Derivation of the magnetization current from the non-relativistic Pauli equation: A comment on “The quantum mechanical current of the Pauli equation” by Marek Nowakowski [Am. J. Phys. 67(10), 916-919 (1999)] M. S. Shikakhwa Physics Program, Middle East Technical University Northern Cyprus Campus, Kalkanlı, Gu¨zelyurt, TRNC, via Mersin 10, Turkey S. Turgut and N. K. Pak Department of Physics, Middle East Technical University, TR-06800, Ankara, Turkey 6 1 0 Some time ago, Nowakowski[1] presented a discussion of the fact that in the non-relativistic limit, the probability 2 current density J of a spin one-half particle contains an extra divergenceless term J M n a J = J0+JM J ¯h ¯h 2 = ψ†∇ψ−(∇ψ†)ψ + ∇×(ψ†σψ) . (1) 2mi 2m 2 (cid:0) (cid:1) whichhederivedbytakingthenon-relativisticlimitoftherelativisticDiracprobabilitycurrentdensity. Thederivation ] h for this additional term essentially relies on two assumptions: (1) In Dirac equation, the probability density of the p particle is given by ρ = ψ†ψ and (2) ρ is the time-component of a Lorentz covariant four-vector (ρ,J). Therefore, - these assumptions uniquely identify the probability current density. Taking the non-relativistic limit of this current t n produces Eq. (1). a Simply because of its existence, J term is important and hence it should be included in textbook discussions of M u the probability current of spin 1/2 particles. A nice discussion of such an additional current term with illustrative q examplesfromvariousquantummechanicalsystems waspublishedinthis journal[2](But, notethatthis workdefines [ the additional term from a different perspective and thus differs by a factor of 2 from the correct J ). M 1 NowakowskicorrectlystatesthattheadditionaltermJ cannotbederivedfromthenon-relativisticPauliequation, M v as the covariance argument can only be applied at the fully-relativistic Dirac equation level. Even though this is the 4 correct state of the affairs, it is still desirable to have an alternative derivation of this additional current term from 3 a non-relativistic “starting point”. If one is to derive the J term for undergraduate or junior graduate students 9 M 5 who have not yet been exposed to relativistic quantum mechanics, one needs to start from the Pauli equation. The 0 purpose of this comment is to point out that there is indeed an alternative derivationof this additionalterm starting . from the non-relativistic quantum mechanics of a spin 1/2 particle. 1 0 Our starting point is an alternative form of the Pauli Hamiltonian, namely 6 1 1 H = (σ·p)2 (2) : 2m v Xi where p = −i¯h∇ is the momentum operator and σi (i = 1,2,3) are the Pauli spin matrices. Using the well-known identity r a (σ·u)(σ·v)=u·v+iσ·(u×v) , (3) which can be easily derived from σ σ =δ I+i ǫ σ , (4) i j i,j ijk k Xk itcanbeseenthatEq.(2)isthesameasthe HamiltonianH =p2/2m. TheforminEq.(2),however,hasthe obvious advantage that, for a chargedparticle, in the presence of coupling to a vector potential A (so that p is replaced with π =p−(e/c)A), we have 1 H = (σ·π)2 (5) 2m 1 e¯h = π2− σ·B , (6) 2m 2mc 2 i.e.,theZeemantermintheHamiltonianisgeneratedautomaticallywiththecorrectg-factorofg =2(see,forexample Ref. 3), rather that being introduced by hand as a phenomenological term, as is usually done. This is intimately related with the fact that Eq. (2) is the first expression obtained for the Hamiltonian when the non-relativistic limit of the Dirac equation is taken, before simplifying it further into the originalPauli Hamiltonian. This teaches us that Eq. (2) is the fundamental non-relativistic Hamiltonian that one should start with; the form H = p2/2m is just a reduced special case. Coupled with these, if we start from the Hamiltonian in Eq. (2), and if we are careful in not canceling some terms, it is possible to derive the additional term J in the probability current density. Below, we M present an alternative derivation of the current density based on the conventional continuity equation. For the sake of completeness, we also sketch a second derivation based on Noether’s theorem. As the probabilitycurrentdensityis usually derivedfromthe continuity equation,itis importanttoshow thatthis approachalso produces J . Note that, since ∇·J =0, this term obviously does not have any contribution to the M M continuity equation, ∂(ψ†ψ)/∂t+∇·J = 0. For this reason, one has to be careful in not dropping some relevant terms. It is important to remember Nowakowski again: in a non-relativistic derivation, one cannot understand the presence,formor the coefficient ofsuchadditionalterms. However,we will showthat “the additionaltermis already there” before it is swept away (canceled) under the divergence operator. In this approach,it is veryuseful to consider the case where there is a vector potential A so that the momentum p are replacedwith π. Of course,for chargelessparticles (e.g., neutrinos), such a change is not physically allowed. Our mainpurpose for incorporatingsucha changeis to remindus notto commute differentcomponents of π. By keeping in mind that they do not commute, the operators will naturally guide us through the derivation. It will be seen that the vector potential will never be important in any stage of the derivation. At the end of the derivation, the vector potential can be set to zero if necessary. This is a reflection of the fact that the derivation is also valid for chargeless particles. Also, the Hamiltonian on the right-hand side of Eq. (2) may contain a scalar potential term. As it has no effectonthecurrentdensity,weomitsuchapotentialterminthefollowingderivation. Thisderivationstartswiththe conventionalconstruction of the continuity equation, namely by taking the time derivative of the probability density ρ=ψ†ψ, ∂ρ ∂ψ† ∂ψ = ψ+ψ† (7) ∂t ∂t ∂t 1 = − ((σ·π)2ψ)†ψ−ψ†(σ·π)2ψ (8) 2mi¯h (cid:16) (cid:17) 1 = − (π π ψ)†σ σ ψ−ψ†σ σ (π π ψ) (9) i j j i i j i j 2mi¯h Xi,j (cid:16) (cid:17) Itisimportanttonotethatthenon-commutativityofthekineticmomentaπ haveguidedustoexpressthetwoterms i as above. Otherwise, if the vector potential were set to zero at the beginning, we would have no rational reason for preferringp p ψ overp p ψ. We willnow try to expressthe right-handside as the divergenceofa currentby “pulling i j j i up” the derivative ∂/∂x to the front. The following identity is very useful for this purpose and, aside from factors i involving the vector potential, is obtained from the chain rule of differentiation ¯h ∂ α†(π β)−(π α)†β = α†β . (10) i i i ∂x i (cid:0) (cid:1) Inthe aboveequationαandβ arearbitrarytwo-componentspinors. Notethatthevectorpotentialdoesnotoccuron the right-hand side of the identity. (It is useful to think that this identity follows from the fact that π is hermitian: i as the integral of the left-hand side must be zero, the right-hand side must be a divergence.) Using this we get ∂ρ 1 ∂ = − (π ψ)†σ σ ψ+ψ†σ σ (π ψ) j j i i j j ∂t 2m ∂x Xi,j i(cid:16) (cid:17) 1 − (π ψ)†σ σ (π ψ)−(π ψ)†σ σ (π ψ) (11) j j i i i i j j 2mi¯h Xi,j (cid:16) (cid:17) Now, the second sum in the equation above gives zero (this can be seen easily by exchanging the labels i↔j for one of the summands). The first sum is in the desired divergence form and the probability-current vector can be read directly as 1 J = (π ψ)†σ σ ψ+ψ†σ σ (π ψ) ; (12) i j j i i j j 2m Xj 3 the continuity equation ∂ρ/∂t+∇·J =0 is then satisfied. What is left is the simplificationof the right-handside to the sum of the conventional and the magnetization current. Using Eq. (4) we get 1 i J = (π ψ)†ψ+ψ†(π ψ) + ǫ −(π ψ)†σ ψ+ψ†σ (π ψ) (13) i i i ijk j k k j 2m 2m (cid:16) (cid:17) Xj,k (cid:16) (cid:17) i ¯h ∂ = J0i+ ǫijk ψ†σkψ (14) 2m i ∂x Xj,k j(cid:16) (cid:17) ¯h = J0i+ 2m ∇×ψ†σψ i =J0i+JMi . (15) (cid:0) (cid:1) Here, we have used the identity in Eq. (10) again for simplifying the magnetization current term. Note that the correct expression of J0 in the presence of a vector potential is as above (i.e., J is the real part of ψ†πψ/m) which basically reduces to the familiar form when A is set to zero. Note also that, at no point in the derivation above the vector potential A appears explicitly. It is only implicitly present by reminding us that we should not commute products π π . i j Note that we have not simplified the products of Pauli matrices until the last point. At Eq. (11), when we have seen that the last sum is zero, we have stopped and identified the current density as (12). It is still possible to not stop at Eq. (11), continue simplifying the first sum in this equation and drop the vanishing terms corresponding to ∇·JM = 0. In such a path, one can obtain unsurprisingly only the J0 term of the current. Of course, these kinds of ambiguities are expected as we are not following the only correct methodology (i.e., using covariance argument in Diracequation). Despite this, the veryfactthatonecanfindthe correcttermJ by not canceling anobviouslyzero M term in the continuity equation, gives some credence to the current approach. The additional current term can also be shown to be a part of the conserved Noether current[4] that follows from the invariance of the non-relativistic Pauli Lagrangianunder the U(1) global phase transformation. In this case, the presence or the absence of the vector potential does not change the derivation. For this reason we take A=0, since a non-zero vector potential does not change the derivation. The Lagrangiandensity is given by i¯h 1 L= ψ†ψ˙ −ψ˙†ψ − (σ·pψ)†(σ·pψ) , (16) 2 2m (cid:16) (cid:17) with the Euler-Lagrangeequations givingEq.(2). Note that, uponsimplificationofthe spin matrices,the Lagrangian density can be written as L=L0+LM where 2 i¯h ¯h L0 = ψ†ψ˙ −ψ˙†ψ − ∂iψ†∂iψ , (17) 2 2m (cid:16) (cid:17) Xi 2 i¯h L = − ǫ ∂ ψ†σ ∂ ψ . (18) M ijk i k j 2m Xijk The last term, L , can be brought into the form of a total divergence M 2 i¯h L = ∇·(ψ†σ×∇ψ) , (19) M 2m and hence its contribution to the action I = Ld4x can be converted into a surface integral. Because of this reason, its presence in the Lagrangian density doesRnot affect the equations of motion. However, if this term is kept, one derives the missing magnetization current term here too. (Note that, if there is magnetic field, then L is not equal M to a divergence. In fact, L contains the Zeeman term in that case.) M Now, consider the global phase transformation i ψ −→e−iα/h¯ψ ≈ 1− δα ψ (20) (cid:18) ¯h (cid:19) whereδα is aninfinitesimalrealnumber independent oftime andspace. The Lagrangiandensity Lis invariantunder this transformation. The corresponding conserved Noether current is given by ∂L ∂L Jµδα=δψ† + δψ (µ=0,1,2,3) . (21) ∂(∂ ψ†) ∂(∂ ψ) µ µ 4 It is then straightforward to see that the time component (µ = 0), J0 = ψ†ψ, is the probability density ρ and the space components produce J =J0+JM. The additional term JM comes from the term LM. Taking this opportunity, we would like to comment on some points related to the interpretation of the current J . It is possible to interpret this term as the effective current density associated with magnetization in classical M electrodynamics[5]. In other words, if the particle is a charged particle with charge e, then its magnetization density is given by M =(e¯h/2mc)ψ†σψ and the associated magnetization current is c∇×M =eJ . In fact, Landau and M Lifshitz have derived this term by taking the functional derivative of the (average)energy with respect to the vector potential[6]. This approach essentially identifies the current density J as the “coupling strength” of the particle to the electromagnetic field. As the internal magnetic moment associated with the spin creates a magnetic field, it is necessary that a spin term is also present in J. Although the argument above for the interpretation of the additional J term seems consistent, it is definitely M incomplete,asthesametermisalsopresentforchargelessspin1/2particlewhichdonotcoupletotheelectromagnetic field, (e.g., neutrinos) as well. For chargeless particles, the additional term is still present. This also explains “the coincidence” that the additive term looks like a magnetization current: For chargedparticles, the charge density and the probability density are directly proportional to each other. By relativistic covariance, the associated currents should also be directly proportional to each other. Therefore, if the charge current contains a term related to the magnetization, then, there should also be such a term in the probability current. Yet, the originof the additive term is not electromagnetism as it is also present for particles that do not interact electromagnetically. [1] Marek Nowakowski, “The quantum mechanical current of thePauli equation”, Am.J. Phys.67, 916-919 (1999). [2] Katsunnori Mita, “Virtual probability current associated with the spin” Am.J. Phys. 68, 259-264 (2000). [3] J. J. Sakurai, Advanced Quantum Mechanics, (Addison Wesley,London, 1967), p. 78. [4] M. E. Peskin and D.V. Schroeder, An Introduction to Quantum Field Theory, (Addison-Wesley,New York,1995), p. 17. [5] J. D.Jackson, Classical Electrodynamics, 2nd ed. (John Wiley and Sons, New York,1975), p.188. [6] L.D.Landau,E.M.Lifshitz,Quantum Mechanics,vol.3ofCourseofTheoreticalPhysics,(PergamonPress,Oxford,1991), pp. 472.