). The fact that no isochorous space-time inversions are included is usually denoted through a zero subscript, but we shall omit it understanding that we deal with the connected Lie group only. Correspondingly, the algebra representation matrices are pseudo- skew-symmetric: i PEt. hel, 2, 0, (3.3) and satisfy the $0(2,1) algebra give by (I,,Jg]- “yy (Jg,Jgl= 4 J}, (JpJyle 4 Jy. (3.4) Notice the minus sign in the first commutator: The ordinary three- dimensional rotation group algebra so(3) has a plus sign in its stead. one cannot bring so(3) to (3.4), no matter how we redefine ge- nerator signs, unless, of course, we introduce i's (J+J, , Jy >J)i but as we work with real groups, its representation structure fould be readically changed. 3.2 The SU(1,1) algebra and group. We are familiar with Pauli matrices, so we may look for another representation of the Lie algebra so(2,1) in (3.4) through 2x2 matri- ces. Indeed, we can associate the representation47 fa 1 chy/2 —éshy/2 ojo tL « iy J) = Fy > sy Z ( :) explains) = (tie chp/2 3.5a) 1 chy/2 bhy/2 itt vy (My 2 Jg> bp i> exploxdo) = (coe ee), (3.50) ae ue Jy * JG of, 1) expléosg)= (; ait). The matrices to the right define a matrix group (3.50) GH (8,8, Y= expl G°la.8yy) An alternative parametrization of the SU(1,1) group which is sometimes preferable to the Euler angles in (3.6) is found through demanding (3.7a) for any 2x2 complex matrix. This leads to noe GYin,0) = (: °), (3.8a) In|?-Jo|? = 1, nO EC. (3.8)48 In this parametrization, the SU(1,J) manifold is a twordimensional complex hyperboloid. The ralation between. the Euler angle parame- ters in (3.6) and the parameters in (3.8) favoured by Bargmann (23, Sect. 3b) is easily found (reinterpreted by Sally: 45, p. 3) as etlo-v/2 snaye, (3.8) oe 3.3 The group SL(2,R). Another parametrization of SU(1,1) is obtained through a sim- ilarity transformation of (3.8) as lab 1-1 1 -1\-1 GS(a,6, 0) (: ). Z ) sti 4 ) - Re n- ReO@ -Imn+ime “\im n+ Imo Re n.tRe Of, (3.9a) ad-be=1, a,b,0,d ER. (3.9) The corresponding 2x2 algebra representation is ee : - che/2 sh¥/2 J #(j i) % exp (ex) = (re chv/2 ), (3.10a) erx/2 9 BF ( 4) exp laxlg) = (, otf, (3.108) So og, ( cost/2 sinb/2 Wore € at exp (aap) “(ait cosd/2 (3.10¢) In this form, two further group isomorphisms are displayed: Since GS{a,b,c) is the most general unimodular (unit determinant) 2x2 real matrix, the group represented here is clearly SL(2,R). The 2x2 unimodular real matrices (3.10) have the further pro- perty: eee Gan) where S is the symplectic metric matrix. The relation (3,11) defi- nes the two-dimensional real symplectic group Sp(2,R). This group and its 2N -dimensional versions Sp(2N,R) are important as dynamical groups -rather, algebras- for the N dimensional harmonic oscillator.49 3.4 The covering group SL(2,R). 7 1:1 We saw $0(2,1) “© su(1,1) ue SL(2,R) “x Sp(2,R). What are their connectivity properties? A convenient handle is provided by the complex hyperboloid in the n-@ plane of SU(1,1). Although we need four digensiong, we can Igok at the surface described by (Ren) 2+(1mn) @-(Reo) 4=1 + (1mo)2. For fixed Im 0=0, the remaining three parameters are constrained to a one-sheeted equilateral hyper- boloid with a circular waist in the (Re n, Imn) plane. The group unit g“(1,0) is on that waist. As we let Im 0 range over R, the outside! of the hyperboloid fills. The group manifold of g-(n,9) is thus the full 0 complex pane, times the n complex punctured by a round hole of radius 1+#|0|* . We may surround that hole any number of times describing a path which may not be continuously deformed to a vanishing loop. The argument of n increases, but whether or not an increase by 2m is considered to bring us back to the starting point depends on the number of distinct Riemann sheets we provide for g4(n,o) in n. The simply connected universal covering group of SU(1,1)* * SL(2,R) * Sp(2,R), denoted by SL(2,R), is that where the argument of n may take any real value without repeating any group element. A convenient parametrization for SL(2,R) is provided by (3.5)- (3.8) with new parameters as : ee a ame 2_,)-1/2 3.12a GY (yo) = (fy|*1) ee ee yec, lyf -u2 su. The maximum of -u2*y occurs at y/=F1/2 and has a value of 1/4; so recalling (3.185) for q > 1/4, tha’fepre- sentation k= 1/2 +4p,o # 0 contains all values of u-m+e,m€Z, for any fixed « © (-1/2, 1/2]. The conditions q >-u2 ¥ uy = 1/4 = (Yay)? may be satisfied52 also for a subset of ¢ © (1/2,1/2] such that the maximum of the right-hand side does not fall in the u-content of the representation: If uz m+ ec, the points falling nearest to the maxima umqy-F1/2 do so at a distance U-imy= 1/2 - le] from it, so that only q>|el(1-le|) 20 is ™féquired. For any fixed q in (0, 1/4], thus, we can find a range of © in (-1/2,1/2] which allows for unbounded values of up . Writing q= k(I-k) , adding -1/4 to both sides of the inequality to complete squares, the condition on © for fixed real k is given by |k-1/2|< 1/2-|e|. In particular, for k=0 alle but e = 1/2 allow for the unbounded range of u . Bargmann (23, see also 45) considers only SU(1,1) representay- tions: C&*" the principal (integer) series for 1/4 1/2i. e. k © [-»,0)U (1,%), where q<0. In those intervals, the distance between u, and uy is larger than 1, sono e can be chosen such that the sequence u=e+m, m€z avoids falling in the negative- |c|“ region unless falls on the boundary zeros. The row index u will then have an upper or lower bpupd. For, consider what happens when f_ acts on the ‘Jouest state! dX" | > ='k: It gives zero since ¢ = 0 , hence @f.7 does not bélong to the irregugible eigenfuncttBa set of the algebra, which consists then of {6} The same statement holds for &", up =~ k +1, and Bally Feflects the inversion symme- try under kUZ@7 5 ky Correspongiggly, for J, , the representation must contain a ‘highest state’ 6°%', uy = k- 1 which the raising operator wil} turn to zero. The 2 irreducible eigenfunction set will then be (ohm } , and the analogue statement holds for aa! ' "these lower- and upper-bound representations were called 'Dis- crete’ series D, and D, by Bargmann (23). Since he worked with su(1,1), he allowed only for integer or half-integer values in the spectrum fu} of lj , so was only allowed to be an integer or half-integer. Here, it may be any real number. When |k - 1/2| < 1/2 i. e. k€ (0,1), @ covers the interval (0,1/4) twice, and the point q = 1/4 once (corresponding to k = 1/2). In this interval, the distance between uj and u5 is less than 1, so that the sequence y =e +m, m@Z may ‘jump! the forbidden range of u . This constitutes the exceptional interval of the con- tinuous series seen in the last Section. Again, lower- and upper- bound representations occur when the lowest yu falls on uy or up andImk FIGURES 1 S1(2,R). (a) as function *% of real q. (b) as function of complex k for q=k(I-k). Shaded regions correspond to allowed self-adjoint representations. Dashed lines indicate that the region does not in- clude the boundary. The parallel boundaries at ¢ =-1/2 are to by ig gti ied. Heavy lines indicate Bargmann's continuous series C_ and cit : q The continuous C° irreducible representation series of54 FIGURES 2 (a) the zeros of |ot 12, (b) the zeros of |e, fe Shaded regions correspond to negative Values of these quantitieS'which must be excluded from the range of u.55 the highest on ut or nf . We can make use of the inversion symmetry k@1-k inorder to“ascribe the lowest ) of the lower-bound rep- resentation to u, = k, for k € (0,1) and, correspondingly, the ghest uy of the upper-bound representation to wy = - k for k (0,1). The points k= 1 and k = 0, both mapping on q = 0 deserve special attention. The first, k = 1,,is on par with the neighbouring points along k, it correspondsto 0] containing u= +(1+m),m=0,1,2,...- The point k = 0, however, is ynique: It lies in the origin of Figs. 2a and 2b, on zeros of both Cy, and c,,. In consequence, JI, = 0, and so the eigenfunction ¢, is the basis for the one-dimensional tri- vial representation of the algebra by zero. As a representation for the group, 69 constitutes the trivial one-dimensional unitary repre- sentation. In summary, the lower- and upper-bound self-adjoint representa- tions-of 51(2,R) are Dy , k > 0 containing the eigenfunctions $, of Wy and C for u = #(kbn), m=0,1,2, If k € (0,1) we can write ue tm as for the continuous series, but keeping the congruence interval of «¢ to be [-1/2,1/2]. The discrete series provide the boundaries of the open regions of the exceptional interval in Figs. la and 1b. In particular, the CY series convers,the interval q > je point q = 0 belongs yto the fiscrete series Dj and to the trivial - Similarly, the C1’* serigs covers the interval q > 1/4, while the point q = 1/4 belokgs to 049. For 0we can give the generator matrix elements u sp le fe, k, ke | gk fe JF Em Uglmmt = ne Ly Omve!, j21,2,0, of #0, (3.22) as (Uperes = by gle # ml = by ge wl, (3.23a)56 + FIGURE 3. Discrete Dy and continuous C® irreducible representation series. 4ol Im k Im k 0 Co aoa c, Otan 5 = ~ abgtddy * Ty ~B) tans/Z , (3.29a) x eX + oX' = (cagtdog)? + (ebgtddy) ? & cos? (HI lagtan $ - og)* + (dy-bytan $17). (3.296) 3.14 s1(2,) algebras of differential operators. Global transformations will be futher analyzed below; here we are interested in infinitessimal transformations which in 3 x 3 ma~ trix form (3.10) appear as 1 1 ‘ay b, I-z 6 (5,4+6)) 99). 1 +2 15,78 +6,7545,95) + 082) -(,7 2 2 1 o C9 4 tay 2010.0 tis a.) tela 2 iO Coe (3.30a) On a dense subspace of the space of functions (3,28a) on the coset space M = P\ SL(2,R), they are given by 6x1") = LTS pH 8g + 59 THF HO (S11 6(x,61, (3.300) where te , k20,1,2 will be first-order differential operators in x and “@ . Replacement of (3.30a) in (3.29), Taylor expansion and collection of §,'s allows us to find them: A} = dcostay + sinba,), (3.31a) uf = ilsinbay ~ cos69,), (3.310)62 M : ty 74 (3.31¢) They satisfy the commutation relations (3.4). Prom here we can find the raising and lowering operators (3.13) as Gt eae alps ce# ay - a, (3.328) Me te . Gre Oi) - A, ee lay + 20,1. (3.32b) Most important, we find the Casimir operator (3.14) in this realiza- tion to be 2 2 2 Cat ap a =~ alta). (3.33) 0 One can see the special feature of the Iwasawa decomposifion (3.27a) and the parabolic coset space M=P\SL(2,R): That J[p is a differential operator in @ and” in yx, . Functions on M~ belong- ing. to a Given eigenvalue k{1-k) under @° can be characterized through & bo) = eo ™ G16). (3.34) On these spaces of functions, 3, may be replaced by -k, so as to obtain from (3.31) and (3.32), ee ee a) (3.35a) a = day - (3.35b) This is the s1(2,R) algebra realization employed by Bargmann (23). Equation (3.34) will be used below to examine the global action of SL(2,R) on this space of functions. 3.15 Basis functions on the circle. In Bargmann's realization, the normalized eigenfunct igng oe in Eq. (3.16) are proportional to the exponential functions 2+?. “The eigenvalue we saw, must range in integer steps y=ctm,m=0,1,2, passing through e(for Cy) or starting from ajyplue use, for svecified in Eqs. (3.24) so that etud = 2% We let @ range on the unit circle S,(i.e. $=$ mod?) and write the(possibly multivalued) function £(¢) ona space associated with Rk and e as fle) = of SUI g gine (3.368)where we are indicating the sum over the proper u-set as {h}_@ + © >: : >: : © to . = [ for Chik)! = [for 0, , withe = 4k]. (3.36b) 7 mo = The possible multivaluation will not be a problem when calculating inner products ~ with a measure stil] go be found- , as any sesqui~ linear inner product involving ,(g®)" and 82? with uzzctm) and ug=etmg will cancel out the e“*” factors, or otherwise render them innocuous. We have not yet written down the nofmalé: ation constants for the eigenfunctions. If we asqyge that L“(Sj;] is the proper space, these constants will be (2m) for all m. But we may have other spaces. In general their normalization will be prescribed by the condition (3.19)-(3.21). k We may fix ghe normalization of $ for u,=e and from this ob- tain all other $-,,, » with € and m ra ging as in (3.36), through ascending or descending by means of tr : (a Lem otter of. ELEN! 8 161 = mls ke - [ TT ey en] Care (le (3.37a) ne : We have the realization (3.35a) for Ji!" and #16) =" o“*, hence m-1 ot (0) = Ke [ 1 Siycen] | Lie (age cee? 7 m-1 ‘ ‘ [ fae | (erty cot agecny MTetlett16 ie 1 ce | py een | etleamis © | n=0 cp peen (3.376) ce bt 1 ett [1/2 dleeme © Nesm eine (TR) © n=0 where nt is the product of the inyerse of the Shy in °°" (3.21) for u=e,etl,...,c4(n-1). We shall not need a definite phase convention in these notes. It should only be noted that phases can be arranged so that the ¥¢,,-1, that is, that the raising and lowering operators have purely pésitive matrix matrix elements. This is the phase definition followed in (23, Sect. 6d) and known as Bargmann's convention. + hases f th pl Ypy Of the 6364, We use the Pochhammer product symbol m-1 T] (atm) = afatt)---(atm-1)=(a),,* Plat) /P (a), (3.37¢) m-1 [] (o-n) = b(b-m+l),, = P(b+7)/T(b+I-m), (3.374) i so as to write (3.37b) in the form hk oe gfletm) ¢ Seam (0) = Kerem (3.38a) be [mn 11/2 [rrestae) rfertetm | 1/2 one = : P(t-ee) 7 (teterm) |1/? ‘tm | (1-kEe}m Tletk) Tle+t-REm) “lrlkte)— TU1-kee+m) (3.38b) We now examine the characteristics of the oft square root fac~ tor for the various representation series. We refer to (3.24) (3.36) for notation and index ranges. For the principal series C- in the nonexceptional interval, k=1/2+p, pER and hence the ratio‘of func- tions of the form y(T-b) /y(k) #y(1/2-d0) /y( 1/2440) has absolute value unity. Thus ch q>d, off = 1 (continuous (3.39a) ~ nonexceptional). For the exceptional interval k&(0,1), on the other hand, for kh S 1/2 we have (kta)/(I-kta) = 1. The ratio of T -functions becomes a product of such factors for a-te,te+,...,te+m-1, so that : ee < ( c o, 1 (continuous (3.39b) qk(I-k), ia, exceptional). Finally, for the discrete series oy , where e=th, we may write expli- citly ot be flr1/2) hath r(2k+m) [1/2 k, (1/2,=), 2" r(zeIr (me7) = 1, (discrete). (3.39¢) The inversion symmetry k + 1-k is not explicitly present in this formula, but a unitary intertwining operator may be set up to bridge these two representations (45, Sects. 2.3 and 2.4).65 3.16 In search of Hilbert spaces. We have now sets of functions @"(6] where wu 4is in the spectrum of Ip in agivendefinite representation, (C- or D,] 68, (the cir- cle) Given by (3.8). We search for inner préducts where these are orthonormal and complete. Completeness of these denumerable bases, when proven, will define the Hilbert spaces as the closure of linear combinations of these functions. The most general sesquilinear inner product on Sj may be written as (4,9) "e) foe fo sto)" a!) (9,67) gio). (3.40) der th id nh Under this, jnner,produdt thd alggbr: Crane Ob tel arg) Boeharg potty tes rements ifposed hy, j-1,2 511. e. by hermiticity of Jy’, integrating by parts: (alg) 4,9) (ee) = fo fore, stl)" ae (9,6") gto") $1 8; = fa fas 566)" -aagal®*)(6,6°N1g(0") ob “(gah gy tele f= ‘fo 610)" a) (6,6) 295,910] 1 1 - fefe 616)" 42g, 0! *) (8,6°)1 916"). Sy) 3; (3.41) g)generators (3.35) must be hermi- + (Wie have gen before the requi- Jp = SUT We now impose the This leads to ' (ke) 1 Ne do fdo" §(6) [(a,-3g.)0 (9,6")]g(') = 0 (3.42) | od which is to hold for all 616) and g(@')in a space I), there is a raphe subtle joint to the ensying reasoning, and that is since may be di ferent frop L°(S7), what (3.42) actually implies is that 2(9,0')=(34-341) 2 7 6,0'1, as a function of @, is orthogonal in the sense of L*(S,) to §(¢), and similarly orthogonal to g(¢') in the same sense. If we were to put 2(9,$') as an inner product measure in Plagg gf 219.0") in (3.40), the result would be zero for all §,g in v£) and hence z(¢,@') is equivalent to the zero func- tion itself. On the basis of this equivalence, we may replace z(¢,@') by zero and thus conclude (ag-2gr)a!®#) (4,9%)= 02a El (6,97) a0 BE (g-4'), (3.430) i. e. 2 is a function only of the angle difference @-$' . As the66 basis fynctions e 8 with as specified by the representation cy or Dy, is assumed complete in eee ee may expand qe ae as albeliy) » > whe a (3.436) 7 as in (3.36), and ubse as yet unknown and put it in (3.40) with the orthonormality condition imposed on the basis functions, using the Fourier series orthogonality results for the exponential func- tions. We thus develop a) with Heine, bee te “fv fon neem efter) mm < Huis flew ies 3 hut capa ot, often") or by 2, fb) ok PT Utes Soe tC ete of) ayn ; fo eilaenlé 4) fo» ila! nh ‘$1 SI . {2n) Pixel? che? uke (34a) We fellows that the proper coefficients for the weight function **(9-$") are (1-kte) Be tance of mo Galo Tee y? (ke), 1 Tle+I-k) _Tletktm) = (3.44b) (2a [Ke]? Tletk) TP (e#T-ketm) ol T(kte) Pl 1-bte+n) (an|Ke |)? Pli-kte) rlktetm |. Bal 7 Local and nonlocal weight functions. Let us now examine the various fepresentation series in order to give the description of the 4 spaces. For the conginuous nonexceptional series Cr we have n©Z and (3.39a), since osm is only a phase. Then a VP) « (eet? Sectors meZ Hanke 5(y) (3.45a)67 We have hence a btrore . 25) (3.456) and the inner product is the ordinary L2-product on the circle. For the continuous exceptional series, ery _ ake gélermy , a ae aE (yp) yt (20 iene 2 me | OD) -m © m=7 es [soa an Suet saltln (2a |e (1? | Lod (bel nt Lb een efev , P = oral (1, T-breskerese) +0F U1 1teesteese™®)-1] (2n[ ke) (3.46a) When e=0 it was proven by Bargmann (23, Eqs. (8.7), (8.9) and (8.11)) that this expression reduces to (1b) pik gO oom. 2 rk) kT (y) = ein Set Tos ¥) eZ (ed, w7r[Kol> rlk-1/2) a (3..46b) Bargmann (23, Sect. 8c) also proves that J? is a Hilbert space when RE[1/2,1), while Sally (45, Sect. 2.4) proves this for the remaining cases. For the discrete series a, finally, we can use (3.39c)-(3. 4h) in order to find oteletmly (3.47) ay oF tte dsthse™) This series converges absolutely for k > 1, conditionally for 1/2 < k< 1 (excluding y=0,2n,... although its integral with boun- ded functions is bounded), and diverges for k=1/2. For 0 You might also like
The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good LifeFrom EverandThe Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good LifeRating: 4 out of 5 stars4/5 (5820) The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You AreFrom EverandThe Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You AreRating: 4 out of 5 stars4/5 (1093) Never Split the Difference: Negotiating As If Your Life Depended On ItFrom EverandNever Split the Difference: Negotiating As If Your Life Depended On ItRating: 4.5 out of 5 stars4.5/5 (845) Grit: The Power of Passion and PerseveranceFrom EverandGrit: The Power of Passion and PerseveranceRating: 4 out of 5 stars4/5 (590) Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space RaceFrom EverandHidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space RaceRating: 4 out of 5 stars4/5 (897) Shoe Dog: A Memoir by the Creator of NikeFrom EverandShoe Dog: A Memoir by the Creator of NikeRating: 4.5 out of 5 stars4.5/5 (540) The Hard Thing About Hard Things: Building a Business When There Are No Easy AnswersFrom EverandThe Hard Thing About Hard Things: Building a Business When There Are No Easy AnswersRating: 4.5 out of 5 stars4.5/5 (348) Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic FutureFrom EverandElon Musk: Tesla, SpaceX, and the Quest for a Fantastic FutureRating: 4.5 out of 5 stars4.5/5 (474) Her Body and Other Parties: StoriesFrom EverandHer Body and Other Parties: StoriesRating: 4 out of 5 stars4/5 (822) The Emperor of All Maladies: A Biography of CancerFrom EverandThe Emperor of All Maladies: A Biography of CancerRating: 4.5 out of 5 stars4.5/5 (271) The Sympathizer: A Novel (Pulitzer Prize for Fiction)From EverandThe Sympathizer: A Novel (Pulitzer Prize for Fiction)Rating: 4.5 out of 5 stars4.5/5 (122) The Little Book of Hygge: Danish Secrets to Happy LivingFrom EverandThe Little Book of Hygge: Danish Secrets to Happy LivingRating: 3.5 out of 5 stars3.5/5 (401) The World Is Flat 3.0: A Brief History of the Twenty-first CenturyFrom EverandThe World Is Flat 3.0: A Brief History of the Twenty-first CenturyRating: 3.5 out of 5 stars3.5/5 (2259) The Yellow House: A Memoir (2019 National Book Award Winner)From EverandThe Yellow House: A Memoir (2019 National Book Award Winner)Rating: 4 out of 5 stars4/5 (98) Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New AmericaFrom EverandDevil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New AmericaRating: 4.5 out of 5 stars4.5/5 (266) A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True StoryFrom EverandA Heartbreaking Work Of Staggering Genius: A Memoir Based on a True StoryRating: 3.5 out of 5 stars3.5/5 (231) Team of Rivals: The Political Genius of Abraham LincolnFrom EverandTeam of Rivals: The Political Genius of Abraham LincolnRating: 4.5 out of 5 stars4.5/5 (234) On Fire: The (Burning) Case for a Green New DealFrom EverandOn Fire: The (Burning) Case for a Green New DealRating: 4 out of 5 stars4/5 (74) The Unwinding: An Inner History of the New AmericaFrom EverandThe Unwinding: An Inner History of the New AmericaRating: 4 out of 5 stars4/5 (45)