Rigid Particle and its Spin Revisited Matej Pavˇsiˇc1 6 Joˇzef Stefan Institute, Jamova 39, SI-1000 Ljubljana, Slovenia 0 0 2 n a J ABSTRACT 7 2 The arguments by Pandres that the double valued spherical harmonics provide a basis 3 v fortheirreduciblespinorrepresentationofthethreedimensionalrotationgrouparefurther 4 2 developedandjustified. Theusualargumentsagainsttheinadmissibilityofsuchfunctions, 3 2 concerning hermiticity, orthogonality, behavior under rotations, etc., are all shown to 1 4 be related to the unsuitable choice of functions representing the states with opposite 0 / projections of angular momentum. By a correct choice of functions and definition of h t - inner product those difficulties do not occur. And yet the orbital angular momentum in p e the ordinary configuration space can have integer eigenvalues only, for the reason which h : v have roots in the nature of quantum mechanics in such space. The situation is different i X in the velocity space of the rigid particle, whose action contains a term with the extrinsic r a curvature. 1Email: [email protected] 1 Introduction The theory of point particle whose action contains not only the length, but also the extrinsic curvature of the worldline has attracted much attention [1]–[5]. Such particle, commonly called “rigid particle”, is a particular case of rigid membranes of any dimension (called “ branes”). The rigid particle behaves in all respects as a particle with spin. The spin occurs because, even if free, the particle traces a worldline which deviates from a straight line. In particular, it can be a helical worldline [3]. In the absence of an external field, the constants of motion are the linear momentum p and the total angular µ momentum J which is the sum of the orbital angular momentum L and the spin µν µν S . In the presence of a gravitational field, the equation of motion for rigid particle was µν shown [3] to be just the Papapetrou equation [6]. The algebra of the (classical) Poisson brackets and the (quantum) commutators resembles that of a spinning particle and it was concluded that the rigid particle leads to the Dirac equation [4]. In refs. [7, 8] a counter argument occurred, namely that the spin of the rigid particle is formally like the orbital momentum, with the only difference that it acts not in the ordinary configuration space, but in the space of velocities. Since orbital momentum is well known to posses integer values only, it was concluded that the rigid particle cannot have half-integer spin values. In the present paper we will challenge that conclusion. A theoretical justification of why orbital angular momentum is allowed to have integer values only, and not half-integer, had turned out to be not so straightforward, and the arguments had changed during the course of investigation. Initially [9] it was taken for granted that the wave function had to be single valued. Then it was realized [10] that only experimental results needed to be unique, but the wave function itself did not need to be single valued. So Pauli [11] found another argument, namely that the appropriate set of basis functions has to provide a representation of the rotation group. He argued that the spherical functions Y with half-integer l fail to provide such representation. lm Amongst many subsequent papers [12],[16]–[18] on the subject there are those by Pan- dres[13,14]who demonstrated that theabove assertionby Pauli wasnot correct. Pandres conclusion was that the functions Y with half-integer l do provide the basis for an ir- lm reducible representation of the rotation group. Pandres explicitly stressed that he had no quarrel with Pauli’s conclusion concerning the inadmissibility of multivalued quantum mechanical wave functions in descriptions of the ordinary orbital angular momentum, 2 although he took issue with the argument through which Pauli had reached that conclu- sion. In the following I am going to clarify and further develop Pandres’ arguments. In particular, I will show that although the usual orbital angular momentum in coordinate space indeed cannot have half-integer values, the situation is different in the velocity space of the rigid particle. In the velocity space the functions Y with half integer l and m lm are acceptable not only because they do provide a basis for representation of the rotation group, but also because the dynamics of the rigid particle, its equations of motion and constants of motion, are different from those of a usual quantum mechanical particle. So the linear momentum π in velocity space is not a constant of motion and the eigenfunc- µ tions of the operator πˆ are not solutions of the wave equation for the quantized rigid µ particle. Since it has turned out [8] much more convenient to formulate the theory not in the velocity, but in the acceleration space, I will explore the ’orbital angular momentum’ operator in the latter space, and show that its eigenvalues can be half-integers. 2 The Schr¨odinger basis for spinor representation of the three-dimensional rotation group Amongst many papers [10]–[12],[16]–[18]onangular momentum andits representation the paper by Pandres [13] —with the above title— is distinct in claiming that the rotation group can be represented by means of double valued spherical harmonics. I will re- examine his arguments and confirm that Pandres’s understanding of the problem was deeperfromthatofotherresearchers. Half-integerspinisspecial—incomparisonwiththe integer spin— in several respects, the most notorious being its property that a 2π rotation does not bring the system in its original state: the additional 2π rotation is necessary if one wishes to arrive at the initial situation. A spin 1 system has an orientation– 2 entanglement with its environment. This has consequences if one tries to describe the system by employing the Schr¨odinger representation. One immediately finds out that this cannot be done in the same way as in the case of a system with an integer value of angular momentum. The spherical harmonics with half-integer values do provide a basis for the irreducible spinor representation of the three-dimensional rotation group, provided that one imposes certain “amendments” to what is meant by “forming a representation”. Such amendments should not be considered as unusual for spinors—which are themselves unusual objects in comparison with the more “usual” objects—and are in close relation 3 to orientation–entanglement of a spinor object with its environment, which is illustrated in the well-known example of a classical object attached to its surroundings by elastic threads. Evidently, as stated by Misner et al.[15], in the case of spinors there is something about the geometry of orientation that is not fully taken into account in the usual concept of orientation. Our first ammendment is related to the fact that under rotations the spinorial wave functions do not transform as scalar functions. The failure of half-integer spherical har- monics to behave as scalars has been taken as one of the crucial reasons to reject them in the description of orbital angular momentum, which was indeed reasonable. But if we want to use them in order to represent spinors, then obviously they should not behave as scalars under rotations. Our second ammendment is related to the fact that half-integer spherical harmonics do not form a fully irreducible representation of the rotation group. The total space splits into an infinite dimensional and a finite dimensional subspace. The latter subspace, denoted , is not invariant; if we start from a state, initially in , after a rotation we l l S S obtain a state with components not only in , but also outside . But it turns out that l l S S the projection of the state onto transforms just as a spinor, its norm being preserved. l S The occurrence of the states outside has no influence on the behavior of the projected l S state. Therefore we can say that the basis states of do form a representation of the l S rotation group in such a generalized, “ammendedd”, sense. This makes sense, because the states outside are unphysical, due to the fact that with respect to the latter states l S the expectation value of the nonnegative definite operator L2 + L2 is negative. So we x y have to disregard those “ghost-like” states, and we do this by performing the projection of a generic state onto the physical space . We also show that an alternative way of l S eliminating the unphysical states from the game is in adopting a suitably renormalized inner product with a consequence that norms of the unphysical states are zero. 4 2.1 Choice of functions Let L be a set of Schr¨odinger-type operators i ∂ ∂ L = i(cotϑcosϕ +sinϕ ) x ∂ϕ ∂ϑ ∂ ∂ L = i(cotϑsinϕ cosϕ ) y ∂ϕ − ∂ϑ ∂ L = i (1) z − ∂ϕ where ϑ and ϕ are the usual polar coordinates. Let us consider the functions Y (ϑ,ϕ) which satisfy the equation lm L2Y = l(l+1)Y (2) lm lm L Y = mY (3) z lm lm where 1 ∂2 1 ∂ ∂ L2 L2 +L2 +L2 = sinϑ (4) ≡ x y z −sin2ϑ∂ϕ2 − sinϑ∂ϑ ∂ϑ (cid:18) (cid:19) For integer values of l the Y are the familiar single-valued spherical harmonics, whilst lm for half-integer values of l the Y are double valued functions. lm Ingeneral,foranyintegerorhalf-integervaluesoflthefunctionsthatsatisfyeqs.(2),(3) are given by 1 2l+1 Π(l+m) 1 1 dl−m Y = eimϕ( 1)l sin2lϑ (5) lm √2π − r 2 sΠ(l m)2lΠ(l)sinmϑ(dcosϑ)l−m − l,l 1,..., l if l integer m = − − (6) l,l 1,..., l, l 1, l 2,... if l halh-integer (cid:26) − − − − − − where Π(l) Γ(l+1) is a generalization of l! to non integer values of l. ≡ Besides (5) there is another set of functions which solves the system (2),(3): 1 2l+1 Π(l m) 1 1 dl+m Z = eimϕ( 1)(l+m) − sin2lϑ (7) lm √2π − r 2 sΠ(l+m)2lΠ(l)sin−mϑ(dcosϑ)l+m l, l+1,...,l if l integer m = − − (8) l,l+1,...,l,l+1,l+2,... if l halh-integer (cid:26) − 5 The function Z coincide with Y for integer values of l only. In the case of half-integer lm l,m l-values, they are different. If we define the raising and lowering operators as usually ∂ ∂ L +iL L = eiϕ +icotϑ (9) x y + ≡ ∂ϑ ∂ϕ (cid:18) (cid:19) ∂ ∂ L iL L = e−iϕ +icotϑ (10) x y − − ≡ −∂ϑ ∂ϕ (cid:18) (cid:19) we find [13] for half-integer values of l Y , m = l 1,l 2,..., l; l 2, l 3,... L Y l,m+1 − − − − − − − (11) + lm ∝ 0 , m = l, m = l 1 (cid:26) − − L Y Y , m = l,l 1,..., l, l 1,... (12) − lm l,m−1 ∝ − − − − L Z Z , m = l, l+1,...,l,l+1,... (13) + lm l,m+1 ∝ − − Z , m = l+1, l+2,...,l; l+2,l+3,... L Z l,m−1 − − (14) − lm ∝ 0 , m = l, m = l+1 (cid:26) − Now let be a function space which is spanned by the basis functions Y for a given l lm S value of l and for m = l,...,l. Further, let be a space spanned by Y for a given l lm − O value of l and for m = l 1, l 2,.... Analogous we have for functions Z . lm − − − − From the relations (11)–(14) we see that although the repeated application of L to − Y does not give zero, but gives Y ,Y ,..., i.e., brings us out of into , the l,−l l,−l−1 l,−l−2 l l S O reverse is not true. If L is applied to Y the result is zero. This comes directly + l,−l−1 l ∈ O from the identity L L Y = (l m)(l m+1)Y (15) ± ∓ lm lm ± ∓ from which we find L L Y = 0 (16) + − l,−l Since L Y Y we have ffrom eq.(16) that − l,−l l,−l−1 ∝ L Y = 0 (17) + l,−l−1 Analogously, from L L Z = 0 we obtain the relation − + l,l L Z = 0 (18) − l,l+1 6 In particular we have, L Y Y (19) − 21,−21 ∝ 12,−32 L Y = 0 (20) + 1,−3 2 2 This can be verified by direct calculations using the differential operators (9),(10) and functions (5),(7). For instance, taking i Y21,−23 = L−Y21,−12 = −π sin−3/2ϑe−i32ϕ (21) we find ∂ ∂ i L+Y21,−23 = eiϕ ∂ϑ +icotϑ ∂ϕ −π sin−3/2ϑe−i32ϕ = 0 (22) (cid:18) (cid:19)(cid:18) (cid:19) Similar is true for the functions Z , with the role of L and L interchanged. lm + − The spherical harmonics for half-integer l-values differ from those for integer l-values in the properties such as those expressed in eqs.(11),(12), i.e., L Y = 0 ,and L Z = 0 (23) − l,−l + l,l 6 6 which imply theexistence of functions outsidethe space . Those functions have negative l S eigenvalues of the operator L2 L2 = L2 +L2, and therefore, according to the ordinary − z x y criteria cannot beconsidered asrepresenting physical states2. Physical states areobtained by projecting an arbitrary state into the subspace . l S In the following I am going to show that using the functions (5),(7) with the properties (11)–(18) the usual arguments against such functions as representing states with half- integer angular momentum do not hold. The confusion has been spread into several directions. Besides Pauli’s very sound argument concerning the behavior of the system under rotations there are other claims such as : functions Ylm and Yl′m are not orthogonal, • they may have infinite norms, • L are not Hermitian, i • other problems [12]. • 2Analogously,negative normstates ingaugetheories areunphysical, andyetthey canbe consistently employed in the formulation of the theories. 7 I will now discuss those claims. Orthogonality - One finds that not only the functions Ylm and Ylm′, but also Ylm and Yl′m,belongingtotheset(5),(7)areorthogonal. Thisisnotthecaseforthesetoffunctions used by Merzbacher and Van Winter who start formally from the same expression (5), but restrict the range of allowed l,m values in a way different from (6), so that for positive m-values they take functions (5), whilst for negative m-values they take functions (7)3. Infinite norms-However, aproblemremainsevenwithourchoiceoffunctions. Certain states of , namely those with m = 3, 5,..., l, have infinite norms, i.e., lm lm is Sl −2 −2 − h | i infinite. But, as stated by Pandres, it is a well known fact that the inner product can be redefined so to obtain finite norms, normalized to unity. Let us consider the quantities Gmm′′(ǫ) = dΩYl∗mYlm′′ (24) ZΩ−ǫ where we have performed a cut off in the integration domain. Instead of integrating over the domain Ω = (ϕ,ϑ) ϕ [0,2π], ϑ [0,π] (25) { | ∈ ∈ } we integrate over a truncated domain Ω ǫ = (ϕ,ϑ) ϕ [0,2π], ϑ [ǫ,π ǫ] (26) − { | ∈ ∈ − } The quantities Gmm′′(ǫ) are zero, if m = m′′, and different from zero and finite, if 6 m = m′′. This is so also if m = 3, 5,..., l. −2 −2 − Now let Gm′′m(ǫ) be the inverse matrix to Gmm′′. So we have dΩYl∗mYlm′′Gm′′m′(ǫ) = Gmm′′(ǫ)Gm′′m′(ǫ) = δmm′ (27) ZΩ−ǫ Definition I of inner product - Thelatterrelationisvalidforanyvalueofǫ, whatever small. Using eq.(27), we define the inner product between two functions according to: (Ylm,Ylm′) = lim dΩYl∗mYlm′′ Gm′′m′(ǫ) = δmm′ (28) ǫ→0 ZΩ−ǫ If m = m′, this can be written as (Y ,Y ) lm lm ǫ (Y ,Y ) = lim , N (ǫ) = (Y ,Y ) (29) lm lm lm lm lm ǫ ǫ→0 Nlm(ǫ) 3Suchunsuitable setoffunctions has beenrecently usedbyHunter et al.[19], who otherwisecorrectly argued that half integer spherical harmonics can represent spin. 8 For values of m and m′ other than {−23,−52,...,−l}, we have limǫ→0Gmm′′(ǫ) = δmm′′ and lim Gm′′m′(ǫ) = δm′′m′, so that in this particular case the inner product coincides with ǫ→0 the usual inner product. Hermiticity - By using the set of functions (5),(7) and the relations (11)–(14) one finds that the operators L , L2 are Hermitian with respect to . This is not the case if one i l S uses a different set of functions—as Merzbacher [17] and Van Winter [12] did—such that, e.g., for l = 1 consists of Sl 2 Y sin1/2ϑeiϕ/2 (30) 11 22 ∝ Y sin1/2ϑe−iϕ/2 (31) 21,−21 ∝ With respect to the above set of functions (30,31) the angular momentum operator is indeed not Hermitian. Hence, the set of functions as used by Merzbacher and Van Winter is indeed not suitable for the representation of angular operator. But the set (5),(7) used in the present paper (and also by Pandres) is free of such a difficulty and/or inconsistency as discussed by Merzbacher and Van Winter. Let me illustrate this on an example. From (5) we have for l = 1 the following subset 2 of normalized functions: 1 S2 i Y12,12 = πsin1/2ϑeiϕ2 (32) i Y12,−21 = −πcosϑsin−21ϑe−i2ϕ (33) If, using (9),(10), we write 1 L = (L +L ) (34) x + − 2 1 L = (L L ) (35) y + − 2i − we find after taking into account L Y = 0 (36) + 1,1 2 2 L Y = Y (37) + 1,−1 1,1 2 2 2 2 L Y = Y (38) − 1,1 1,−1 2 2 2 2 L Y = Y (39) − 1,−1 1,−3 2 2 2 2 L Y = 0 (40) + 1,−3 2 2 9 that the matrix elements of angular momentum operator satisfy: 1 11 L 1 1 = = 1, 1 L 11 ∗ (41) h22| x|2 − 2i 2 h2 −2| x|22i i 11 L 1, 1 = = 1, 1 L 11 ∗ (42) h22| y|2 −2i −2 h2 −2| x|22i where hlm|Li|lm′i = Yl∗mLiYlm′dΩ , dΩ ≡ sinϑdϑdϕ (43) Z Here we have also taken into account that the states with the same l but different m values are orthogonal. The matrix values (41)–(42) are just the standard ones. The fact thatL Y = 0 has no influence on the values of matrix elements of angularmomentum − 21,−21 6 operator, calculated with respect to the basis states of . This is so because of eq. (40). l S In eqs. (41)–(42) we have just the property that the matrix elements of a Hermitian operator have to satisfy. Let us now check by explicit integration whether the operators L , i = 1,2,3, satisfy i the requirement for self-adjointness (φ,L ψ) = (L φ,ψ) for all φ, ψ (44) i i l ∈ S withtheinnerproduct beingdefinedaccording toeq.(28). Sinceanyphysically admissible φ, ψ is by definition a superposition of Y , it is sufficient to show the relation (44) lm l ∈ S for functions Y only. Taking into account the relations (34),(35) we find that the lm l ∈ S condition for self-adjointnes of the operators L becomes i (Ylm,L+Ylm′) = (L−Ylm,Ylm′) if m = m′ +1 (45) (Ylm,L−Ylm′) = (L+Ylm,Ylm′) if m = m′ 1 (46) − If we calculate the matrix elements by adopting the usual definition of the inner product, we have (m = m′ +1): 2π π ∂ ∂ dϕ dϑsinϑYlm∗ e−iϕ −∂ϑ +icotϑ ∂ϕ Ylm′ Z0 Z0 (cid:18) (cid:19) 2π π ∂ ∂ = dϕ dϑ e−iϕ∂ϑ (Yl∗msinϑ)−icosϑ ∂ϕ Ylm∗ e−iϕ ϑ Ylm′ +Bmm′ Z0 Z0 (cid:20) (cid:21) 2π π ∂ ∂ (cid:0)∗ (cid:1) = dϕ dϑsinϑ eiϕ +icotϑ Ylm Ylm′ +Bmm′ (47) ∂ϑ ∂ϕ Z0 Z0 (cid:20) (cid:18) (cid:19) (cid:21) 10