LAEFF 95/26 Introducing spin to 6 9 classical phase space 9 1 n J. Le´on∗ and J. M. Mart´ın† a J Laboratorio de Astrof´ısica Espacial y F´ısica Fundamental, INTA 5 Ap. 50727, E 28080 MADRID, Spain 1 and 1 Instituto de Matem´aticas y F´ısica Fundamental, CSIC v Serrano 123, E 28001 MADRID, Spain 1 1 December 26, 1995 0 1 Abstract 0 The kinematic degrees of freedom of spinning particles are analyzed 6 and an explicit construction of the phase space and the simplectic struc- 9 turethat accomodates them is presented. A Poincare invariant theory of / h classical spinning particles that generalizes the work of Proca and Barut p to arbitrary spin is given using spinor variables. Second quantization is - naturallyconnectedtotheunphysicalnatureofzitterbewegung. Position t n variables can not be disentangled from spin in a canonical way, nor can a the phase space be reduced to the usual description (x,p) and a vector u spin. q : Pacs: 03.20.+i, 03.65.Sq, 03.30.+p,11.30.Cp v i X r a ∗e-mail: leon@laeff.esa.es †e-mail: [email protected] 1 ClassicalMechanics is based on the idea of the point particle endowed with mass. It successfully determines the evolution of the particle by the sole virtue of the Newton Principles. These have been formulated in a variety of ways of which in this paper we will use the economical and elegant Action Principle. Assuming that the action A describes the motion of a free particle through the Euler-Lagrange equations, we can impose δA = 0 for translations and Lorentz transformations to ensure that any pair of independent inertial observers will detectthe samemotionofthe particle. Thatthis givesthe Galileoprinciple (or its equivalent, the First Newton Law) can be seen as follows: Be xµ,x˙µ and τ the coordinates, velocities and proper time of the particle (x˙µ =dxµ/dτ). The invariance of the action implies ∂L δA=(p δxµ)|τ2 =0, p ≡ (1) µ τ1 µ ∂x˙µ where δxµ = ǫµ or δxµ = ǫµxν for translations or Lorentz transformations ν respectively. From the above one obtains the conservation of momentum p µ and angular momentum L = x p −x p . In practice there are only seven µν µ ν ν µ independentconservedquantitiesp andL pν,becauseL∗ pν,whereL∗ isthe µ µν µν dual of L, vanish by construction. The object X = 1 L pν gives the part of µ p2 µν x orthogonal to p, so the particle trajectory is simply: x =X +f(τ)p (2) µ µ µ where f(τ) is a scalar function of the proper time. From this one gets: ~x(τ)−~x(0) ~p = =~v, (3) x0(τ)−x0(0) p0 which is the sought Galileo Principle written in the most accesible way to an inertial observer. Summarizing, the trajectory is a straight line completely determinedbythesixconstantsofmotionp / p2 andX . Theparticletravels µ µ alongthe trajectorywith constantspeedv. The physicalrequirementthatτ be p the proper time, i.e. dτ2 =dx dxµ, or x˙2 =1, gives f˙=1/m, where m= p2 µ istheparticlemass. Inaddition,theobservercansetuptime-adjustingittothe p particletime-byfixingtheinitialconditionf(0). Thisderivationisindependent of the explicit form chosen or the lagrangian, but a consequence of combining the action principle with the invariance under Poincare transformations. Classical spinning particles are typically described [1] like points endowed withamassandaspinangularmomentumS . Alongwithpthetotalangular µν momentum J =L +S . (4) µν µν µν is conserved. What canbe known about particle kinematics for a generic S ?. µν J is a constant, but whe have not proved the same for L nor for S, we only 2 know that S˙ = −L˙. The six constants of motion contained in J can be made explicit by 1 X = J pν, W =−J∗ pν =−S∗ pν (5) µ p2 µν µ µν µν W is the Pauli-Lubansky vector describing the spin of the particle, while X is a fixed position in terms of which we can express the trajectory as x (τ)=X +f(τ)p +r (τ) (6) µ µ µ µ The vector r, the radius, is defined by 1 r =− S pν. (7) µ p2 µν Itsorigincanbetracedbacktotheindependentexistenceoforbitalandinternal angular momentum. This vector yields a tiny deviation of order (spin/mass) from the trajectory of the spinless free particle Eq. (2). It is not restricted to be a constant , opening the possibility for a non rectilinear x(τ) (a classical versionofzitterbewegung). Sufficient coarseningwillonlyrevealthe “external” trajectory of (2) and possibly a finite constant spin W. Coarsening erases the internal angular momentum generated by zitterbewegung, the first term in 1 S =−(r p −r p )+ ǫ Wρpσ. (8) µν µ ν ν µ p2 µνρσ Ingeneralonewouldneedofadditionalsubsidiaryconditionstofixthe time evolution of r. For instance, a constant spin modulus gives a constant norm radius: 1 W2 S Sµν =− +p2r2 (9) 2 µν p2 An arbitrary antisymmetric tensor decomposes into the sum of two orthogonal planes. The conditionthatS correspondstooneandonlyone ofthese planes µν is 1 S∗ Sµν =2Wr=0 (10) 2 µν Finally,r isaspacelikefourvectorofconstantnorm,whoseonlyallowedmotion wouldbe a rotationin the planeπ orthogonalto W andp. The particle would r performahelicalmotioncomposedoftheuniformmotionproperofthespinless particle andthe rotationin the plane π . The helix threadnumber is not given r yet. A particular case of (9) and (10) is S x˙ν = 0 which would add that µν information. To describe the above situation beyond kinematics, one needs a lagrangian linear in velocities to avoid getting x˙ in terms of p. This is not enough still, since the spin degrees of freedom have to participate in the hamiltonian, and hence in the canonical formalism, in order to get a r dependent trajectory. 3 Here, we describe classical spinning particles in mathematical terms as belong- ing to irreducible representations of the Poincare group of mass m and spin s. We implement this idea in classical mechanics by including in the configura- tion space of the particle additional variables ξ and ξ˙ (functions of the proper time) transforming according to the representation chosen. Invariance of the actionunder translations(δξ =0)willbe achievedbymomentumconservation, whileLorentztransformations(δξ =ǫ(s)ξ)willrequiretheconservationoftotal angular momentum. The idea of using Dirac spinors in classical mechanics is originally due to Proca [2] and Schiller [3]. The latter author demonstrated that the hamiltonian proposed by Kramers [4] to describe the evolution of a classical dipole in an external electromagnetic field could be reformulated in terms of a Dirac bispinor. He then generalized the Hamilton-Jacobi formalism to include the motion of the dipole in a classicalfield theory framework. Proca intendedtobuildaanewpointmechanicsintermsofclassicalspinorstolaythe foundations of a new quantization programme. Barut and collaborators gave a vigorousimpulsetothis proposal-promotedbythemtoaclassicalmodelofthe Dirac electron [5]- and were able to show the emergence of QED from the path integral of this classical particle [6]. We now focus on the phase space of the spinning particle and recall that we are dealing with elementary systems, i.e. with irreducible representationsof thePoincareGroup[7]. Theξ’swillthenbethespinorstransformingaccording to the representation chosen. To the pair of variables ξ and its time derivative ξ˙ we will associate a pair of canonical conjugate variables (ξ,η), and enlarge the coordinatesandmomentaoftheusualphasespacetoP ={(xµ,p ),(ξ,η)}. ν Thus, eachelementary system is characterizedby the representationchosen for the phase space. P can be labeled with two indices (m,s) giving the mass and spin of the particle. We will form Lorentz scalars, vectors and tensors out of these by forming bilinears ηΓ ξ and combinations of x and p in the µ1···µn standard form.We also form higher spinors from ξ or η. We can give a simple recipe for the symplectic structure on P: Given any pair of functions A,B of P, we define the canonical braket as ∂A ∂B ∂A ∂B ∂A∂B ∂B∂A {A,B}= − + − (11) ∂xµ∂p ∂p ∂xµ ∂ξ ∂η ∂ξ ∂η µ µ wheretheorderofthefactorsinthelasttermofther.h.s. ischosenbynotational convenience. It can be seen easily that the above fulfills all the conditions necessary to become an apropriate Poissonbracket [8]. We now require Poincare invariance: 1. Under translations (the p are constants, so the hamiltonian is indepen- µ dent of x). 2. UnderLorentztransformations(theJ areconstants,sothehamiltonian µν must produce S˙ =−L˙). 4 3. Underparityandtimereversal(wewillonlyconsiderrepresentations(j,j) or (j,0)⊕(0,j) and their combinations for the Lorentz part, so η ∼ ξ†β where β is the parity matrix). The solution to the equations of motion have to be physically consistent. This can be translated into three physical requirements: 1. Physical trajectories are in Minkowski space-time at all times. We nor- malize the proper time to the lenght of the trajectory, x˙2 =1 . 2. Physical variables remain in the representation space chosen (the invari- ants labeling the representation have to be constants). 3. Finally, there is only one object in phase space (combined with item 3. above this will lead to charge conjugation and CPT [9]). We will not analyze these issues in general, as they are well known [10] from relativistic particle theory. As said above, to disentangle momentum from velocity, one needs a la- grangian linear in velocities. This singular case is best treated by the method of Fadeev and Jackiw [11] where the momenta are considered as independent variables in configuration space, obtaining the same results than the standard canonicaltreatmentwithlessereffort. WewilluseageneralizationoftheProca lagrangian[2] for arbitrary spin L=iλ(ξ¯ξ˙−ξ˙¯ξ)+ 1(p x˙µ−p˙ xµ)−ξ¯β pµξ (12) µ µ µ 2 2 where β is the irreducible part of 1 β = γ ⊗I ⊗...⊗I+.2.s.+I⊗...⊗I ⊗γ (13) µ µ µ 2s and s is the spin label.(cid:0)The Euler Lagrange equations are: (cid:1) x˙µ =ξ¯βµξ, p˙ =0, iλξ˙=β·pξ, iλξ˙¯=−ξ¯β·p (14) µ In (12,14) λ is an arbitrary scalar constant with the dimensions of action. We have also taken the spinor momentum η = iλξ¯without loss of generality. Ob- serve how momentum and velocity are decoupled, the latter being fixed by the spinor degrees of freedom alone. The spin tensor, Pauli-Lubansky vector, and radius are λs Sµν =−λsξ¯βµνξ, rµ = ξ¯βµνξp , Wµ =λsξ¯β∗µνξp (15) p2 ν ν where βµν =is[βµ,βν]. After some algebra one arrives to the Poissonbrackets {W ,W } = ǫ pρWλ, {W ,r }=ǫ pρrλ, µ ν µνρλ µ ν µνρλ 1 {r ,r } = − ǫ pρWλ (16) µ ν p4 µνρλ 5 which show the close relation between these variables and the rotations and boosts of the Lorentz group. We can use the constant Wµ as giving the spin in the rest frame. Also, it is possible to give the spin tensor (8) in terms of p and r solely Sµν = pµrν −pνrµ −ip2{rµ,rν}. Observe that r can be made to vanish and still the Poisson Bracket {r,r} of (16) be finite. Therefore, spin is not a consequence of zitterbewegung. We may have no zitterbewegung (r =0), but finite spin (W 6= 0). There is also a loose relation between the product of bilinears and the symplectic structure; here we are freed of the problems [12] that other approaches have with the simultaneous treatment [13] of classical and -the would be- quantum variables. Thesolutionoftheequationsofmotion(14)willofcourseleadtoatrajectory of the form (6). We get it explicitly with 1 Xµ =xµ(0)−rµ(0), f(τ)= ξ¯β·pξτ, (17) p2 p2τ λs p2τ rµ(τ)=rµ(0)cos + r˙µ(0)sin (18) pλs ! p2 pλs ! where r˙µ(0) = dµν(p)ξ¯β ξ and rµ(0) is givpen by (15). All the spinors in (18) ν are given by their initial values at τ =0 and dµν(p) is the projector orthogonal to p. Equations (2,18) reveal the presence of zitterbewegung through a radius joining the instantaneous centre of the particle (X +fp) to its true position x. It is the centre which follows the external trajectory along p of a spinless particle. The radius rapidly rotates around the centre (with the frequency 2 p2c2/¯h = 1.5 1021s−1 for an electron and λ = 1) tracing an ellipse whose parameters depend solely of the spinorial degrees of freedom of the particle. p It is also worth to note here that the particle does not spend any energy nor momentum in keeping zitterbewegung on. One would also expect that classical external perturbations would act adiabatically on this internal rotation. The momentumwouldchangeaccordingto the appliedexternalforce asdictatedby the Newtonsecondlaw,butthe particlewouldcontinueto windateachinstant around the momentum to form a helix embracing the external trajectory. We will now show that the above picture is not consistent for charged par- ticles with spin. The reason is simply that zitterbewegung produces a ra- diation field that should carry fourmomentum away from the free particle!. We first introduce the electromagnetic interaction by the minimal substitu- tion rule p → p − eA in (12). Then we obtain the electromagnetic current jµ =eξ¯βµξ =euµ. We now recallthat we can use all the results of the electro- dynamicsofspinlessparticlesthatdonotuseofrelationsofthetypepµ =muµ. Our spinning free particle will radiate at a rate dpµ 2 rad =− e2(r¨ r¨µ)uµ (19) µ dτ 3 6 Somecommentsareinorderhere: First,radiationwillnotdrainfourmomentum off the particle (p˙ = 0). Second, radiation will occur even in the “rest system” (p~ = 0) of the charge. Third, radiation will not slow zitterbewegung, that proceeds at fixed frequency. The cure to these accumulation of catastrophes is to forbid the free particle to radiate at all. Being r¨ spacelike, this can be achieved only if r¨=0. Inspecting (18) we see that in this case r =0 also. This is the condition of parallelism that can be attained when the spinor is one of the eigenvectors of the operator β ·p. For instance, for s = 1/2, r = 0 implies γ ·pξ = ± p2ξ . It is consistent [14] to normalize ξ¯ ξ = ±1 and that ± ± ± ± when p0 < 0, ξ represents an antiparticle of momentum −p. Therefore, the p− conditionofnospontaneousemisionleadstofreeparticlesbeingeitherparticles or antiparticles, never a mixture of these. Second cuantization is a natural outcome of the above requisites. In addition, one recovers the old form of the Galileo principle. The switch on of the electromagnetic interaction begins to complicate this verysimplebehaviourofthefreespinningchargedparticle. Zitterbewegungwill start with its proper invariable frequency, but with a small growing amplitude as the external field feeds into a finite r (for instance, for s = 1/2 this will proceed by creating a finite ξ− (ξ+) component out of the initial ξ+ (ξ−)). Conversely, when exiting from the interaction region the particle, generally off the mass shell, will radiate till reaching the pure state. One may think that the description in terms of spinors is a too ellaborated one to deal with the spinning particle, at least in the free case. There may be a strong temptation to disentangle the radius from the Minkowski space coordinates. This could be achieved by performing a canonical transformation from x to z =x−r getting rid of the cumbersome vanishing radius. However it is easy to check that this new coordinate is not canonical; and the other way around: any canonical and covariant coordinate without the radius will acquire an imaginary part, which in addition turns out to be spin(-or) dependent. The lesson is clear: even if the kinematicsofthespinningparticletracesthatofthespinlesscase,thereare additional degrees of freedom present, and these have to be taken into account to implement the canonical formalism. In this sense, the study of spinning particles in ordinary (x,p) phase space is hopeless. The work of J. M. Mart´ın was supported by Fundaci´on Calvo Rod´es and a grant from the Comunidad Aut´onoma de Madrid. References [1] H. Corben, Classicaland Quantum Theories of Spinning Particles,Holden Day, Inc. San Francisco, 1968. [2] A. Proca, J. Phys. Rad. 15, 65 (1954). [3] R. Schiller, Phys. Rev. 128, 1402 (1962). 7 [4] H. A. Kramers,Verhandl. Zeeman Jubil. 403 (1935). [5] A. O. Barut and N. Zanghi, Phys. Rev. Lett. 52, 2009 (1984). [6] A. O. Barut and I. H. Duru, Phys. Rev. Lett. 53, 2355 (1984), and Phys. Rep. 172, 1 (1989). [7] V. Bargmann and E. P. Wigner, Proc. N. A. S. 34, 211 (1948). [8] P. A. M. Dirac, Lectures in Quantum Mechanics, Yeshiva University, New York, 1964. [9] S. Weinberg, Phys. Rev. 133, B1318 (1964). [10] S. Weinberg, Quantum Theory of Fields, Cambridge University Press, Cambridge, 1995. [11] L. Fadeev and R. Jackiw, Phys. Rev. Lett, 60, 1692 (1988). [12] A. Anderson, preprint quant-ph/9511014. [13] A. Anderson, Phys. Rev. Lett. 74, 621 (1995). [14] F. H. Gaioli and E. T. Garc´ıa Alvarez, Am. J. Phys. 63, 177 (1995). 8