Exercise: The Non-Relativistic Limit and the Pauli Equation

Introduction

This is one of the worked exercises attached to the article The Dirac Equation in Biquaternionic Form — Solutions and the Non-Relativistic Limit. That article states the two forms of the biquaternion Dirac equation, constructs the plane-wave solutions of the one-sided pair, and carries the pair into the non-relativistic regime far enough to exhibit the Pauli equation and the gyromagnetic factor $g=2$. This exercise takes the limiting procedure as given and executes it in full: every order that contributes to the first corrections is kept, every coefficient is derived, and every claim is checked.

Five problems are worked:

  1. the Foldy–Wouthuysen elimination of the small component, carried out step by step and to the order at which the first corrections appear;
  2. the minimal substitution $p\to p-q\vec A$ and the electromagnetic terms it produces, including the spin term and the coefficient that fixes $g=2$;
  3. the next order in $1/c$: the relativistic kinetic correction, the Darwin term, and the spin–orbit coupling;
  4. the Hermiticity of every term, and which terms lie in the Hermitian subspace $\mathbb{B}_{\Psi}$ and which do not;
  5. the physical dimensions of every term.

The operators, fixed once. Every derivative in this exercise is the biquaternionic gradient

$$ \tilde\nabla_\Phi=\sum_{\mu\in\{t,x,y,z\}}e_\mu\frac{\partial}{\partial R_\mu}, \qquad (\tilde\nabla_\Phi)^{\natural}=\sum_{\mu\in\{t,x,y,z\}}\varepsilon_\mu e_\mu\frac{\partial}{\partial R_\mu}, \qquad \varepsilon=(1,-1,-1,-1), $$

and its norm is the d'Alembertian

$$ \Box_\Phi=\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}=(\tilde\nabla_\Phi)^{\natural} \tilde\nabla_\Phi=\frac{\partial^2}{\partial R_t^2}+\Delta_\Phi . $$

On the material sector, where $Q_t=i q'_t$, the gradient and its conjugate read

$$ \tilde\nabla_{\Phi}=-i\,e_t\frac{\partial}{\partial r'_t}+\sum_\eta\vec e_\eta\frac{\partial}{\partial r_\eta}, \qquad (\tilde\nabla_\Phi)^{\natural}=-i\,e_t\frac{\partial}{\partial r'_t}-\sum_\eta\vec e_\eta\frac{\partial}{\partial r_\eta}, \qquad \Box_\Phi=\Delta_\Phi-\frac{\partial^2}{\partial {r'_t}^2}, $$

and the pure part of the gradient is written $D_3=\sum_\eta\vec e_\eta\partial/\partial r_\eta$. Minimal coupling shifts each derivative, $\partial/\partial R_\mu\to\partial/\partial R_\mu-\frac{iqA_{\mu}}{\hbar}$, and with it the gradient:

$$ \tilde\nabla_\Phi\longmapsto\tilde\nabla_\Phi-\frac{i\,q}{\hbar}\tilde A_{\Phi}, \qquad \tilde A_{\Phi}=Ξ e_t-\vec A, \qquad \vec A=\sum_\eta A_\eta\vec e_\eta, \qquad \sum_\eta\vec e_\eta\pi_\eta=-i\,\hbar\,D_{3}^{A}, $$

with $\pi_\eta=p_\eta-qA_\eta$ and $D_3^A=\sum_\eta\vec e_\eta(\partial/\partial r_\eta-\frac{iqA_{\eta}}{\hbar})$ the pure part of the gauged gradient. The corpus names this family $\tilde\nabla_{\mathrm{gco},\Phi}$ in The Covariant Derivative and the Gauge Connection in Biquaternionic Form, where the potential is written $i\,a'_t\,e_t+\vec A$; the exercise keeps its own $\tilde A_{\Phi}=Ξ e_t-\vec A$ and writes the operator out rather than borrowing the name. A plane wave $\exp\!\bigl(i(k_{t}^{\prime}x_0-\vec k_\Phi\cdot\vec q)\bigr)$, with $x_0=r'_{t}$, replaces the gradient by the element

$$ -i\tilde K_{\Phi}=k_{t}^{\prime}-i\vec k_\Phi, \qquad -i\tilde K_{\Phi}f=0\ \text{on the light cone}, \qquad (-i\tilde K_{\Phi})^{\natural}(-i\tilde K_{\Phi})=k_{t}^{\prime\,2}-\lvert\vec k_\Phi\rvert^2 . $$

What the gradient can do, and what it cannot. The massless equation is one first-order equation in one field,

$$ \tilde\nabla_\Phi F=0 . $$

The massive case is not of that form, and the reason is a square. The square of the gradient is not scalar: the time unit commutes with the three spatial units instead of anticommuting with them, so the mixed derivatives add, and each spatial unit squares to $-e_t$, which puts the wrong sign on the Laplacian. Explicitly, on a general field,

$$ \tilde\nabla_\Phi^2=\Box_\Phi-2\Delta_\Phi-2i\sum_\eta\vec e_\eta\frac{\partial^2}{\partial r'_t\,\partial r_\eta}, $$

so $\tilde\nabla_\Phi^2\neq\Box_\Phi$ and no scalar mass can be subtracted from the square. The massive equation is therefore a pair of first-order equations in two fields, the chiral pair, and it is stated below. No one-field first-order massive equation is written anywhere in this exercise.

On the parent article. The parent's non-relativistic section works only to leading order. It factors out the rest energy, eliminates the small component with the approximation $\partial/\partial r'_t\,\tilde{Ξ}_{\mathrm{odd}}\approx 0$, and reads off the Pauli Hamiltonian; it stops there. It contains no spin–orbit and no Darwin term, and it does not perform a systematic Foldy–Wouthuysen transformation. It also does not state the sector ($\mathbb{B}_{\Psi}$ and $\mathbb{B}_{\Phi}$) status of the intermediate operators it uses. Problems 3 and 4 supply those orders and that status. Where the parent is silent this is stated explicitly and not silently filled in.

All the derivatives below are $\tilde\nabla_\Phi$, $(\tilde\nabla_\Phi)^{\natural}$ and their pure part $D_3$, and no other differential operator appears. Every display carries $\hbar$ where a mass is meant and $c$ where a rate meets a length; nothing is set to one. The basis operator that realises the split is the right multiplication by $-i\vec e_z$; it takes the two values $\mp1$, and its side is the right slot; the field is one element $F\in\mathbb B$ split into its two parts, $F=F_L+F_R$, of value $-1$ and $+1$. The side is position: the chirality acts as $F\mapsto-iF\vec e_z$, and the left action $F\mapsto-i\vec e_zF$ is a different element, never used here. Since $\vec e_z=-\omega$ with $\omega$ the pseudoscalar of the frame, $\omega^2=-e_t$, the chirality is also written $F\mapsto iF\omega$; the $i$ in front of it is the central scalar $i$ of the algebra and not $\omega$.

The two Dirac equations

The parent carries two equations, and they must not be read as one.

The one-sided pair. The corpus's own equation is the chiral pair. It acts on the two chiral halves $F_R$ and $F_L$ of the field by left multiplication, and the mass is the off-diagonal pairing of the two halves:

$$ \tilde\nabla_\Phi F_R=mF_L, $$

$$ (\tilde\nabla_\Phi)^{\natural}F_L=mF_R . $$

Both gradients act on the left, and no right multiplication and no conjugation enters. On the Dirac module $\mathcal{S}_{\mathrm{rep}}=\mathbb B$ the pair is a four-component equation, and it is the equation reduced below.

The two-sided form. Kravchenko's dictionary writes the same physics as a single equation on the whole algebra. The derivative acts on the left, the mass acts on the right through $\vec e_z=-\omega$, and a componentwise conjugation supplies the antilinear part. Here $D_3=\sum_\eta\vec e_\eta\,\partial/\partial R_\eta$ is the pure part of the gradient and $\mathcal N$ names the left-hand side. The operator reads

$$ \mathcal N F=\Bigl(i\frac{\partial}{\partial r'_t}+D_3\Bigr)F-k_mi\,\overline F\,\vec e_z=0 . $$

Equivalently,

$$ -(\tilde\nabla_\Phi)^{\natural}F-mi\,\overline F\,\vec e_z=0 $$

for a field $F\in\mathbb B$, since $i\partial/\partial r'_t=-\partial/\partial R_t$ in these units. The two forms are related by a change of representation, but they are different operators: the pair is linear in the field, while the mass term of the two-sided form is antilinear because it carries $\overline F$.

| | One-sided pair | Two-sided form | |+-|+-|+-| | Equation | $\tilde\nabla_\Phi F_R=mF_L$, $(\tilde\nabla_\Phi)^{\natural}F_L=mF_R$ | $\mathcal NF=0$, with $\mathcal NF=(i\partial/\partial r'_t+D_3)F-mi\overline F\vec e_z$ | | Multiplication | left by $\tilde\nabla_\Phi$ and $(\tilde\nabla_\Phi)^{\natural}$ | left by $i\partial/\partial r'_t$ and $D_3$; right by $\vec e_z$; antilinear bar | | Field | $F=F_L+F_R\in\mathcal{S}_{\mathrm{rep}}=\mathbb B$ | $F\in\mathbb B$ | | Mass | off-diagonal pairing of the chiral halves | $-mi\overline F\vec e_z$: a conjugation and a direction | | Chirality | the central sign, $\mp1$ on the two halves | the right multiplication by $-i\vec e_z$ | | Use | the formulation and its kinematics | the dictionary, the Cauchy theory, the integral representation |

The reduction below is of the one-sided pair, read on the Dirac module. The Foldy–Wouthuysen transformation is applied to the Hamiltonian of that pair. The two-sided form is a different operator, antilinear in its mass term; its non-relativistic reduction is not attempted here.

The Hamiltonian form

The two gradings. The reduction needs two involutions of the module, and both are right multiplications of the algebra, written by position.

  • The chirality $\Gamma F=-iF\vec e_z$, of values $\mp1$ on $F_L$ and $F_R$. It separates the two members of the pair, and it commutes with every left multiplication, hence with the gradient.
  • The mass grading $\beta F=iF\vec e_x$. It squares to the identity, $\beta^2=1$; it anticommutes with the chirality, $\beta\Gamma=-\Gamma\beta$; and it commutes with the gradient. Its two values split the module into a large part and a small part.

The Hamiltonian. The pair, read on the module, is the equation of motion

$$i\,\hbar\,c\frac{\partial F}{\partial r'_t}=HF, \qquad H=\beta mc^2+O+qΞ, \qquad OF=c\sum_\eta\vec e_\eta\pi_\eta\,F\,\vec e_y=-i\,\hbar\,c\bigl(D_3^AF\bigr)\vec e_y . $$

The mass term is the grading itself, multiplied by the mass. The kinetic operator $O$ is the pure part of the gauged gradient, multiplied on the right by $\vec e_y$. Both factors are present and both are needed: left multiplication by $\vec e_\eta$ with the derivative on one side, and the multiplier $\vec e_y$ on the other, so $O$ is a two-sided operator and not the left multiplication by an element. It is odd because $\vec e_y$ anticommutes with $\vec e_x$; it reverses the chirality because $\vec e_y$ anticommutes with $\vec e_z$ as well. The relations used below are four:

$$ \beta^2=1, \qquad \beta O=-O\beta, \qquad \beta\,qΞ=qΞ\,\beta, \qquad O=O^{\dagger},\quad \beta=\beta^{\dagger}, $$

so that $O$ is odd, $qΞ$ is even, and $H$ is Hermitian. The eigenvalues of $H$ at zero potential are $\pm\sqrt{m^2c^4+c^2\vec\pi^2}$, which is the mass shell $(-i\tilde K_{\Phi})^{\natural}(-i\tilde K_{\Phi})=k_m^{2}$ of the pair.

The two halves. With $\beta\tilde{Ξ}_{\mathrm{evn}}=+\tilde{Ξ}_{\mathrm{evn}}$ and $\beta\tilde{Ξ}_{\mathrm{odd}}=-\tilde{Ξ}_{\mathrm{odd}}$, the equation of motion reads on the two parts

$$i\,\hbar\frac{\partial\tilde{Ξ}_{\mathrm{evn}}}{\partial r'_t}=mc^2\tilde{Ξ}_{\mathrm{evn}}+O\tilde{Ξ}_{\mathrm{odd}}+qΞ\tilde{Ξ}_{\mathrm{evn}},i\,\hbar\frac{\partial\tilde{Ξ}_{\mathrm{odd}}}{\partial r'_t}=-mc^2\tilde{Ξ}_{\mathrm{odd}}+O\tilde{Ξ}_{\mathrm{evn}}+qΞ\tilde{Ξ}_{\mathrm{odd}} . $$

The even operator $qΞ$ is the electrostatic energy and the odd operator $O$ is what couples the large and the small parts: $O\tilde{Ξ}_{\mathrm{evn}}$ is small and $O\tilde{Ξ}_{\mathrm{odd}}$ is large. This is the system reduced in Problem 1.

Problem 1: The Foldy–Wouthuysen elimination of the small component

Statement. Starting from the four-component Dirac equation with minimal coupling, eliminate the small component and obtain the Pauli Hamiltonian, keeping the orders through the first corrections.

Solution.

Step 1: the split. The Hamiltonian of the pair is $H=\beta mc^2+O+E$ with

$$ O=c\sum_\eta\vec e_\eta\pi_\eta\,(\cdot)\,\vec e_y\quad(\text{odd},\ \beta O\beta=-O), \qquad E=qΞ\quad(\text{even},\ \beta E\beta=E), $$

and the odd operator is the pure part of the gauged gradient, $\sum_\eta\vec e_\eta\pi_\eta=-i\,\hbar\,D_{3}^{A}$, multiplied on the right by $\vec e_y$. Writing it as $\hbar\,c\,\vec \pi$, that is as the left multiplication by an element, would make it even; the right multiplier $\vec e_y$ is what makes it odd, and it is also what reverses the chirality.

Step 2: factoring out the rest energy. Write $F=\exp(-i\theta_{\mathrm{rst}})\,G$ with $\theta_{\mathrm{rst}}=r'_t\,k_m$. The factor is the central element $\cos\theta_{\mathrm{rst}}-i\sin\theta_{\mathrm{rst}}$, which commutes with every operator of the reduction, so it can be cancelled on both sides. What is left of the mass term is $(\beta-1)mc^2$, which vanishes on the large part and is $-2mc^2$ on the small part, and the parent's two coupled equations follow:

$$i\,\hbar\frac{\partial\tilde{Ξ}_{\mathrm{evn}}}{\partial r'_t}=O\tilde{Ξ}_{\mathrm{odd}}+qΞ\tilde{Ξ}_{\mathrm{evn}}, $$

$$i\,\hbar\frac{\partial\tilde{Ξ}_{\mathrm{odd}}}{\partial r'_t}=O\tilde{Ξ}_{\mathrm{evn}}-2mc^2\tilde{Ξ}_{\mathrm{odd}}+qΞ\tilde{Ξ}_{\mathrm{odd}} . $$

Why the rest energy comes out first. The $-2mc^2$ in the small-part equation is the rest energy already added in, and its presence is what makes the elimination converge. Solving that equation exactly,

$$ \tilde{Ξ}_{\mathrm{odd}}=\frac{1}{i\,\hbar\partial_{r'_t}+2mc^2-qΞ}\,O\tilde{Ξ}_{\mathrm{evn}} , $$

the denominator is $2mc^2+\varepsilon-qΞ$ on a field whose temporal eigenvalue is the kinetic energy $\varepsilon=E-mc^2$, so the expansion parameter is $(\varepsilon-qΞ)/(2mc^2)$; it vanishes at rest and is small exactly in the non-relativistic regime. Remove the phase after the elimination instead of before, and the same computation divides by $i\,\hbar\partial_{r'_t}+mc^2$, whose eigenvalue is $E\approx mc^2$ and not $\varepsilon$; the parameter is then of order one, the series does not converge, and its first term overshoots by a factor of two. The symptom is unambiguous. The small part comes out as $\sigma\cdot\vec\pi\,\tilde{Ξ}_{\mathrm{evn}}/(mc)$ instead of $/(2mc)$, which turns the leading term into $\vec p^{\,2}/m$ rather than $\vec p^{\,2}/(2m)$. Removing the phase after the expansion instead of before does not repair it either: that order lands on $+\vec p^{\,2}/(4m)$, half the correct kinetic term, an effective mass of $2m$. Neither endpoint is repaired by a local redefinition. A scalar factor, phase or amplitude, multiplies both terms of a differential operator by the same function and leaves the coefficient of $\vec p^{\,2}$ alone, and a momentum-dependent redefinition $1+\alpha\vec p^{\,2}/(m^{2}c^{2})$ changes the kinetic term only at relative order $c^{-2}$, so it cannot undo a factor of two at order $c^{0}$. The order of the two operations is therefore not a matter of taste: the phase first, the elimination second, and the coefficient $1/2m$, with the $g=2$ that it carries, is a consequence of that order.

Step 3: the leading elimination. The term $2mc^2\tilde{Ξ}_{\mathrm{odd}}$ dominates the second equation whenever the fields, the kinetic energy and $\partial\tilde{Ξ}_{\mathrm{odd}}/\partial r'_t$ are small compared with $mc^2$. Neglecting the other terms there,

$$ \tilde{Ξ}_{\mathrm{odd}}=\frac{O}{2mc^2}\,\tilde{Ξ}_{\mathrm{evn}}+O(c^{-3}), $$

so the small part is of relative order $v/c$, as it must be. Substituting into the first equation eliminates $\tilde{Ξ}_{\mathrm{odd}}$ exactly at this order:

$$i\,\hbar\frac{\partial\tilde{Ξ}_{\mathrm{evn}}}{\partial r'_t}=\left[\frac{O^2}{2mc^2}+qΞ\right]\tilde{Ξ}_{\mathrm{evn}} , $$

and on the large part $O^2/(2mc^2)=(\vec\pi^2-\hbar\,q\,\boldsymbol{\sigma}\,\vec B)/2m$, whose leading term is $\vec\pi^2 c/(2\,m\,c)$ and whose remainder is evaluated in Problem 2.

Step 4: the systematic transformation. The approximation of Step 3 is the first term of a controlled expansion. The Foldy–Wouthuysen generator is $S=S_1+S_2+\cdots$ with

$$ S_1=-\frac{i}{2\,m\,c^{2}}\,\beta O =-\frac{1}{2\,m\,c}\sum_\eta\vec e_\eta\pi_\eta\,(\cdot)\,\vec e_z =\frac{i}{2\,k_m}\bigl(D_3^A\cdot\bigr)\vec e_z . $$

The second form is the first one after $\beta O=-ic\sum_\eta\vec e_\eta\pi_\eta(\cdot)\vec e_z$, the two multipliers of $\beta O$ being $i\vec e_x$ and $\vec e_y$, whose product is $i\vec e_z$; the third is the same operator written with the pure part of the gauged gradient, since $\sum_\eta\vec e_\eta\pi_\eta=-i\,\hbar\,D_{3}^{A}$. The generator is two-sided as well, left by $\vec e_\eta$ and right by $\vec e_z$, and it is Hermitian, being the product of an anti-Hermitian left factor and an anti-Hermitian right multiplier with a Hermitian $\pi_\eta$ between them; so $iS_1$ is anti-Hermitian and the transformation is unitary. For static fields the generator is time-independent, and the transformed Hamiltonian is

$$ e^{iS_1}He^{-iS_1} =H+i[S_1,H]-\tfrac12[S_1,[S_1,H]]-\tfrac{i}{6}[S_1,[S_1,[S_1,H]]]+\cdots $$

the full transformed operator being this expression minus $i\,\hbar\,c\partial/\partial r'_t$. Using $\beta O=-O\beta$ and $\beta E=E\beta$, the three elementary commutators are

$$ i[S_1,\beta mc^2]=-O,\qquad i[S_1,O]=\frac{\beta O^2}{mc^2},\qquad [S_1,[S_1,\beta mc^2]]=\frac{\beta O^2}{mc^2}. $$

The first cancels the odd term $O$. The second contributes $+\frac{\beta O^2}{mc^2}$ to the expansion and the third contributes $-\frac12$ of the same, so together they give $+\frac{\beta O^2}{2mc^2}$, the leading non-relativistic energy. There is also a residual odd operator, of relative order $(v/c)^2$ compared with $O$ for a Coulomb field,

$$ O_1=i[S_1,E] =\frac{q}{2\,k_m}\sum_\eta\vec e_\eta E_\eta\,(\cdot)\,\vec e_z =\frac{q}{2\,k_m}\,(\vec E\,\cdot)\,\vec e_z , $$

which the second transformation $S_2=-\frac{i}{2\,m\,c^{2}}\beta O_1$ removes. To the order at which the first physical corrections appear, therefore,

$$ H_1=\beta mc^2+qΞ+\frac{\beta O^2}{2mc^2}+O_1+O(c^{-2}). $$

The omitted even term of order $c^{-2}$ is the second-order term $-\tfrac12[S_1,[S_1,E]]$; it is evaluated in Problem 3.

Step 5: evaluating $O^2$. Each term of $O$ ends in the right multiplication by $\vec e_y$, and $\vec e_y\vec e_y=-e_t$, so the two multipliers collapse to a scalar and $O^2$ is a left multiplication by an element:

$$ O^2=c^2\sum_{j,\eta}\vec e_j\vec e_\eta\,\pi_j\pi_\eta\,\vec e_y\vec e_y =-c^2\sum_{j,\eta}\vec e_j\vec e_\eta\,\pi_j\pi_\eta . $$

The product identity of the algebra, $\vec e_j\vec e_\eta=-\delta_{j\eta}+\epsilon_{j\eta l}\vec e_l$, splits the sum into a symmetric and an antisymmetric part. The symmetric part is $-\vec\pi^2$. The antisymmetric part is the commutator, the ordering being kept since $\vec p$ and $\vec A$ do not commute: with $[\pi_j,\pi_\eta]=i\,\hbar\,q\,\epsilon_{j\eta l}\,B_{l}$ and $\sum_{j<\eta}\epsilon_{j\eta l}\vec e_j\vec e_\eta=\vec e_l$, it gives $i\,\hbar\,q\,\vec B$. Hence

$$ O^2=c^2\left(\vec\pi^2-i\,\hbar\,q\,\vec B\right), \qquad \vec B=\sum_\eta B_\eta\vec e_\eta . $$

This is a left multiplication by an element of the algebra, so it is even, and it commutes with $\beta$. On the large part the element $-i\,\hbar\,q\,\vec B$ acts as the matrix $-\hbar\,q\,\boldsymbol{\sigma}\,\vec B$, under the isomorphism $\mathsf M_2(\vec e_\eta)=-i\sigma_\eta$, so

$$ O^2_{\rm large}=c^2\left(\vec\pi^2-\hbar\,q\,\boldsymbol{\sigma}\,\vec B\right), \qquad H_{\rm large}=mc^2+qΞ+\frac{\vec \pi^{2}}{2\,m}-\frac{qc}{2\,k_m}\boldsymbol\sigma\cdot\vec B+O(c^{-2}). $$

Here $k_m=mc/\hbar$ is the mass wave number, the inverse of the mass length.

Subtracting the rest energy,

$$ \boxed{\;i\,\hbar\frac{\partial\tilde{Ξ}_{\mathrm{evn}}}{\partial r'_t}=\left[\frac{(p-q\vec A)^{2}}{2\,m}+qΞ-\frac{\hbar\,q}{2\,m}\,\boldsymbol\sigma\cdot\vec B\right]\tilde{Ξ}_{\mathrm{evn}}\;} $$

which is the Pauli equation exactly as the parent reports it. The coefficient $1/2m$ is the whole content of the $g=2$ result of Problem 2: it is produced by the $2mc^2$ in the denominator of the small-component elimination, not put in by hand.

Problem 2: Minimal substitution, the electromagnetic terms, and $g=2$

Statement. Carry out the minimal substitution, evaluate the electromagnetic terms it generates, and verify the gyromagnetic factor $g=2$ together with the biquaternion form of the spin term.

Solution.

Step 1: the substitution. The minimal coupling of this exercise is a shift of every derivative, $\partial/\partial R_\mu\to\partial/\partial R_\mu-\frac{iqA_{\mu}}{\hbar}$; on the gradient it is $\tilde\nabla_\Phi\to\tilde\nabla_\Phi-\frac{i\,q\,\tilde{A}_{\Phi}}{\hbar}$, with $A_\mu=(Ξ,-\vec A)$ in this mostly-minus metric, so that $\tilde A_{\Phi}=Ξ e_t-\vec A$. The time component gives

$$i\,\hbar\,c\frac{\partial}{\partial r'_t}\ \longrightarrow\ i\,\hbar\,c\frac{\partial}{\partial r'_t}-qΞ , $$

and the spatial components give $\vec\pi=p-q\vec A$, that is the pure part of the gradient becomes $D_3^A$ with $\sum_\eta\vec e_\eta\pi_\eta=-i\,\hbar\,D_{3}^{A}$. The electrostatic energy is collected in $V=qΞ$. The sign of the coupling is tied to the metric and to the convention for $A_\mu$; the mostly-plus convention often writes the opposite sign. Within this article the statement above is used throughout.

Step 2: the commutator of the kinetic momenta. With $p=-i\,\hbar\nabla$ and $\vec\nabla_\Phi=\sum_\eta\vec e_\eta\partial/\partial r_\eta$ the pure part of the gradient,

$$ [\pi_i,\pi_j]=-q\bigl([p_i,A_j]+[A_i,p_j]\bigr) =-q\bigl(-i\,\hbar\partial/\partial r_i\,A_j+i\,\hbar\partial/\partial r_j\,A_i\bigr) =i\,\hbar\,q\,\epsilon_{ij\eta}\,B_{\eta}, $$

so that

$$ \vec\pi\times\vec\pi=i\,\hbar\,q\,\vec B,\qquad \vec B=\vec\nabla_\Phi\times\vec A . $$

Equivalently, $p\times\vec A+\vec A\times p=-i\,\hbar\,\vec B$; the two terms are individually non-Hermitian but their sum is Hermitian. This is the parent's $\vec\pi\times\vec\pi=iq\vec B$, with $q$ the charge and the coupling $\frac{ϙ}{\hbar}$ carried in $\vec\pi$. The curl here is the curl of the corpus's pure gradient: for $\vec A=\sum_\eta A_\eta\vec e_\eta$ the product $D_3\vec A$ is $-\vec\nabla_\Phi\cdot\vec A$ in its scalar part and $\vec\nabla_\Phi\times\vec A$ in its pure part, with no reversed sign, so $B_\eta=\epsilon_{\eta ji}\partial/\partial r_jA_i$ is the Cartesian curl and no sign of this exercise had to be changed on that account.

Step 3: the square of the gradient's pure part. The product identity of the algebra is the product of two pure elements,

$$ \Bigl(\sum_\eta\vec e_\eta a_\eta\Bigr)\Bigl(\sum_\eta\vec e_\eta b_\eta\Bigr) =-\vec a\cdot\vec b\,e_t+\sum_l\epsilon_{lj\eta}a_jb_k\,\vec e_l , $$

which is $-\vec a\cdot\vec b-i\,\boldsymbol\sigma\cdot(\vec a\times\vec b)$ in the matrix reading, and it is valid whenever the coefficients commute with the units, which they do. Applying it to the two factors of $O$ and keeping the ordering, the commutator of Step 2 gives

$$ O^2=c^2\bigl(\vec\pi^2-i\,\hbar\,q\,\vec B\bigr) \ \longleftrightarrow\ c^2\bigl(\vec\pi^2-\hbar\,q\,\boldsymbol{\sigma}\,\vec B\bigr) $$

on the large part, which is the parent's result, with $\hbar$ restored. The coefficient carries no power of $c$: the two factors of $c$ in the elimination cancel, since $\tilde{Ξ}_{\mathrm{odd}}=O\tilde{Ξ}_{\mathrm{evn}}/(2mc^2)$ is multiplied by $O$ in the first equation. The identification $\vec S=\frac{\hbar}{2}\boldsymbol\sigma$ gives the spin term

$$ -\frac{\hbar\,q}{2\,m}\,\boldsymbol\sigma\cdot\vec B=-\frac{q}{m}\,\vec S\cdot\vec B=-\vec\mu\cdot\vec B . $$

Step 4: the gyromagnetic factor. Writing the coupling as $-\vec\mu\cdot\vec B$ identifies

$$ \vec\mu=\frac{\hbar\,q}{2\,m}\boldsymbol\sigma=\frac{q}{m}\,\vec S,\qquad \vec S=\frac{\hbar}{2}\boldsymbol\sigma . $$

The general parametrization of a magnetic dipole is $\vec\mu=g\dfrac{q}{2m}\vec S$, so that

$$ \boxed{\;g=2\;} $$

for a structureless spin-$\tfrac12$ particle. For the electron, $q=-e$ and

$$ \vec\mu_e=-\frac{e}{m_e}\vec S=-g\frac{e}{2m_e}\vec S,\qquad g=2, $$

whose magnitude at $S_z=\frac{\hbar}{2}$ is one Bohr magneton $\mu_B=ec\dfrac{1}{2\,k_{m_e}}$. The measured anomalous part $a=(g-2)/2\approx\alpha/2\pi$ is a radiative correction that lies outside the equation treated here.

Here $1/k_{m_e}=\hbar/(m_ec)$ is the electron mass length.

Step 5: the biquaternion form. Under the isomorphism $\mathsf M_2(\vec e_\eta)=-i\sigma_\eta$ the spin components are the elements $\tilde S_{\Psi,\eta}=\frac{\hbar}{2}i\vec e_\eta$, and a magnetic field written as the pure real quaternion $\vec B=\sum_\eta B_\eta\vec e_\eta$ maps the spin term to

$$ -\frac{\hbar\,q}{2\,m}\boldsymbol\sigma\cdot\vec B\ \longleftrightarrow\ -\frac{\hbar\,q}{2\,m}\,i\vec B\ \in\ \mathbb{B}_{\Psi} , $$

the Hermitian element of the informational sector, exactly as the parent states: $(i\vec B)^{*}=i\vec B$, because $(\vec e_\eta)^{*}=-\vec e_\eta$ and the coefficient is imaginary. The kinetic and electrostatic terms are the scalar multiples $(\vec\pi^2 c/(2\,m\,c))e_t$ and $qΞ e_t$ of $\mathbb{B}_{\Psi}$.

Numerical check. The commutator identity of Step 2 was verified by finite differences on a $41\times41$ grid with spacing $h=0.1$ on $[-2,2]^2$, with $\vec A=(-y,0,0)$, for which $\vec B=(0,0,1)$, and with the smooth test function $\tilde{Ξ}=\exp(-(x^2+y^2))\cos(2x)\sin(3y)$. With $q=1$ and the action in units of $\hbar$, the identity $[\pi_x,\pi_y]\tilde{Ξ}=iB_z\tilde{Ξ}$ was confirmed at the interior points with a maximum absolute error $4.8\times10^{-2}$; halving the spacing to $h=0.05$ gave $1.2\times10^{-2}$, the factor four of the second-order central difference.

Problem 3: The spin–orbit and Darwin terms at the next order

Statement. Extend the elimination to the next order in $1/c$ and obtain the relativistic kinetic correction, the Darwin term, and the spin–orbit coupling.

Solution. This order is not contained in the parent article, which stops at the leading Pauli Hamiltonian. The even terms of order $c^{-2}$ have two sources. The second-order term $-\tfrac12[S_1,[S_1,E]]$ of the first transformation gives the Darwin and spin–orbit couplings; the expansion of the free large-part energy gives the relativistic kinetic correction. The second transformation removes the residual odd term $O_1$ and adds no even term at this order.

Step 1: the even term of the first transformation. From Step 4 of Problem 1,

$$ [S_1,E]=-\frac{i\,q}{2\,k_m}\sum_\eta\vec e_\eta E_\eta\,(\cdot)\,\vec e_z =-\frac{i\,q}{2\,k_m}\,(\vec E\,\cdot)\,\vec e_z . $$

A direct use of the algebra, with $[S_1,E]$ above and $S_1=-\frac{1}{2\,m\,c}\sum_j\vec e_j\pi_j(\cdot)\vec e_z$, gives

$$ -\tfrac12[S_1,[S_1,E]] =-\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\,i(\vec E\times\vec\pi) -\frac{q}{8\,k_m^{2}}\,\vec\nabla_\Phi\cdot\vec E . $$

The first term comes from the antisymmetric part of $\vec e_j\vec e_\eta=-\delta_{j\eta}+\epsilon_{j\eta l}\vec e_l$; the second comes from the symmetric part. Since $\vec\nabla_\Phi\cdot\vec E=-\vec\nabla_\Phi^2V/q$, the second term is the Darwin term $+\hbar^{2}\vec\nabla_\Phi^2 V/(8m^2c^2)$. In the large part, $i\vec a$ acts as $\boldsymbol\sigma\cdot\vec a$, so the two terms read

$$ -\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\,\boldsymbol\sigma\cdot(\vec E\times\vec\pi) -\frac{q}{8\,k_m^{2}}\,\vec\nabla_\Phi\cdot\vec E . $$

Step 2: the Darwin term. Since $\vec E=-\vec\nabla_\PhiΞ$ and $V=qΞ$,

$$ \frac{1}{8\,k_m^{2}}\vec\nabla_\Phi^2V=-\frac{q}{8\,k_m^{2}}\vec\nabla_\Phi\cdot\vec E . $$

This is the Darwin term. Its Hermiticity is immediate: $\vec\nabla_\Phi\cdot\vec E$ is a real function of position.

Step 3: the spin–orbit term. Since $\vec\nabla_\Phi V=q\nablaΞ=-q\vec E$,

$$ -\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\boldsymbol\sigma\cdot(\vec E\times\vec\pi) =\frac{\hbar}{4\,m^{2}\,c^{2}}\,\boldsymbol\sigma\cdot(\vec\nabla_\Phi V\times\vec\pi) =\frac{\hbar}{4\,m^{2}\,c^{2}}\,\frac1r\frac{dV}{dr}\,\boldsymbol\sigma\cdot\vec L \qquad(\text{central }V), $$

where $\vec E\times\vec\pi$ has been written with $\vec E=-\vec\nabla_\PhiΞ$. For a central electrostatic field, whose vector potential may be neglected so that $\vec\pi\to p$, the last factor is the standard spin–orbit coupling with $\vec L=\vec r\times p$ and $\vec S=\frac{\hbar}{2}\boldsymbol\sigma$. The coefficient $1/(4m^2c^2)$ already contains the Thomas factor of $\tfrac12$: a naive Lorentz transformation to the instantaneous rest frame of the electron would give twice this value, and the factor of $\tfrac12$ is the Thomas precession correction. With $\vec S=\frac{\hbar}{2}\boldsymbol\sigma$, the term reads $\dfrac{1}{2m^2c^2}\dfrac1r\dfrac{dV}{dr}\vec L\cdot\vec S$, the standard spin–orbit coupling.

Step 4: the relativistic kinetic correction. The free large-part energy is the expansion of $\sqrt{m^2c^4+O^2}$, and the next term is fixed by the same $O^2$:

$$ \sqrt{m^2c^4+O^2} =mc^2+\frac{O^2}{2mc^2}-\frac{O^4}{8m^3c^6}+\cdots . $$

In the large part $O^4=c^4(\boldsymbol\sigma\cdot\vec\pi)^4$, so the correction is

$$ -\frac{(\boldsymbol\sigma\cdot\vec\pi)^4}{8m^3c^2} =-\frac{1}{8m^3c^2}\bigl(\vec\pi^2-i\,\hbar\,q\,\vec B\bigr)^2 . $$

It is $(\boldsymbol\sigma\cdot\vec\pi)^4$, not $\vec\pi^4$, because $(\boldsymbol\sigma\cdot\vec\pi)^2=\vec\pi^2-\hbar\,q\,\boldsymbol{\sigma}\,\vec B$; the difference is of the same $c^{-2}$ order and is not dropped.

Step 5: the second generator and the assembled Hamiltonian. The odd remainder is $O_1$, and the transformation that removes it is $S_2=-\frac{i}{2\,m\,c^{2}}\beta O_1$. With $O_1$ from Problem 1 and $\beta(\vec E\,\cdot)\vec e_z=i(\vec E\,\cdot)\vec e_y$,

$$ \beta O_1=\frac{i\,q}{2\,k_m}\sum_\eta\vec e_\eta E_\eta\,(\cdot)\,\vec e_y, \qquad S_2=-\frac{i}{2\,m\,c^{2}}\beta O_1=\frac{\hbar\,q}{4\,c^{3}\,m^{2}}\sum_\eta\vec e_\eta E_\eta\,(\cdot)\,\vec e_y . $$

The transformation cancels $O_1$; its even terms are of higher order and are dropped. Collecting the leading and next-to-leading orders, the large-part Hamiltonian is

$$ \boxed{\; H=\frac{\vec \pi^{2}}{2\,m}+qΞ-\frac{\hbar\,q}{2\,m}\boldsymbol\sigma\cdot\vec B -\frac{(\boldsymbol\sigma\cdot\vec\pi)^4}{8m^3c^2} +\frac{1}{8\,k_m^{2}}\vec\nabla_\Phi^2 V -\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\boldsymbol\sigma\cdot(\vec E\times\vec\pi) \;} $$

to relative order $c^{-2}$. The first three terms are the Pauli Hamiltonian of Problem 1; the last three are the corrections that the parent article does not reach.

Step 6: numerical verification against the exact Dirac spectrum. The three corrections are fixed by their coefficients; a single overall check is therefore decisive. In a hydrogenic atom with nuclear charge $\mathfrak{Z}$, the first-order shifts of the three corrections are the standard ones, each carrying the rest energy $m\,c^{2}$,

$$ \delta E_{\rm kin}=-\frac{m\,c^{2}(\mathfrak{Z}\alpha)^4}{2n^3}\!\left[\frac{1}{\ell+\tfrac12}-\frac{3}{4n}\right], $$

$$ \delta E_{\rm SO}=\frac{m\,c^{2}(\mathfrak{Z}\alpha)^4}{2n^3}\,\frac{j(j+1)-\ell(\ell+1)-\tfrac34}{2\ell(\ell+\tfrac12)(\ell+1)}\quad(\ell\neq0), $$

$$ \delta E_{\rm D}=\frac{m\,c^{2}(\mathfrak{Z}\alpha)^4}{2n^3}\quad(\ell=0), \qquad \delta E_{\rm D}=0\quad(\ell\neq0), $$

with the exact Dirac shift $\delta E_{\rm exact}=-\dfrac{m(\mathfrak{Z}\alpha)^4}{2n^3}\Bigl[\dfrac{1}{j+\tfrac12}-\dfrac{3}{4n}\Bigr]$. Evaluating all of these with exact rational arithmetic for every level with $n\le6$ (36 levels, with $j$ the total angular momentum, $j=\ell\pm\tfrac12$ and one value at $\ell=0$), the sum $\delta E_{\rm kin}+\delta E_{\rm SO}+\delta E_{\rm D}$ equals $\delta E_{\rm exact}$ in every case, with difference exactly $0$. The coefficients are also those of the two sources named at the head of this problem: the binomial coefficients $1,\tfrac12,-\tfrac18$ of $\sqrt{m^2c^4+O^2}$ and the Darwin coefficient $+\tfrac18$ of the second-order term.

Problem 4: Hermiticity and the $\mathbb{B}_{\Psi}/\mathbb{B}_{\Phi}$ decomposition

Statement. Check the Hermiticity of every term and determine which terms lie in $\mathbb{B}_{\Psi}$.

Solution.

Step 1: Hermiticity. Each term of the Hamiltonian of Problem 3 is Hermitian:

  • $\vec\pi^2 c/(2\,m\,c)$: $\pi_i^\dagger=\pi_i$, hence $\sum_i\pi_i^2$ is Hermitian;
  • $qΞ$: multiplication by the real function $qΞ$, Hermitian;
  • $-\frac{\hbar\,q}{2\,m}\boldsymbol\sigma\cdot\vec B$: $\sigma_\eta$ and the real field $B_\eta$ are Hermitian;
  • $-\dfrac{(\boldsymbol\sigma\cdot\vec\pi)^4}{8m^3c^2}$: minus a real multiple of the square of the Hermitian $(\boldsymbol\sigma\cdot\vec\pi)^2$;
  • $\frac{1}{8\,k_m^{2}}\vec\nabla_\Phi^2 V$: multiplication by a real function;
  • $-\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\boldsymbol\sigma\cdot(\vec E\times\vec\pi)$: this is the one term whose Hermiticity is not automatic. Using $[\pi_i,E_j]=-i\,\hbar\partial/\partial r_i\,E_j$,

$$ (\vec E\times\vec\pi)_i^\dagger=(\vec E\times\vec\pi)_i+i\,\hbar\,(\vec\nabla_\Phi\times\vec E)_i , $$

so the term is Hermitian if and only if $\vec\nabla_\Phi\times\vec E=0$, its failure of Hermiticity being proportional to the curl. Since $\vec E=-\vec\nabla_\PhiΞ$ for an electrostatic field, $\vec\nabla_\Phi\times\vec E=-\vec\nabla_\Phi\times\nablaΞ=0$ identically, and the term is Hermitian. Writing it instead as $\boldsymbol\sigma\cdot(\vec\nabla_\Phi V\times\vec\pi)$ makes this manifest, because $\vec\nabla_\Phi V$ is a gradient. For a general time-dependent field, where $\vec E$ has a solenoidal part and $\vec\nabla_\Phi\times\vec E=-\partial/\partial r'_t\,\vec B$, the antisymmetric part does not vanish and the non-relativistic reduction requires the symmetrized ordering; this is the only place in the reduction where the result is convention- and ordering-sensitive beyond the metric conventions already noted.

Step 2: the $\mathbb{B}_{\Psi}$ representatives. Under the isomorphism, a Hermitian $2\times2$ matrix $h_t\,I_2+\vec h_\Psi\cdot\boldsymbol\sigma$ corresponds to the Hermitian biquaternion $h_t\,e_t+i\vec h_\Psi\in\mathbb{B}_{\Psi}$. Applying it term by term, with $\vec B=\sum_\eta B_\eta\vec e_\eta$ and $(\vec E\times\vec\pi)=(\vec E\times\vec\pi)_k\vec e_\eta$,

| Term of $H$ | Biquaternion representative | Sector | |+-|+-|+-| | $\vec\pi^2 c/(2\,m\,c)$ | $(\vec\pi^2 c/(2\,m\,c))\,e_t$ | $\mathbb{B}_{\Psi}$ | | $qΞ$ | $(qΞ)\,e_t$ | $\mathbb{B}_{\Psi}$ | | $-\frac{\hbar\,q}{2\,m}\boldsymbol\sigma\cdot\vec B$ | $-\frac{\hbar\,q}{2\,m}\,i\vec B$ | $\mathbb{B}_{\Psi}$ | | $-\dfrac{(\boldsymbol\sigma\cdot\vec\pi)^4}{8m^3c^2}$ | $-\dfrac{1}{8m^3c^2}\bigl(\vec\pi^2-i\,\hbar\,q\,\vec B\bigr)^2$ | $\mathbb{B}_{\Psi}$ | | $\frac{1}{8\,k_m^{2}}\vec\nabla_\Phi^2 V$ | $\frac{1}{8\,k_m^{2}}(\vec\nabla_\Phi^2V)\,e_t$ | $\mathbb{B}_{\Psi}$ | | $-\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\boldsymbol\sigma\cdot(\vec E\times\vec\pi)$ | $-\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\,i(\vec E\times\vec\pi)$ | $\mathbb{B}_{\Psi}$ |

Every term of the reduced, two-component Hamiltonian is Hermitian, hence every term lands in $\mathbb{B}_{\Psi}$. This is the precise content of the parent's remark that the spin term is an observable of the informational sector; the remark applies term by term. Each representative is a left multiplication by an element, which is what makes the identification possible at all.

Step 3: what does not land in $\mathbb{B}_{\Psi}$. Three classes of object in the reduction do not.

  1. The right multiplications. The mass term $\beta mc^2$ multiplies by $i\vec e_x$ on the right, and the odd operator $O$ ends in a right multiplication by $\vec e_y$. The two multipliers are $i\vec e_x$, which is Hermitian, and $\vec e_y$, which is anti-Hermitian; each operator is Hermitian anyway, because the left factor $\vec e_\eta$ is anti-Hermitian too and a product of two anti-Hermitian factors with a Hermitian $\pi_\eta$ between them is Hermitian. What matters is that neither operator is a left multiplication, so no element of $\mathbb B$ represents either. This is why the four-component Hamiltonian is not an element of $\mathbb B$ and why a transformation is needed before the biquaternion identification can be made: the Foldy–Wouthuysen transformation is exactly the operation that replaces the two-sided operators by left multiplications and leaves a sum of elements of $\mathbb{B}_{\Psi}$. A right multiplication is a different operator from the left multiplication by the same element, and neither may be substituted for the other.

  2. The anti-Hermitian generator. $S_1$ is Hermitian, so $iS_1$ is anti-Hermitian: it plays the rôle that the elements of $\mathbb{B}_{\Phi}$ play in the companion quantum article, where the Lie algebra of the $*$-unitary group is $\mathbb{B}_{\Phi}$. It is not itself in $\mathbb{B}_{\Phi}$, because it is built from right multiplications and is not a left multiplication by an element. The two gradings must not be conflated: the Foldy–Wouthuysen grading is by conjugation with $\beta$, the $\mathbb{B}_{\Psi}/\mathbb{B}_{\Phi}$ grading is by ${}^{*}$. Thus the odd operator $O$ is Foldy–Wouthuysen-odd and Hermitian, yet it is not the left multiplication by the element $\hbar\,c\,\vec \pi$ — that element acts on one side and is even — and the identification of "odd" with "not in $\mathbb{B}_{\Psi}$" is incorrect.

  3. The real structure and the mass term. In the biquaternion equation the conjugation $\flat=-{}^{*}$ is the algebra's real structure. It maps $\mathbb{B}_{\Psi}$ to $\mathbb{B}_{\Phi}$ and fixes $\mathbb{B}_{\Phi}$: if $F\in\mathbb{B}_{\Psi}$ then $F^\flat=-F\in\mathbb{B}_{\Phi}$, and if $F\in\mathbb{B}_{\Phi}$ then $F^\flat=F$. Its action is diagonal on the sector decomposition with opposite signs $\mp1$: a coupling built on $\flat$ — a Majorana-type mass, $mF^\flat$ — would carry those opposite signs. The parent's mass term is not that coupling; it is the linear, chirality-off-diagonal pair $\tilde\nabla_\Phi F_R=mF_L$, $(\tilde\nabla_\Phi)^{\natural}F_L=mF_R$, which relates the two parts of the chirality and leaves each element's sector membership untouched. The operation that exchanges the two sectors is multiplication by the central $i$. The rest-energy factor that is removed before the reduction, the central element $\cos\theta_{\mathrm{rst}}-i\sin\theta_{\mathrm{rst}}$ with $\theta_{\mathrm{rst}}=r'_t\,k_m$, is not Hermitian either: the conjugation inverts $\theta_{\mathrm{rst}}$, so the element is fixed by ${}^{*}$ neither as it stands nor after multiplication by $-1$, and it lies in neither $\mathbb{B}_{\Psi}$ nor $\mathbb{B}_{\Phi}$. The phase that removes the mass term from a $\beta$-diagonal equation, $\exp(-i\beta\theta_{\mathrm{rst}})=\cos\theta_{\mathrm{rst}}\,e_t+\sin\theta_{\mathrm{rst}}\,\vec e_x$, is a right multiplication, and it splits the other way: its scalar part lies in $\mathbb{B}_{\Psi}$ and its pure part in $\mathbb{B}_{\Phi}$, since $\vec e_x$ is anti-Hermitian. Either way the rest-energy factor is a phase and not an observable, and it must be factored out before a two-component (hence $\mathbb{B}_{\Psi}$) description of the dynamics is reached.

Step 4: the sides, one line per operator. Every operator of the reduction is listed with the element that multiplies on the left and the element that multiplies on the right. The column headed "side" is the answer to the only question that matters here: whether the operator is a single multiplication or a product of two.

| Operator | Left element | Right element | Side | |+-|+-|+-|+-| | $\beta$ | $e_t$ | $i\vec e_x$ | right multiplication alone | | $\Gamma$ | $e_t$ | $-i\vec e_z$ | right multiplication alone | | $O$ | $\vec e_\eta$, with $\pi_\eta$ | $\vec e_y$ | two-sided | | $S_1$ | $\vec e_\eta$, with $\pi_\eta$ | $\vec e_z$ | two-sided | | $O_1$ | the element $\vec E$ | $\vec e_z$ | two-sided | | $S_2$ | the element $\vec E$ | $\vec e_y$ | two-sided | | $O^2$ | $\vec e_j\vec e_\eta$, with $\pi_j\pi_\eta$ | $e_t$, the two $\vec e_y$ having collapsed | left multiplication by an element | | $\vec\pi^2$, $qΞ$, $\boldsymbol\sigma\cdot\vec B$, the $c^{-2}$ terms | an element of $\mathbb{B}_{\Psi}$ | $e_t$ | left multiplication by an element |

No other right multiplication and no other differential operator occurs in the reduction: the only derivative is the pure part of the gradient, $\sum_\eta\vec e_\eta\pi_\eta=-i\,\hbar\,D_{3}^{A}$, and it always appears with the single left element $\vec e_\eta$ shown. The elements $\vec e_y$ and $\vec e_z$ that close $O$, $S_1$, $O_1$ and $S_2$ are the only right multipliers, and no symbol of the article stands for an operator whose sides are left unstated.

Problem 5: The physical dimensions of each term

Statement. Determine the physical dimensions of each term.

Solution. Every term is an energy, since each is a term of a Hamiltonian. The individual coefficients nevertheless carry different powers of $\hbar$ and $c$, and checking them is a useful consistency test. In SI units, with $[\hbar]=\mathrm{kg\,m^2\,s^{-1}}$, $[c]=\mathrm{m\,s^{-1}}$, $[q]=\mathrm C$, $[Ξ]=\mathrm V=\mathrm{kg\,m^2\,s^{-3}A^{-1}}$, $[\vec B]=\mathrm T=\mathrm{kg\,s^{-2}A^{-1}}$ and $[\vec E]=\mathrm{V\,m^{-1}}=\mathrm{kg\,m\,s^{-3}A^{-1}}$:

| Term | Coefficient units | Field/operator units | Product | |+-|+-|+-|+-| | $\vec\pi^2 c/(2\,m\,c)$ | $1/m$ | $(\mathrm{kg\,m\,s^{-1}})^2$ | $\mathrm{kg\,m^2\,s^{-2}}$ | | $qΞ$ | $\mathrm C$ | $\mathrm{kg\,m^2\,s^{-3}A^{-1}}$ | $\mathrm{kg\,m^2\,s^{-2}}$ | | $-\frac{\hbar\,q}{2\,m}\boldsymbol\sigma\cdot\vec B$ | $\mathrm C\,\mathrm{kg\,m^2\,s^{-1}\,kg^{-1}}$ | $\mathrm{kg\,s^{-2}A^{-1}}$ | $\mathrm{kg\,m^2\,s^{-2}}$ | | $-\frac{q}{8\,k_m^{2}}\vec\nabla_\Phi\cdot\vec E$ | $\mathrm C\,\mathrm{kg^2m^4s^{-2}}\,\mathrm{kg^{-2}m^{-2}s^{2}}$ | $\mathrm{kg\,s^{-3}A^{-1}}$ | $\mathrm{kg\,m^2\,s^{-2}}$ | | $-\frac{\hbar\,q}{4\,m^{2}\,c^{2}}\boldsymbol\sigma\cdot(\vec E\times\vec\pi)$ | $\mathrm C\,\mathrm{kg\,m^2s^{-1}}\,\mathrm{kg^{-2}m^{-2}s^{2}}$ | $\mathrm{kg^2m^2s^{-4}A^{-1}}$ | $\mathrm{kg\,m^2\,s^{-2}}$ | | $-\dfrac{(\boldsymbol\sigma\cdot\vec\pi)^4}{8m^3c^2}$ | $\mathrm{kg^{-3}m^{-2}s^{2}}$ | $\mathrm{kg^4m^4s^{-4}}$ | $\mathrm{kg\,m^2\,s^{-2}}$ |

The pattern is informative. The leading terms ($\vec\pi^2 c/(2\,m\,c)$, $qΞ$, the spin term) carry no power of $c$; the three corrections each carry $c^{-2}$, which is the statement that they are relativistic corrections of order $v^2/c^2$. The spin term carries one power of $\hbar$ in its coefficient, while the leading kinetic and electrostatic terms carry none. Written with $\vec S=\frac{\hbar}{2}\boldsymbol\sigma$, it is $-(q/m)\vec S\cdot\vec B$, the classical coupling of a magnetic moment; the power of $\hbar$ is the conversion between the dimensionless $\boldsymbol\sigma$ and the spin. The Darwin term carries $\hbar^{2}$ and has no classical counterpart. The spin–orbit term carries one power of $\hbar$ in its coefficient, but it is quadratic in the angular momenta once $\vec S$ and $\vec L=\vec r\times p$ are used, so it becomes the classical Thomas-precession coupling when $\vec S$ and $\vec L$ are treated as classical vectors.

Summary

The non-relativistic limit of the biquaternion Dirac equation was carried through in five steps, for the one-sided chiral pair read on the Dirac module. Every derivative in the reduction is the biquaternionic gradient: the free equation is the first-order one-field equation $\tilde\nabla_\Phi F=0$, the massive equation is the chiral pair $\tilde\nabla_\Phi F_R=mF_L$, $(\tilde\nabla_\Phi)^{\natural}F_L=mF_R$, and the square of the gradient is neither scalar nor the d'Alembertian, so no one-field first-order massive equation was written at any point. The two-sided form of the parent is a different operator, antilinear in its mass term, and is not reduced here.

The Foldy–Wouthuysen elimination of the small component, executed systematically with the generator

$$ S_1=-\frac{i}{2\,m\,c^{2}}\beta O=-\frac{1}{2\,m\,c}\sum_\eta\vec e_\eta\pi_\eta\,(\cdot)\,\vec e_z=\frac{i}{2\,k_m}\bigl(D_3^A\cdot\bigr)\vec e_z , $$

produces the Pauli Hamiltonian $i\,\hbar\,c\partial\tilde{Ξ}_{\mathrm{evn}}/\partial r'_t=\bigl[\vec\pi^2 c/(2\,m\,c)+qΞ-(qc/k_m/2)\boldsymbol\sigma\cdot\vec B\bigr]\tilde{Ξ}_{\mathrm{evn}}$, reproducing the parent article. The minimal substitution in the gradient, $\tilde\nabla_\Phi\to\tilde\nabla_\Phi-\frac{i\,q\,\tilde{A}_{\Phi}}{\hbar}$, gives $[\pi_i,\pi_j]=i\,\hbar\,q\,\epsilon_{ij\eta}\,B_{\eta}$ and $(\boldsymbol\sigma\cdot\vec\pi)^2=\vec\pi^2-\hbar\,q\,\boldsymbol{\sigma}\,\vec B$, and the coefficient $1/2m$ produced by the elimination fixes $g=2$ and $\vec\mu=(q/m)\vec S$, hence $\mu_B=ec\dfrac{1}{2\,k_{m_e}}$ for the electron. At the next order in $1/c$, the second-order term $-\tfrac12[S_1,[S_1,E]]$ of the first transformation gives the Darwin term $+\hbar^{2}\vec\nabla_\Phi^2 V/(8m^2c^2)=-(\frac{q}{8\,k_m^{2}})\vec\nabla_\Phi\cdot\vec E$ and the spin–orbit coupling $-(\frac{\hbar\,q}{4\,m^{2}\,c^{2}})\boldsymbol\sigma\cdot(\vec E\times\vec\pi)=(\frac{\hbar}{4\,m^{2}\,c^{2}})r^{-1}(dV/dr)\boldsymbol\sigma\cdot\vec L$, while the expansion of the free large-part energy gives the relativistic kinetic correction $-(\boldsymbol\sigma\cdot\vec\pi)^4/(8m^3c^2)$. The second transformation removes the residual odd term $O_1$ and adds no even term at this order. Summed, these reproduce the exact Dirac fine structure of hydrogenic ions for 36 levels with $n\le6$, with difference exactly zero.

Every term of the reduced Hamiltonian is Hermitian and therefore lies in $\mathbb{B}_{\Psi}$. The objects that do not are the two-sided operators of the reduction — the mass term $\beta mc^2$ and the odd operator $O$, which end in right multiplications by $i\vec e_x$ and by $\vec e_y$ and so are not left multiplications by an element — together with the anti-Hermitian generator $iS_1$ and the rest-energy phase, which is not Hermitian and belongs to neither sector. The real structure $\flat$ — whose opposite-sign action on the sectors a Majorana-type coupling would carry — is not the parent's mass term, which is the linear coupling of the two chiralities. The Foldy–Wouthuysen grading by $\beta$ must not be confused with the $\mathbb{B}_{\Psi}/\mathbb{B}_{\Phi}$ grading by ${}^{*}$. All terms are energies, the three corrections carrying the expected $c^{-2}$.

Summary of Notation

| Symbol | Meaning | |+-|+-| | $\mathbb B=\mathbb C\otimes_\mathbb R\mathbb H$ | Biquaternion algebra | | $e_t=1,\vec e_x,\vec e_y,\vec e_z$, $\vec e_\eta^2=-e_t$ | Quaternion basis | | $\mathbb{B}_{\Psi},\mathbb{B}_{\Phi}$ | Hermitian / anti-Hermitian subspaces | | $\mathsf{M}_2(\vec e_\eta)=-i\sigma_\eta$, $\sigma_\eta\leftrightarrow i\vec e_\eta$ | Isomorphism, Pauli matrices | | $\tilde\nabla_\Phi$, $(\tilde\nabla_\Phi)^{\natural}$ | The gradient and its conjugate, $\varepsilon=(1,-1,-1,-1)$ | | $\Box_\Phi=\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}$ | The d'Alembertian | | $D_3=\sum_\eta\vec e_\eta\partial/\partial r_\eta$, $D_3^A$ | Pure part of the gradient, free and gauged | | $\tilde\nabla_\Phi-\frac{i\,q\,\tilde{A}_{\Phi}}{\hbar}$, $\tilde A_{\Phi}=Ξ e_t-\vec A$ | The gauged gradient; minimal substitution | | $-i\tilde K_{\Phi}=k_{t}^{\prime}-i\vec k_\Phi$ | Symbol of the gradient on a plane wave; $(-i\tilde K_{\Phi})^{\natural}(-i\tilde K_{\Phi})=k_{t}^{\prime\,2}-\lvert\vec k_\Phi\rvert^2$ | | $F=F_L+F_R$; $\Gamma F=-iF\vec e_z$ | The field and its chirality, values $\mp1$ | | $\beta F=iF\vec e_x$, $\beta^2=1$, $\beta\Gamma=-\Gamma\beta$ | The mass grading; $\beta O=-O\beta$ | | $O=c\sum_\eta\vec e_\eta\pi_\eta(\cdot)\vec e_y=-i\,\hbar\,c\,D_{3}^{A}\,\vec e_{y}$ | The odd operator; $O^2=c^2(\vec\pi^2-i\,\hbar\,q\,\vec B)$ | | $\omega=-\vec e_z$, $\omega^2=-e_t$ | The pseudoscalar of the frame | | $\vec\pi=p-q\vec A$ | Kinetic momentum | | $\vec B=\vec\nabla_\Phi\times\vec A$, $\vec E$ | Magnetic, electric fields | | $V=qΞ$ | Potential energy | | $H=\beta mc^2+O+E$, $E=qΞ$ | Split Hamiltonian; odd and even parts | | $S_1=-\frac{i}{2\,m\,c^{2}}\beta O=-\frac{1}{2\,m\,c}\sum_\eta\vec e_\eta\pi_\eta(\cdot)\vec e_z$ | Foldy–Wouthuysen generator | | $O_1=i[S_1,E]=\frac{q}{2\,k_m}(\vec E\cdot)\vec e_z$ | Residual odd operator | | $S_2=-\frac{i}{2\,m\,c^{2}}\beta O_1=\frac{\hbar\,q}{4\,c^{3}\,m^{2}}(\vec E\cdot)\vec e_y$ | Second generator | | $\vec S=\frac{\hbar}{2}\boldsymbol\sigma$, $\vec\mu=g\frac{q}{2m}\vec S$ | Spin and magnetic moment | | $g=2$ | Gyromagnetic factor | | $\mu_B=ec\dfrac{1}{2\,k_{m_e}}$ | Bohr magneton |

Further Reading

  • P. A. M. Dirac, "The quantum theory of the electron," Proceedings of the Royal Society A 117 (1928) 610–624, for the original prediction of $g=2$.
  • L. L. Foldy and S. A. Wouthuysen, "On the Dirac theory of spin 1/2 particles and its non-relativistic limit," Physical Review 78 (1950) 29–36, for the transformation used here.
  • J. D. Bjorken and S. D. Drell, Relativistic Quantum Mechanics (McGraw-Hill, 1964), for the plane-wave solutions and the standard non-relativistic reduction.
  • J. J. Sakurai, Advanced Quantum Mechanics (Addison-Wesley, 1967), for the emergence of the Pauli equation and the spin–orbit coupling.
  • W. Greiner, Relativistic Quantum Mechanics: Wave Equations (Springer, 2000), for a step-by-step Foldy–Wouthuysen elimination and Pauli limit.
  • C. Itzykson and J.-B. Zuber, Quantum Field Theory (McGraw-Hill, 1980), for the fine structure and the hydrogen spectrum.
  • The Dirac Equation in Biquaternionic Form, for the two forms and the mass term.
  • The Dirac Equation in Biquaternionic Form — Solutions and the Non-Relativistic Limit, the parent of this exercise, for the Pauli limit at leading order.
  • The Foldy–Wouthuysen Transformation in Biquaternionic Form, for the transformation in the algebra.
  • Quantum Physics in Biquaternionic Form, for the sector structure and the observables.
  • Operator and Calculus Conventions, for the gradient, its conjugate, the d'Alembertian and the gauged gradient.