The Fourier Transform and the Coupled Proca Equation in Biquaternionic Form
Introduction
The Proca field is the first in the corpus whose carrier has to be widened before a connection can be written, and the widening is forced by the connection itself. The Coupled Proca Equation in Biquaternionic Form states the reason in its §The Carrier: the minimal substitution replaces the gradient by $\tilde\nabla_\Phi+\frac{i\varkappa}{\hbar}\tilde A_\Phi$, the product of two material potentials is not material because of the $i$ in the time slot, and the field therefore has to be a general biquaternion. The $i$ that breaks closure is the $i$ the mass term needs, which is why the massless Proca field can be a real potential and the massive one cannot.
The article's subject is what the transform says about that widening, and the answer is a two-term symbol. In the coupled theory the operator's symbol is the free symbol at a shifted four-wave-number, plus one extra term that is the curvature:
$$ \sigma_{\mathrm{gco}}\bigl(\pi\bigr)=\underbrace{\bigl(-\pi^{2}\delta_{\mu\nu}-\pi_\mu\pi_\nu\bigr)}_{\text{the free symbol at }\pi} \;+\;\underbrace{\frac{i\varkappa}{\hbar}\,F^{(\Phi)\mu}{}_{\nu}}_{\text{the obstruction}} , \qquad \pi_\mu=\hbar k_\mu-q A_\mu . $$
The first bracket is the free Proca symbol with $\pi$ in place of $k$ — translation and rank-one projector together — and the second is the commutator of two covariant derivatives. The first term is what the transform says and the second is what the transform cannot say, and the article is the separation of the two.
The translation is exact and it is the corpus's own identity. The parent proves $(\tilde\nabla_\Phi)^{\natural}_{\mathrm{gco}}\tilde\nabla_{\mathrm{gco},\Phi}=\sum_\mu\tilde\nabla_{\Phi,\mathrm{gco},\mu}^{2}+\frac{i\varkappa}{\hbar}\tilde F^{(\Phi)}$, a decomposition of the covariant d'Alembertian into a symmetric sum of squares and a curvature. In frequency space the first piece is $-\pi^{2}$ and the second is the curvature term, so the identity is exactly the symbol's two terms and the article's first finding is that no third term exists.
The second finding is that the same four-wave-number can give different symbols. Two connections can agree in $\pi$ at a point and disagree in $F$ there, and the symbol's difference is $\frac{i\varkappa}{\hbar}$ times the difference of the curvatures: for two connections used below, $\lvert F_{21}\rvert$ is $0$ and $0.9$ while $\pi$ agrees at the point of evaluation. So the symbol is not a function of the momentum alone, and that is the exact statement of what a Fourier description of a gauge theory loses.
The third finding is that the mass does not enter the obstruction. The mass term is $-k_m^{2}\delta$ on the amplitude space, which is inert under the curvature; the curvature term carries no $k_m$. So the obstruction and the mass are two independent additions to the Klein–Gordon symbol, and §The Mass Does Not Enter the Obstruction states the check.
The fourth finding is a floor on how much a coefficient can hide. The companions' bench measures the mode occupancy of a multiplication, a derivative of a delta and a general function on a $256$-mode circle: one mode, first order, and all $256$. Covariant differentiation is the derivative's floor, so a connection moves the symbol away from the free one by an amount the occupancy does not see, and the article quotes the bench rather than rerunning it.
The denial is stated plainly. No sector is exchanged. The coupled field is an element of $\mathbb{B}$, its transform is $\mathbb{B}$-valued, and the symbol is a matrix of numbers only at a constant connection. The article adds the decomposition and its two failures: the failure of the symbol to be a function of $\pi$, and the failure of the ladder to reach the floor.
Conventions
The algebra, the products, the two forms, the invertibles and the sectors are those of Algebra and Element Conventions, Product and Form Conventions, Norm and Pairing Conventions, Physics Units and Common Elements Conventions. The gradient and the covariant derivative are those of Operator and Calculus Conventions and Gauge Fields and the Covariant Derivative, and the transform is the one of Derivatives, Fields and Transforms, whose symbol rule for a gauge-covariant operator is quoted in derived form:
$$ \tilde\nabla_{\mathrm{gco},\Phi,\mu}\longmapsto i\pi_\mu , \qquad \pi_\mu=\hbar k_\mu-q A_\mu , \qquad k_\mu=\bigl(k_t',\vec k_\Phi\bigr) , \qquad k_t'=\frac{\omega}{c} , $$
with $\varkappa$ the coupling and $A_\mu$ the background potential. The shift is a translation of the wave number and not a scaling: the coupling enters as a length and the potential as its inverse.
The minimal substitution and the field strength are the parent's. The substitution is $\tilde\nabla_\Phi\mapsto\tilde\nabla_\Phi+\frac{i\varkappa}{\hbar}\tilde A_\Phi$, and the covariant field strength is
$$ \tilde F=\mathrm{Vect}\bigl((\tilde\nabla_{\mathrm{gco},\Phi})^{\natural}\tilde A\bigr) =\mathrm{Vect}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde A\bigr) +\frac{i\varkappa}{\hbar}\,\mathrm{Vect}\bigl(\tilde A_\Phi^{\natural}\tilde A\bigr) , $$
whose second term is the parent's extra piece, measured between $0.7$ and $12$ over one hundred random draws. That term is a product of two fields and not a derivative, and §The Field Strength's Extra Term is about what the transform does with it.
The square-splitting identity is the parent's and it is the article's spine. For a connection with curvature $F^{(\Phi)}$,
$$ (\tilde\nabla_{\mathrm{gco},\Phi})^{\natural}\tilde\nabla_{\mathrm{gco},\Phi} =\sum_\mu\tilde\nabla_{\Phi,\mathrm{gco},\mu}^{2}+\frac{i\varkappa}{\hbar}\tilde F^{(\Phi)} , \qquad \bigl[D_{\mathrm{gco},\mu},D_{\mathrm{gco},\nu}\bigr]=\frac{i\varkappa}{\hbar}F^{(\Phi)}_{\mu\nu} , $$
the commutator checked on a test function to $1.1\times10^{-5}$. The first display is the symbol's decomposition before the transform is applied, and the transform turns the two terms into $-\pi^{2}$ and $\frac{i\varkappa}{\hbar}F^{(\Phi)}$.
The constant-connection licence. The symbol below is the symbol of a differential operator with constant coefficients, which requires a constant background. A linear background is not constant and its symbol is not defined; §The Obstruction: Same Momentum, Different Curvature handles that case by evaluating at a point and saying so.
The denial of the Legendre reading, once. The transform's variable is not a canonical momentum, and the parent's $d_\Phi=-i\hbar c\tilde\partial_\Phi+i\varkappa c\tilde A_\Phi$ is a covariant derivative and not a conjugate momentum. The corpus's second conjugacy is the canonical one, and it is not used here.
The Symbol in Two Terms
The transform acts on the identity and gives the decomposition, term by term. Applying the symbol rule to the square splitting, the symmetric sum of squares becomes $-\pi^{2}$ — the four-momentum's interval with a sign — and the curvature term becomes $\frac{i\varkappa}{\hbar}F^{(\Phi)}$. There is no third term because the identity has two, and the article's first claim is only that the identity survives the transform unchanged.
The identity is the d'Alembertian's and the Proca symbol adds two pieces of its own. The square splitting decomposes $(\tilde\nabla_\Phi)^{\natural}_{\mathrm{gco}}\tilde\nabla_{\mathrm{gco},\Phi}$, which is the covariant d'Alembertian; the Proca operator is that d'Alembertian less the projector onto the Lorenz form and less $k_m^{2}$. So the symbol's three pieces are the interval, the projector and the curvature, and the first two together are the free symbol at $\pi$. The curvature is the only piece the free article has no counterpart for, and every statement below about the obstruction is a statement about that piece alone.
The translation term is a scalar and the curvature term is a matrix. $-\pi^{2}$ is one number times the identity in the amplitude space; $\frac{i\varkappa}{\hbar}F^{(\Phi)}$ is a two-form's matrix, antisymmetric and traceless in the four-vector reading. So the symbol is a scalar plus a traceless part, and any statement about a trace is a statement about the translation alone.
The curvature term's trace vanishes, which is the parent's Lorenz argument in frequency space. The parent shows that taking the trace of the covariant equation removes the commutator, so the curvature does not enter the covariant Lorenz condition. In the symbol's language: the trace of $\sigma_{\mathrm{gco}}$ is $-\pi^{2}\cdot4$ and has no curvature in it, so the constraint's symbol is the free one at $\pi$. That single sentence is the whole of the parent's §The Lorenz Condition in Covariant Form, and the article states it as a trace.
The curvature term is what makes the symbol non-scalar and it is what a gauge transformation cannot remove. A gauge transformation shifts $A_\mu$ by a gradient, which shifts $\pi_\mu$ and leaves $F$ alone; so the symbol's translation moves and its obstruction does not. That is the frequency-space form of gauge covariance: the symbol is not invariant, its two parts transform differently, and the next section separates them.
The Translation Is Exact
For a constant connection the symbol is the free one at shifted momentum, and the shift is the only change. With $F=0$ the second term vanishes identically, so
$$ \sigma_{\mathrm{gco}}\bigl(\pi\bigr)=-\pi^{2}\mathbb{1} \qquad\text{when}\qquad F^{(\Phi)}_{\mu\nu}=0 , $$
and the symbol's dependence on the coupling is entirely through $\pi_\mu=\hbar k_\mu-q A_\mu$. The mass shell then reads $\pi^{2}=k_m^{2}$, which is the parent's constant-background dispersion with the coupling in the momentum and nowhere else.
The check is one line and it is exact. A constant connection has $F_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu=0$ identically, so the commutator vanishes and the square splitting reduces to the symmetric sum of squares. The symbol of that sum is $-\pi^{2}$, and the identity is an identity and not an approximation.
What the translation does to the spectrum is the parent's shift, read as a symbol. The temporal component of $\pi$ is $\hbar\omega/c+\varkappa\varphi$, so the symbol's evenness is about $\omega=-\varkappa c\varphi/\hbar$ and not about zero. A constant potential moves the axis of the symbol's parity, which is the frequency-space form of "a constant potential shifts every energy by $\varkappa\varphi$".
The spatial components shift the wave vector's origin, and that is the same statement. $\vec\pi=\hbar\vec k_\Phi-q\vec A$ means the symbol's dependence on $\vec k_\Phi$ is a translation, so the dispersion's minimum sits at $\vec k_\Phi=-\varkappa\vec A/\hbar$. A constant vector potential moves the dispersion's centre and not its shape, and the symbol's translation is exactly that statement.
The translation is therefore the entire content of "the coupling is a shift" and the obstruction is the entire content of "the coupling is not only a shift". The article separates them because the corpus's literature separates them: the constant-background section of the parent is the translation, and its covariant field strength is the obstruction.
The Obstruction: Same Momentum, Different Curvature
Two connections can agree in $\pi$ at a point and disagree in $F$ there, and then the symbols differ. That is the article's sharpest statement and it is a statement about what a Fourier description can and cannot represent. Written as a display, at a point $r_0$:
$$ \pi^{(1)}(r_0)=\pi^{(2)}(r_0) \qquad\text{while}\qquad F^{(1)}(r_0)\neq F^{(2)}(r_0) \qquad\Longrightarrow\qquad \sigma^{(1)}(r_0)\neq\sigma^{(2)}(r_0) , $$
the difference being $\frac{i\varkappa}{\hbar}(F^{(1)}-F^{(2)})$ and nothing else. So the symbol at a point is not a function of the momentum at that point, and a reading that computed the symbol from $\pi$ alone would be wrong by the curvature.
The example is two connections that differ by a term linear in the coordinate, and the numbers are the parent's. Take $A^{(1)}=0$ and $A^{(2)}$ with $A_2=0$ and $A_1=b\,r_2'$; the second is the parent's linear connection, whose curvature has $\lvert F_{21}\rvert=b=0.9$ and whose other components vanish. At the point $r_0=(0,0,0,0)$ the two connections agree in every component, because a linear connection vanishes at its own origin, and their curvatures differ there by a matrix with one independent component of size $0.9$.
The consequence for a Fourier description is a limit and not a defect. The transform diagonalises a constant-coefficient operator and does nothing for a variable one. A background with curvature has no symbol, and the closest a symbol can come is the constant-connection symbol at the background's value at a point. The article's statement is that this approximation is exact in $\pi$ and wrong in $F$, which is the sharpest form the limitation takes.
What the parent's covariant field strength does with the same fact is the position-space version. The extra term $\frac{i\varkappa}{\hbar}\mathrm{Vect}(\tilde A_\Phi^{\natural}\tilde A)$, measured between $0.7$ and $12$, is the same curvature's appearance in the field strength rather than in the operator. Two registers, one object: the symbol's obstruction and the field strength's extra term are both the commutator, and the article's §The Field Strength's Extra Term is the second register.
The obstruction's rank is the curvature's rank and its trace is zero. A two-form in four dimensions has rank four unless its invariants vanish, so the obstruction is a full-rank traceless addition in general; the trace's vanishing is what leaves the constraint untouched. So the obstruction moves every eigenvalue of the symbol's transverse part and moves the constraint not at all, and that asymmetry is why the covariant Lorenz condition is the free one in $\pi$.
The Mass Does Not Enter the Obstruction
The mass term is a multiple of the identity and the obstruction is traceless, so the two additions are independent. The coupled equation's symbol in the four-component reading is
$$ \sigma_{\mathrm{gco}}=\bigl(-\pi^{2}-k_m^{2}\bigr)\delta_{\mu\nu}+\frac{i\varkappa}{\hbar}F^{(\Phi)\mu}{}_{\nu} -\pi_\mu\pi_\nu , $$
so the mass appears in one diagonal entry of the decomposition and the curvature in a different one. Neither can be written in terms of the other, and the check is that the obstruction's trace is zero at every $k_m$ and the mass's is $-4k_m^{2}$ at every $F$.
The consequence is that the count cannot be changed by the coupling. The mass's effect on the symbol's rank is the linear factor $-k_m^{2}$, which is the same at every curvature; the curvature's effect is traceless, so it cannot create or destroy the mass's factor. The degrees-of-freedom count is therefore a property of the mass and of the constraint's structure, not of the background, and the form companion's rank table survives the coupling with $\pi$ in place of $k$.
The independence is visible in the eigenvalues at a constant connection. With $F=0$ the eigenvalues are the free ones at $\pi$, and the two multiplicities $\{\pi^{2}-k_m^{2},\pi^{2}-k_m^{2},\pi^{2}-k_m^{2},-k_m^{2}\}$ are unchanged in structure; what changes is the value of $\pi^{2}$. The coupling moves the numbers and not the multiplicities, which is the same sentence as the mass's independence read forwards.
What the independence does not say is that the coupling is spectroscopically invisible. It says the coupling's effect on the count is nil. Its effect on the values is the whole of the constant-background dispersion, and its effect on the eigenvectors is the curvature's. The article states three things and not one, because the three are cheap and the confusion between them is not.
The Field Strength's Extra Term
The covariant field strength has a term the free one does not, and the transform makes it a product. The parent's display
$$ \tilde F=\mathrm{Vect}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde A\bigr)+\frac{i\varkappa}{\hbar}\mathrm{Vect}\bigl(\tilde A_\Phi^{\natural}\tilde A\bigr) $$
has two terms of different kinds: a derivative of the field and a product of two fields. The first has a symbol and the second does not, because a product is a convolution in frequency space and a convolution is not a multiplier.
The parent's measurement of the second term is the reason it cannot be dropped. Over one hundred random draws at unit amplitude the term ranges from $0.7$ to $12$ in the units of the display, so it is not a small correction to the derivative term; it is comparable to it and sometimes larger. A frequency-space account that kept only the derivative term would be wrong by an $O(1)$ amount, and the article records the range rather than a mean.
In frequency space the product is a convolution and it is the obstruction's second appearance. The first appearance is the symbol's $\frac{i\varkappa}{\hbar}F$, which is a multiplier because a constant curvature is a constant; the second is the product term, which is not a multiplier because the potential is a function. The two are one object seen at two levels of resolution, and the article says so once and does not claim they are numerically equal.
What the transform can say about the product is its support, and the companions' bench gives the numbers. On a $256$-mode circle, a delta is a multiplier and occupies one mode; a derivative of a delta is first order and occupies a small window; a general function occupies all $256$. A product with a general potential therefore occupies every mode, which is the frequency-space statement of "the covariant field strength is not a local operator in the free field".
The Ladder and Its Floor
The companions' bench measures how far a multiplication can be from a diagonal action, and the floor is the derivative. The three numbers are the three rungs: a delta is a multiplier and occupies one mode of $256$; a derivative of a delta is first order; a function is a convolution and occupies all $256$. Covariant differentiation is the derivative's rung, so it is first order and never diagonal.
The article's use of the bench is a floor and not a measurement. A connection's curvature term is a multiplier at constant curvature, so it sits on the delta's rung there; but the covariant derivative that produced it sits on the derivative's rung, and the product term sits on the function's. The three rungs are occupied simultaneously in one expression, which is why the coupled theory has a symbol in one limit and not in general.
The floor is the reason the free theory's exactness is not misleading. The free Proca symbol is exact because the coefficients are constant; the coupled one is exact only where the connection is, and the bench says how much of the structure the exactness buys: everything at constant curvature, the translation term at every curvature, and nothing else.
The rungs are the companions' and the article does not rerun them. The Fourier Transform and the Coupled Klein–Gordon Equation in Biquaternionic Form and The Fourier Transform and the Coupled Dirac Equation in Biquaternionic Form establish the occupancy counts, and the article quotes them because the Proca case adds no rung. The Proca case's addition is the second matrix index: the field strength has two, so its product term is a matrix product and its convolution is a convolution of matrices, and the occupancy count is the same.
Limits
The constant-connection licence bounds the symbol and nothing else. The translation term is exact for a constant connection; the obstruction is a multiplier only where the curvature is constant. Every statement about the symbol above carries that condition, and the article repeats it because the condition is easy to lose in a display.
The Abelian case is the one treated. The commutator's value is the curvature and there is no second term from the algebra's non-commutativity, because the coupling is Abelian. A non-Abelian coupling would add a term to the field strength and the symbol would acquire a second obstruction, and the article does not treat that case.
The mass's independence is a statement about the symbol's structure and not about the spectrum's values. The values move with the coupling through $\pi$; the multiplicities do not. A reader who took "independent" for "invisible" would be reading the wrong half of the sentence.
The denial, once more. No sector is exchanged. The coupled field is one element with eight real parameters, its transform is $\mathbb{B}$-valued, and the symbol is a matrix only at a constant connection. None of the three is a statement about $\mathbb{B}_\Phi$ and $\mathbb{B}_\Psi$.
Open Questions
Whether a frequency-space account of the obstruction exists at all is open. The symbol's curvature term is exact at constant curvature, and the product term is a convolution; whether a transform-like object exists that diagonalises the second is not known to the article, and the companions raise the same question for their own fields.
Whether the extra term's range $[0.7,12]$ has a bound is open. The parent measures the range over random draws at unit amplitude and does not bound it. A bound in terms of the amplitude's norm and the coupling would turn the range into an inequality, and nothing above supplies one.
Whether the mass's independence survives a non-Abelian coupling is open. An obstruction of a different kind would have a different trace, and the trace's vanishing is what the constraint's survival rests on. The article's independence argument uses the trace and would need redoing.
The covariant Lorenz condition's exactness is the parent's and it is quoted, not proved here. The article's §The Symbol in Two Terms reads it as a trace; whether the trace argument and the parent's divergence argument are the same argument is a question about the two orders of differentiation, and it is not answered here.
Summary
The coupled Proca operator's symbol is two terms. The first is the free symbol at the shifted four-momentum $\pi_\mu=\hbar k_\mu-q A_\mu$, which is the translation; the second is $\frac{i\varkappa}{\hbar}F^{(\Phi)}$, which is the commutator of two covariant derivatives and is the obstruction. The decomposition is the parent's square-splitting identity, transformed term by term, and there is no third term because the identity has two.
At a constant connection the obstruction vanishes identically and the symbol is $-\pi^{2}$ times the identity: the translation is exact, the dispersion is $\pi^{2}=k_m^{2}$, and a constant potential moves the axis of the symbol's parity to $\omega=-\varkappa c\varphi/\hbar$.
At a varying connection the obstruction is not removable. Two connections can agree in $\pi$ at a point and differ in $F$ there — the example is $A^{(1)}=0$ against $A^{(2)}$ with $A_1=b\,r_2'$, whose $\lvert F_{21}\rvert=0.9$ while both connections vanish at the origin. So the symbol at a point is not a function of the momentum at that point, and a Fourier description is wrong by exactly the curvature.
The mass does not enter the obstruction: the mass term is $-k_m^{2}\delta$, whose trace is $-4k_m^{2}$, and the obstruction is traceless at every $k_m$. The coupling therefore cannot change the count, only the values and the eigenvectors, and the form companion's rank table survives with $\pi$ in place of $k$. The field strength's extra term, the parent's product of two fields ranging from $0.7$ to $12$, is the same object in the other register, and in frequency space it is a convolution: the bench's ladder reads one mode for a delta, first order for its derivative, and all $256$ for a general function.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\varkappa$ | The coupling; a length whose inverse multiplies the potential |
| $A_\mu$ | The background potential |
| $\pi_\mu=\hbar k_\mu-q A_\mu$ | The shifted four-momentum; the symbol's argument |
| $k_t'=\omega/c$, $k^{2}=k_t'^{2}-\vec k_\Phi^{2}$ | The free four-wave-number and its interval |
| $k_m=mc/\hbar$ | The mass wave number |
| $\sigma_{\mathrm{gco}}=-\pi^{2}+\frac{i\varkappa}{\hbar}F^{(\Phi)}$ | The coupled symbol; translation plus obstruction |
| $F^{(\Phi)}_{\mu\nu}$ | The curvature; $\lvert F_{21}\rvert=0.9$ in the example |
| $\frac{i\varkappa}{\hbar}\mathrm{Vect}(\tilde A_\Phi^{\natural}\tilde A)$ | The field strength's extra term; $0.7$ to $12$ over $100$ draws |
| $-\pi^{2}$ | The trace part; the constraint's symbol |
| $256$ | The mode count of the companions' bench; one, a few, all |
Further Reading
- The Coupled Proca Equation in Biquaternionic Form, for the carrier argument, the minimal substitution, the covariant field strength and the constant background.
- The Coupled Proca Equation in Biquaternionic Polar Variables, for the same theory read in the chart, and for the gauge transformation as a phase translation.
- Gauge Fields and the Covariant Derivative, for the commutator, the curvature and the square-splitting identity.
- The Fourier Transform and the Proca Equation in Biquaternionic Form, for the free symbol's determinant, rank and kernel.
- The Fourier Transform and the Coupled Klein–Gordon Equation in Biquaternionic Form, for the occupancy bench and the ladder's rungs.
- The Fourier Transform and the Coupled Dirac Equation in Biquaternionic Form, for the same ladder with a first-order operator.
- The Fourier Transform and the Coupled Proca Equation in Biquaternionic Polar Variables, for the chart reading of this symbol.
- Derivatives, Fields and Transforms, for the symbol rule.
- M. Nakahara, Geometry, Topology and Physics (Adam Hilger, 1990), Chapter 10, for the curvature as a commutator and the square splitting.
- L. H. Ryder, Quantum Field Theory (Cambridge University Press, 2nd ed., 1996), Chapter 3, for the covariant derivative's symbol and the constant-background dispersion.