The Feynman Propagator in Biquaternionic Form
Introduction
The companion article Canonical Quantization of the Biquaternion Dirac Field quantizes the biquaternion Dirac field: it promotes the field to an operator-valued distribution, imposes the equal-time anticommutators, expands the field in its plane-wave solutions, and builds the Fock space. It closes with an explicit accounting of what is standard and what remains open, and it concludes that the construction is a transcription of ordinary canonical quantization rather than a derivation from the algebra $\mathbb{B}$.
The four-momentum is $\tilde P_{\Phi}=i(E/c)\,e_t+\vec p$, so that $\tilde P_{\Phi}\tilde P_{\Phi}^\natural= -m^{2}\,c^{2}$ on the mass shell with $k_m=mc/\hbar$ the mass wave number.
The dynamics is the one-sided chiral pair
$$ \tilde\nabla_\Phi\tilde{Ξ}_{R}=k_m\tilde{Ξ}_{L},\qquad (\tilde\nabla_\Phi)^{\natural}\tilde{Ξ}_{L}=k_m\tilde{Ξ}_{R} . $$
Here $k_m=mc/\hbar$ is the mass wave number, the inverse of the mass length.
The framework carries two equations called the Dirac equation, and they must not be read as one. The pair displayed above is the first: it is the definition. The gradient acts by left multiplication, the field is the element $\tilde{Ξ}=\tilde{Ξ}_{L}+\tilde{Ξ}_{R}$ of $\mathbb{B}$, and the mass is the off-diagonal pairing of the two parts named by the subscript, $L$ and $R$ (Special Element Conventions, §The Chirality Convention).
The second is the two-sided form of The Dirac Equation in Biquaternionic Form, a single equation on $\mathbb{B}$,
$$ \mathcal{N}F=\Big(i\frac{\partial}{\partial r'_t}+D_3\Big)F-k_mi\,\overline{F}\,\vec e_z=0, $$
in which $D_3$ is the spatial Dirac operator of that article, the mass acts by right multiplication through $\vec e_z$, and the conjugation supplies the antilinear part. Because $i\partial/\partial r'_t+D_3=-(\tilde\nabla_\Phi)^{\natural}$, the two-sided operator is
$$ \mathcal{N}=-(\tilde\nabla_\Phi)^{\natural}-k_mi\,\overline{\,\cdot}\,\vec e_z . $$
That display is the whole relation between the two equations, and it says three things. The leading term is the $\natural$-member of the pair, not a new operator. The mass sits on the far side, and it reaches the field only through its conjugate: comparing $(\tilde\nabla_\Phi)^{\natural}\tilde{Ξ}_{L}=k_m\tilde{Ξ}_{R}$ with $\mathcal{N}F=0$ reads the second field as $\tilde{Ξ}_{R}=-i\overline{F}\,\vec e_z$, a conjugation and a fixed direction and nothing else. The price is that $\mathcal{N}$ is not left multiplication by an element of $\mathbb{B}$, which is why the two forms are not the same operator. The change of variable is a real-linear bijection, so the two forms describe the same field; what changes is the shape of the operator, not the physics. The conjugation in the mass term pairs $F$ with $\overline{F}$, so a solution carries both frequency branches at once; in the time-harmonic case the parent article exhibits exactly that, as a sum of two exponentials of opposite sign.
Everything below is written for the one-sided pair, whose plane-wave solutions are elements of $\mathbb{B}$: the element $\tilde{Ξ}$, split by the subscript into its two parts $\tilde{Ξ}_{L}$ and $\tilde{Ξ}_{R}$, and no second carrier is introduced to hold them.
This article is about the next object in that quantization: the two-point function of the quantized field, the Feynman propagator. It is an action on $\mathbb{B}$, read on a test element $\tilde Q$,
$$ \boxed{\;S_F(x-y):\ \tilde Q\ \longmapsto\ \big\langle 0\big|\,T\,\tilde{Ξ}(x)\,\big(\tilde{Ξ}^*(y)\,i\,\tilde Q\,\vec e_x\big)\,\big|0\big\rangle\;}, $$
with $T$ the fermionic time-ordering operator. The adjoint of the field enters as the Hermitian conjugate $\tilde{Ξ}^*$, and the fixed pair $i\,\vec e_x$ that closes it on the test element is the timelike frame basis operator. The test element stands where the older display puts the object it multiplies; here it is an element of the algebra, and the propagator carries an element to an element.
The two-point function is where the boundary condition of the theory is stored. Its momentum-space amplitude is not a function but a distribution, and what makes it the Feynman amplitude rather than the retarded or advanced one is a contour prescription in the complex energy plane — the $i\epsilon$.
The specific question this article asks is a question about the algebra. The biquaternion framework already carries a complex structure: the scalar imaginary $i$ is central and fixed (Algebra and Element Conventions), it is what turns the material time coordinate into $i r'_t$, and it is what lets the mass shell be written as the single quadratic condition $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}= -m^{2}\,c^{2}$ whose two roots in the energy are the positive- and negative-frequency branches (the particle and antiparticle modes in the standard interpretation). The Feynman prescription is also an insertion of $i$: $m^2\mapsto m^2-i\epsilon$. Are these the same $i$ doing the same work? Is the contour deformation the algebra's own complex structure, or is it an independent analytic input that the algebra can only write down?
We answer this on a concrete case, by computing the pole structure and the contour explicitly and by recomputing every displayed formula on a momentum chosen for the purpose. The finding is stated at the outset because it is the article's content.
The algebra supplies the complex plane in which the contour is drawn, and a natural notation for the prescription — the deformed mass-shell scalar $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}+m^{2}\,c^{2}-i\epsilon$, whose added term $-i\epsilon\,e_t$ is a central element along the $i r'_t$ axis of the material sector — but it does not select the contour.
The deformation's axis is algebraically natural; its orientation, which is what distinguishes the Feynman propagator from the retarded and advanced ones, is an analytic boundary condition. In this problem the algebra adds notation and a home for the $i\epsilon$, not the physics that fixes it. If that is all it adds, we say so.
The Two-Point Function of the Quantized Dirac Field
The companion article quantizes the field on its module; here the same modes are read as elements of $\mathbb{B}$, and the operator-valued field is
$$ \tilde{Ξ}_{\mathrm{fop}}(x)=\int\!\frac{d^3p}{(2\pi\hbar)^3}\sqrt{\frac{\hbar c}{2E_{\vec p}}} \sum_{r=1}^{2}\Big[a_{\mathrm{fop},r}(\vec p)\,\tilde w^{(r)}(\vec p)\,\exp(-i\theta_{\mathrm{pwp}}) +b_{\mathrm{fop},r}^{\dagger}(\vec p)\,\tilde v^{(r)}(\vec p)\,\exp(+i\theta_{\mathrm{pwp}})\Big], $$
Here $\theta_{\mathrm{pwp}}:=p\cdot x/\hbar=\omega_{\vec p}r'_t-\vec p\cdot\vec r$ is the plane-wave angle in momentum notation: $p$ and $x$ are real, so the exponent is $i$ times a real number and the mode factor is a phase of modulus one.
with $E_{\vec p}=+\sqrt{\vec p^{\,2}c^2+m^{2}\,c^{4}}$, and with $\tilde w^{(r)}$ and $\tilde v^{(r)}$ the plane-wave elements of that momentum: the two solutions of the chiral pair and their two negative-frequency partners, each an element of $\mathbb{B}$. Its adjoint is the Hermitian conjugate $\tilde{Ξ}_{\mathrm{fop}}^*(x)$, carried with the same fixed pair $i\,\vec e_x$. That pair sits inside the time-ordered product, between the adjoint field and the test element, and the contraction over the two solutions consumes it: what acts on the test element is the kernel $\mathcal{A}_p+mc$ derived below, and no basis operator survives in the amplitude.
The momentum action. The four frame basis operators of Frame and Metric Conventions, written out as actions instead of as basis operators, are
$$ G_t:\ \tilde Q\longmapsto i\,\tilde Q\,\vec e_x, \qquad G_\eta:\ \tilde Q\longmapsto i\,\vec e_\eta\,\tilde Q\,\vec e_y\quad(\eta=1,2,3), $$
and the momentum action is their contraction with the four-momentum,
$$ \mathcal{A}_p:\ \tilde Q\ \longmapsto\ i\,p_0\,\tilde Q\,\vec e_x-i\sum_{\eta=1}^{3}p_\eta\,\vec e_\eta\,\tilde Q\,\vec e_y \;=\;p_0\,G_t(\tilde Q)-\sum_{\eta=1}^{3}p_\eta\,G_\eta(\tilde Q). $$
The energy enters on the right through $\vec e_x$; each momentum component enters on the left through $\vec e_\eta$ and on the right through $\vec e_y$. That asymmetry is the reference frame's, not a slip of notation: it is what makes the action square to the Lorentz square of the momentum.
The four squares do not share a sign. The signs of $G_t$ and $G_\eta$ are read from one application followed by another, and they are the plain weight $\varepsilon$ of Frame and Metric Conventions:
$$ G_t^2=\mathrm{id},\qquad G_\eta^2=-\mathrm{id},\qquad \{G_\mu,G_\nu\}=2\varepsilon_\mu\delta_{\mu\nu}\,\mathrm{id}, $$
with $\mathrm{id}$ the identity. The energy term squares to $+1$ because two minus signs meet, $i^2=-1$ from the central unit and $\vec e_x^2=-e_t$ from the quaternion,
$$ G_t:\ \tilde Q\longmapsto i^2\,\tilde Q\,\vec e_x^2=(-1)(-1)\,\tilde Q=+\tilde Q, $$
while a spatial term squares to $-1$, because three signs meet there, the same $i^2$ together with $\vec e_\eta^2=-e_t$ and $\vec e_y^2=-e_t$,
$$ G_\eta:\ \tilde Q\longmapsto i^2\,\vec e_\eta^2\,\tilde Q\,\vec e_y^2=(-1)(-1)(-1)\,\tilde Q=-\tilde Q . $$
The cross terms vanish because each action puts an invertible element on the right, and the two possible orders differ by $\vec e_y\vec e_x=-\vec e_x\vec e_y$ for the energy term against a spatial one; two distinct spatial terms cancel the same way, because distinct spatial units anticommute. Hence
$$ \mathcal{A}_p^2=-N(\tilde Q_{\Phi})\,\mathrm{id},\qquad N(\tilde Q_{\Phi})=-p_0^2+\vec p^2 . $$
The body of this article is written in these actions; the closing section gives their standard names.
The mode operators obey
$$ \{a_{\mathrm{fop},r}(\vec p),a_{\mathrm{fop},s}^{\dagger}(\vec p^{\,\prime})\} =\{b_{\mathrm{fop},r}(\vec p),b_{\mathrm{fop},s}^{\dagger}(\vec k_\Phi)\} =(2\pi\hbar)^3\delta_{rs}\delta^{(3)}(\vec p-\vec p^{\,\prime}), $$
with all other anticommutators vanishing. The plane-wave elements of the parent article obey the normalizations it records there, and the two sums over its two solutions are the numerators of this article, read as actions on $\mathbb{B}$: the positive-frequency sum is the momentum action shifted up by the mass, $\mathcal{A}_p+mc$, and the negative-frequency sum is the same action shifted down, $\mathcal{A}_p-mc$, both on shell.
The separation angle is $\theta_{\mathrm{sep}}=p\cdot(x-y)/\hbar$, real because the momentum and the separation are real. Contracting the mode operators against the vacuum gives the two elementary two-point actions. The positive-frequency piece is
$$ W_+(x-y):\ \tilde Q\ \longmapsto\ \int\!\frac{d^3p}{(2\pi\hbar)^3}\frac{1}{2E_{\vec p}}\,\big(\mathcal{A}_{p}(\tilde Q)+mc\tilde Q\big)\,\exp(-i\theta_{\mathrm{sep}}), $$
in which only the $a_{\mathrm{fop}} a_{\mathrm{fop}}^{\dagger}$ contraction survives, and the negative-frequency piece is
$$ W_-(x-y):\ \tilde Q\ \longmapsto\ \int\!\frac{d^3p}{(2\pi\hbar)^3}\frac{1}{2E_{\vec p}}\,\big(\mathcal{A}_{p}(\tilde Q)-mc\tilde Q\big)\,\exp(+i\theta_{\mathrm{sep}}), $$
in which only $b_{\mathrm{fop}} b_{\mathrm{fop}}^{\dagger}$ survives. In both expressions the momentum action is on shell, because the free mode expansion contains the on-shell modes and nothing else. These two objects are not yet the propagator; each is a boundary value of an analytic function, and it is the time ordering that combines them into $S_F$.
Time ordering for a fermion field carries the anticommutation sign:
$$ S_F(x-y)=\mathrm{H}_{\mathrm{hvs}}(x_0-y_0)\,W_+(x-y)-\mathrm{H}_{\mathrm{hvs}}(y_0-x_0)\,W_-(x-y), $$
an equality of actions, read on a generic element. The relative minus is not a convention; it is the statement that the two field operators anticommute. Every sign in what follows traces back to it, and it is the quantity a recomputation must reproduce.
The Pole Structure and the Contour
The two $\mathrm{H}_{\mathrm{hvs}}$ functions can be represented by a single four-dimensional integral. The standard result, and the object whose biquaternion form is the subject of this article, is
$$ \boxed{\;S_F(x-y):\ \tilde Q\longmapsto i\int\!\frac{d^4p}{(2\pi\hbar)^4}\; \exp(-i\theta_{\mathrm{pwp}})\frac{\mathcal{A}_p(\tilde Q)+mc\tilde Q}{-\big(N(\tilde Q_{\Phi})+m^{2}\,c^{2}\big)+i\epsilon}, \qquad \epsilon\to0^+\;}. $$
The action is written by what it does, so the display carries the test element and no second symbol. The overall $i$ and the $i\epsilon$ are fixed by requiring that the $p_0$ integral reproduce the two branches, with the normalization of the time-ordered product and the fermionic relative sign. It is worth performing that check explicitly, because it is where a sign can silently go wrong.
Write the $p_0$ integral at fixed $\vec p$, and put $\delta t=x_0-y_0$ and $\delta\vec s_{\Psi}=\vec s_{\Psi}-\vec y$. Since
$$ -\big(N(\tilde Q_{\Phi})+m^{2}\,c^{2}\big)=p_0^2-\vec p^2-m^{2}\,c^{2}=p_0^2-\frac{E_{\vec p}^2}{c^{2}}, $$
the integrand has poles where
$$ \boxed{\;p_0^2-\frac{E_{\vec p}^2}{c^{2}}+i\epsilon=0, \qquad\text{i.e.}\qquad p_0=\pm\sqrt{\frac{E_{\vec p}^2}{c^{2}}-i\epsilon} =\pm\Big(\frac{E_{\vec p}}{c}-\frac{i\epsilon\,c}{2E_{\vec p}}\Big)+O(\epsilon^2)\;}. $$
The sign is worth fixing once explicitly. Let $p_0=E_{\vec p}/c+\delta$. Then the denominator is $2(E_{\vec p}/c)\,\delta+i\epsilon$, and it vanishes at $\delta=-i\epsilon\,c/(2E_{\vec p})$, which is below the axis. At $p_0=-E_{\vec p}/c+\delta$ the denominator is $-2(E_{\vec p}/c)\,\delta+i\epsilon$, and it vanishes at $\delta=+i\epsilon\,c/(2E_{\vec p})$, which is above it.
This is the whole content of the prescription: the two poles are displaced in opposite directions. The decay of $\exp(-ip_0\delta t)$ then selects one pole or the other:
- for $\delta t>0$, close the contour in the lower half-plane and pick up the pole at $+E_{\vec p}/c$;
- for $\delta t<0$, close in the upper half-plane and pick up the pole at $-E_{\vec p}/c$.
Write $p_+=(E_{\vec p}/c,\vec p)$ and $p_-=(-E_{\vec p}/c,\vec p)$. The residues are
$$ \mathrm{Res}_{p_0=+E_{\vec p}/c}\frac{\mathcal{A}_p(\tilde Q)+mc\tilde Q}{p_0^2-E_{\vec p}^2/c^{2}} =\frac{\mathcal{A}_{p_+}(\tilde Q)+mc\tilde Q}{2E_{\vec p}/c}, \qquad \mathrm{Res}_{p_0=-E_{\vec p}/c}\frac{\mathcal{A}_p(\tilde Q)+mc\tilde Q}{p_0^2-E_{\vec p}^2/c^{2}} =\frac{\mathcal{A}_{p_-}(\tilde Q)+mc\tilde Q}{-2E_{\vec p}/c}. $$
The contour contributes $-2\pi i$ when it closes below and $+2\pi i$ when it closes above. With the residues above and the overall factor $i$ of $S_F$, the two cases are
$$ \delta t>0:\quad i\,(-2\pi i)\,\frac{\mathcal{A}_{p_+}(\tilde Q)+mc\tilde Q}{2E_{\vec p}/c} =2\pi\,\frac{\mathcal{A}_{p_+}(\tilde Q)+mc\tilde Q}{2E_{\vec p}/c}, $$
$$ \delta t<0:\quad i\,(+2\pi i)\,\frac{\mathcal{A}_{p_-}(\tilde Q)+mc\tilde Q}{-2E_{\vec p}/c} =2\pi\,\frac{\mathcal{A}_{p_-}(\tilde Q)+mc\tilde Q}{2E_{\vec p}/c}, $$
in both lines the product of the two imaginary units being $i^2=-1$.
Both carry the same positive weight: the sign of the half-plane and the sign inside the residue cancel, and the $2\pi$ cancels against the $1/(2\pi)$ of the $p_0$ measure, leaving $c/(2E_{\vec p})$, which is the normalisation of $W_\pm$. The fermionic minus is therefore not in the contour, and it must come from the numerator. For $\delta t>0$ the $p_0$ integral is already the positive-frequency integrand of $W_+$. For $\delta t<0$ it is
$$ \frac{\mathcal{A}_{p_-}(\tilde Q)+mc\tilde Q}{2E_{\vec p}/c}\,\exp\!\big(+iE_{\vec p}\delta h_t/(\hbar c)+i\vec p\cdot\delta\vec s_{\Psi}/\hbar\big). $$
Now flip the integration variable, $\vec p\to-\vec p$: the measure and $E_{\vec p}$ are unchanged, and the spatial phase reverses. The action is linear in the four-momentum, so flipping all four components flips its sign,
$$ \mathcal{A}_{(-E_{\vec p}/c,-\vec p)}=-\mathcal{A}_{(E_{\vec p}/c,\vec p)}, $$
and the numerator of the flipped expression becomes
$$ \mathcal{A}_{(-E_{\vec p}/c,-\vec p)}(\tilde Q)+mc\tilde Q =-\big(\mathcal{A}_{(E_{\vec p}/c,\vec p)}(\tilde Q)-mc\tilde Q\big), $$
which is the negative-frequency numerator of $W_-$ with a minus. The single covariant expression, with its single factor of $i$ in front, therefore reproduces both the time ordering and the fermionic minus.
Three things are conflated in the phrase "the $i\epsilon$ does the time ordering", and they are worth separating.
- The $i\epsilon$ fixes which pole lies in which half-plane; that is the boundary condition, and its sign is the choice of propagator.
- The overall $i$ fixes the normalization of the amplitude relative to the two $\mathrm{H}_{\mathrm{hvs}}$ terms.
- The minus in front of $W_-$ is carried by the numerator, through the two numerators $\mathcal{A}_p+mc$ and $\mathcal{A}_p-mc$. It is the two-point form of the anticommutation, and it has no counterpart in the scalar case, where the two residues share one numerator and the time-ordered combination is $\mathrm{H}_{\mathrm{hvs}}W_++\mathrm{H}_{\mathrm{hvs}}W_-$.
A recomputation on a case chosen for it. Take $m\,c^{2}=0.7$ and $\vec p=(0.3,-0.9,1.1)\,m\,c$, so that $|\vec p|=1.4525839046\,m\,c$ and $E_{\vec p}=1.6124515497\,m\,c^{2}$; every momentum below is quoted in units of $m\,c$ and every energy in units of $m\,c^{2}$. Applying the four coordinate actions twice on a hundred random elements, the identity $\mathcal{A}_p^2=-N(\tilde Q_{\Phi})\,\mathrm{id}$ holds to better than $10^{-14}$ and $(\mathcal{A}_p-mc)(\mathcal{A}_p+mc)=-\big(N(\tilde Q_{\Phi})+m^{2}\,c^{2}\big)\,\mathrm{id}=-\mathcal{M}(\tilde Q_{\Phi})\,\mathrm{id}$ to the same, the anticommutator $\{G_\mu,G_\nu\}=2\varepsilon_\mu\delta_{\mu\nu}\,\mathrm{id}$ holds exactly on the grid tested, and $\mathcal{A}_{-p}=-\mathcal{A}_p$ exactly. The identity
$$ (\mathcal{A}_p-mc)(\mathcal{A}_p+mc)=\mathcal{A}_p^2-m^{2}c^{2}\,\mathrm{id}=-\big(N(\tilde Q_{\Phi})+m^{2}\,c^{2}\big)\,\mathrm{id} $$
is not a coincidence: it is the square of the action derived above, and it holds for off-shell $p$ as well. For $\epsilon=10^{-3}$ the Feynman poles come out at
$$ p_0=+1.61245158-i\,3.101\times10^{-4} \qquad\text{and}\qquad p_0=-1.61245158+i\,3.101\times10^{-4}, $$
that is, one on each side of the real axis, at the distance $\epsilon\,c/(2E_{\vec p})$ the boxed display predicts.
Feynman, Retarded, and Advanced: What Selects the Contour
The pole structure makes the choice of propagator explicit. Three prescriptions are available, and they differ only in how the two poles are displaced:
| Prescription | Denominator | Poles (to first order in $\epsilon$) |
|---|---|---|
| Feynman | $p_0^2-E_{\vec p}^2/c^{2}+i\epsilon$ | $\pm E_{\vec p}/c\mp i\epsilon\,c/(2E_{\vec p})$ (opposite sides) |
| Retarded | $(p_0+i\epsilon)^2-E_{\vec p}^2/c^{2}$ | $\pm E_{\vec p}/c-i\epsilon$ (both below) |
| Advanced | $(p_0-i\epsilon)^2-E_{\vec p}^2/c^{2}$ | $\pm E_{\vec p}/c+i\epsilon$ (both above) |
The three are distinguished by whether the displacements are opposite (Feynman) or common (retarded and advanced). The displacement in the first row is smaller by the factor $2E_{\vec p}$ because that prescription shifts the denominator by $i\epsilon$ while the other two shift $p_0$ by $\pm i\epsilon$; only the pattern of sides matters for the contour. For $\epsilon=10^{-3}$ and the same momentum, the retarded poles are at $1.61245124-i\,10^{-3}$ and $-1.61245124-i\,10^{-3}$, and the advanced poles at the same points with $+i\,10^{-3}$: both on the same side, as the table states. Closing the contour then shows at once why the retarded propagator vanishes for $x_0-y_0<0$ and the advanced one for $x_0-y_0>0$: with both poles on one side, one of the two half-planes contains no pole.
The companion The Klein–Gordon Propagator and Its Green's Functions in Biquaternionic Form lists the same three displacements in the frequency $\omega$ and with the denominator written as $\vec p^2+\mu^2-\omega^2$, the negative of the one here; the pole positions agree.
This is the physical content of the $i\epsilon$, and it is exactly the content the algebra does not supply. The Feynman choice encodes the time-ordered vacuum correlation; the retarded choice encodes causal response to a source; the advanced choice is its mirror. They solve the same differential equation off the source and differ only in support. A choice among them is a boundary condition on the distributional inverse, and boundary conditions are not algebraic data. Maxwell's Equations in the Biquaternionic Formulation meets the same situation: it selects the retarded kernel as "the physically relevant" one, and it states that the retarded kernel is a different object from the elliptic Cauchy kernel of the analysis articles, the two being related by the Wick rotation. The selection there is made on physical grounds, not read off the algebra, and the same is true here.
The Biquaternion Form of the Prescription
The Feynman denominator has a compact biquaternion form. With the four-momentum biquaternion $\tilde P_{\Phi}=i(E/c)\,e_t+\vec p$, where $E$ is the energy and $p_0=E/c$ is the variable the contour integral runs over, the quaternion conjugate is $\tilde P_{\Phi}^{\natural}=i(E/c)\,e_t-\vec p$, and the quaternion product is
$$ \tilde P_{\Phi}\tilde P_{\Phi}^{\natural}=(iE/c)^2+\vec p^2=-E^{2}/c^{2}+\vec p^{\,2}=N . $$
Two signs enter, and they come from two different places. The scalar one is $i^2=-1$, which takes $(iE/c)^2$ to $-E^2/c^2$. The vector one is a plus: the vector part multiplies $-\vec p$, and $\vec p\,(-\vec p)=+\vec p^2$, because the quaternion square of a vector is $\vec p\,\vec p=-\vec p^2$ with $\vec p^2=p_x^2+p_y^2+p_z^2$ the Euclidean square. The last step is $N=-E^{2}/c^{2}+\vec p^{\,2}$.
The mass shell is then the single algebraic condition
$$ \tilde P_{\Phi}\tilde P_{\Phi}^{\natural}= -m^{2}\,c^{2}\,\Longleftrightarrow\,-E^2/c^2+\vec p^{\,2}=-m^{2}\,c^{2} \;\Longleftrightarrow\; E^{2}=\vec p^{\,2}c^2+m^{2}\,c^{4}, $$
whose two frequency branches $E=\pm E_{\vec p}$ are the two poles $p_\pm$ of section The Pole Structure and the Contour. Substituting $N(\tilde Q_{\Phi})=\tilde Q_{\Phi}\tilde Q_{\Phi}^{\natural}$ in the amplitude gives the biquaternion form,
$$ \boxed{\;S_F(p):\ \tilde{Q}\,\longmapsto-\frac{i\big(\mathcal{A}_p(\tilde Q)+mc\tilde Q\big)}{N(\tilde Q_{\Phi})+m^{2}\,c^{2}-i\epsilon} \;=\;-\,\frac{i\big(\mathcal{A}_p(\tilde Q)+mc\tilde Q\big)}{\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}+m^{2}\,c^{2}-i\epsilon}\;}, $$
the denominator being the deformed mass-shell operator $\mathcal{M}(\tilde Q_{\Phi})=N(\tilde Q_{\Phi})-N(\tilde P_{\Phi})=N(\tilde Q_{\Phi})+m^{2}\,c^{2}$, shifted by $-i\epsilon$. The two slots are distinct objects: $\tilde Q_{\Phi}\in\mathbb{B}_{\Phi}$ is the free element at which the operator is read, and $\tilde P_{\Phi}=m\tilde V_{\Phi}$ is the particle's own momentum, fixed, with $\langle\tilde V_{\Phi}\rangle_{\natural}=-c^{2}$ and $N(\tilde P_{\Phi})=-m^{2}c^{2}$. Everything the propagator does happens at a free $\tilde Q_{\Phi}$: the pole is the zero set $N(\tilde Q_{\Phi})=-m^{2}c^{2}$, a three-dimensional level set, and not a property of the fixed datum. The prescription is thus a deformation of the scalar on which the mass shell is defined, not an addition to the algebra.
The shift has a natural algebraic home. The deformation is $-i\epsilon\,e_t$: a central element, purely imaginary as a scalar coefficient and with no vector part. Being anti-Hermitian it lies on the material side, and being central it lies in $\mathbb{B}_{C}=\mathbb{C}e_t$. Since $\epsilon$ is real, $-i\epsilon\,e_t=-\epsilon\,(ie_t)$, so it is a real multiple of the $i r'_t$ direction, the direction in which the framework writes the material time. Both statements describe one element: the shift is imaginary relative to the scalar unit $e_t$ and real relative to the time direction $ie_t$, exactly as $Q_t=i q'_t$ is.
One caveat belongs here, because it is the kind of sign that hides in prose. The Feynman term is added to the scalar $\mathcal{M}(\tilde Q_{\Phi})$, and as an element it is anti-Hermitian. Displacing the four-momentum biquaternion is a different operation. With $\tilde P_{\Phi}\to\tilde P_{\Phi}+\delta e_t$,
$$ \tilde P_{\Phi}\tilde P_{\Phi}^{\natural}\;\longmapsto\;N+2i(E/c)\delta+\delta^2 , $$
so reproducing the term requires $\delta=-\epsilon c/(2E)$, which is real: the equivalent displacement of the momentum runs along the Hermitian direction, not the material one. The two readings agree on the pole positions to first order in $\epsilon$ and disagree on which sector the deformation lives in. The article's claim is the first: $\tilde P_{\Phi}$ is not displaced, the central scalar is, and it is that added term which lies in $\mathbb{B}_{\Phi}$. Read as a displacement of the momentum, the $i\epsilon$ would sit in the informational sector, because $\tilde P_{\Phi}$ carries its energy through $i$.
The algebra therefore does three things here, and it is worth separating them:
- It provides the complex plane in which the contour is drawn — the coefficient complex structure of $\mathbb{B}_{C}$.
- It locates the direction of the added term, the $i r'_t$ axis of $\mathbb{B}_{\Phi}$.
- It makes the undeformed mass shell a single algebraic condition $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}= -m^{2}\,c^{2}$, so that the two poles are the two branches of one equation rather than two unrelated objects.
What it does not do is choose the orientation. Replacing $\epsilon$ by $-\epsilon$ reverses the pole displacements, turning the Feynman propagator into the anti-Feynman one; replacing the opposite displacement by a common one gives the retarded or advanced propagator. All three denominators are built from the same $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}+m^{2}\,c^{2}$ and the same central $i$. The algebra cannot tell them apart, because the difference is not in the algebra; it is in which half-plane the contour closes, equivalently in which boundary condition defines the distribution. The $i\epsilon$ is therefore written with the algebra's complex structure and lives in its material sector, but it is not produced by either. The algebra names the axis; the physics chooses the direction.
The closest the complex structure comes to doing real work is the Wick rotation. On the material sector the d'Alembertian reads
$$ \Box_\Phi=-\frac{\partial^2}{\partial {r'_t}^2}+\Delta_\Phi , $$
a hyperbolic operator: one sign against three. The substitution $r'_t=-i\tau$ turns $\partial^2/\partial {r'_t}^2$ into $-\partial^2/\partial\tau^2$, so the same operator becomes
$$ \Box_\Phi \;\longmapsto\; \frac{\partial^2}{\partial\tau^2}+\Delta_\Phi , $$
which is elliptic: all four directions now carry the same sign, and that single sign flip is exactly what separates the hyperbolic reading from the elliptic one. The Euclidean Green's function of the elliptic operator is unique once one demands decay at infinity, and its analytic continuation back to real time is the Feynman propagator. In that route the complexification of time is exactly what makes the continuation available.
But the selection is still analytic, not algebraic. The step "continue the unique Euclidean Green's function" is the input: uniqueness rests on a decay condition in the Euclidean variables, and the direction of the rotation is a further choice, equivalent to the sign of $\epsilon$. Nothing in $\mathbb{B}$ prefers decaying-in-Euclidean over growing. So the Wick rotation exhibits the algebra's complex structure as a necessary carrier of the continuation, and the analytic input as the thing that fixes which continuation; it does not convert the algebra into a derivation of the Feynman prescription.
The Propagator as a Distribution
It matters that $S_F$ is a distribution, and not only because the statement is true. The integral defining it converges for no real $p$ on the mass shell; the mass shell is precisely the pole set, and the unregulated amplitude $(\mathcal{A}_p+mc)/\big(-\mathcal{M}(\tilde Q_{\Phi})\big)$ is undefined there. The $i\epsilon$ is what turns the algebraic inverse of $\mathcal{A}_p-mc$ off the mass shell into a distributional inverse on all of momentum space.
The two-point action of the chiral pair is the pair of two kernels, and it satisfies, in the distributional sense,
$$ \bigl(\tilde\nabla_\Phi+k_m\bigr)G_F \quad\text{and}\quad \bigl((\tilde\nabla_\Phi)^{\natural}+k_m\bigr)G_F, \qquad (\Box_\Phi-k_m^{2})\,G_F=i\,\delta^{(4)}(x-y), $$
where the first kernel is the first line of the pair applied to $G_F$ and the second kernel is the second line. Each kernel is annihilated by its own line with the sign of the mass flipped, $\bigl(\tilde\nabla_\Phi-k_m\bigr)(\tilde\nabla_\Phi+k_m)G_F=i\,\delta^{(4)}$ on the first and $\bigl((\tilde\nabla_\Phi)^{\natural}-k_m\bigr)((\tilde\nabla_\Phi)^{\natural}+k_m)G_F=i\,\delta^{(4)}$ on the second. The two lines and the momentum action are two writings of one first-order symbol, the first one line per part and the second by the four basis operator actions read on a single element; the change of variable of the introduction relates them, they are not the same map, and they share the pole set because each line composed with its partner is multiplication by $-N$. The overall factor is a normalization: the conventions article writes the Green matrix of the pair as the two entries $(\tilde\nabla_\Phi+k_m)G$ and $((\tilde\nabla_\Phi)^{\natural}+k_m)G$ with $(\Box_\Phi-k_m^{2})G=\delta$, whereas the time-ordered product of this article carries $S_F=i\mathcal{S}$; the companion Klein–Gordon article fixes $(\Box_\Phi-\mu^2)G=-\delta^{(4)}$ at $\mu=m$, so its $G$ is $i$ times the $G_F$ here. None of these normalizations affects a pole position or a support. The differential equation holds for the Feynman, retarded, and advanced choices alike, because the difference between them is supported only on the mass shell and carries no additional source off it. The differential equation, therefore, does not determine the propagator. What determines it is the time-ordering prescription, which in momentum space is the contour and in configuration space is the split of support into $x_0>y_0$ and $x_0 This is the sharpest way to see why the question "does the algebra supply the $i\epsilon$?" cannot be answered by finding an $i$ in the algebra. The equation of motion supplies a differential operator, and a differential operator has many distributional inverses; the algebra supplies the operator $\tilde\nabla_\Phi$ (or its second-order companion $\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}$) and hence the same many inverses. Distinguishing the Feynman one requires a condition that is not an equation of motion. The complex structure enters when that condition is represented — as a contour in $\mathbb{B}_{C}$, or as a point in the complexified time plane — and not when it is chosen. One further structure is worth recording, because it is the place where the framework's complex time does carry a genuinely analytic condition. The companion article The KMS Condition and the Biquaternion Framework characterizes thermal equilibrium by the analyticity of the correlation function $F_{AB}(r'_t)=\omega_\beta(A\,\alpha_{r'_{t}}(B))$ in the strip $0<\mathrm{Im}(r'_t)<\beta$, with boundary relation $F_{AB}(r'_t+i\beta)=F_{BA}(-r'_t)$; the imaginary-time direction is intrinsic to the material sector $\mathbb{B}_{\Phi}$, and the companion quantization article shows that the anticommutator turns the boundary relation into the Fermi–Dirac distribution and the antiperiodic Matsubara frequencies $\omega_n=(2n+1)\pi/\beta$. At finite temperature the time-ordered two-point function is not given by a vacuum contour integral but by the discrete Matsubara sum over the compact imaginary-time circle; the real-time retarded and advanced propagators still carry their own $\epsilon$ prescriptions, and they are obtained from the Matsubara sum by analytic continuation, not by an algebraic operation. The imaginary-time complexification that the algebra carries is a real feature of the thermal state — it is where the KMS analyticity lives — but it does not select the vacuum contour either. It is a consistency: the imaginary direction the framework already uses for the material time is the direction in which the thermal correlation functions are analytic, and the vacuum contour is recovered as the $\beta\to\infty$ limit rather than derived from the algebra. Standard field theory, transcribed. The mode expansion and the mode anticommutators of the companion article; the two elementary two-point actions $W_+$ and $W_-$ and their numerators; the covariant representation of $S_F$ with its contour; the identification of the pole displacements with the Feynman, retarded, and advanced propagators; the distributional equation $(\Box_\Phi-k_m^{2})S_F=i\delta$; and the finite-temperature reduction to the Matsubara sum. None of this is new, and none of it depends on the biquaternion structure beyond the kinematical conventions already fixed by the read-list articles. What the biquaternion notation provides. The deformed mass-shell scalar $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}+m^{2}\,c^{2}-i\epsilon$; the pole set as the two branches of one algebraic condition; and a place for the added term in the centre $\mathbb{B}_{C}$, along the $i r'_t$ axis of the material sector $\mathbb{B}_{\Phi}$ — a place, not a necessity, for the reason recorded above. What remains open in the framework. The Pauli and gamma formulation is legacy: the reasons are explained in the article The AXF Matrices: Axis-Free, Integer Entries, §Why the Pauli and the Gamma Displays Are Refused. The generators carry no frame label: $\gamma_\mu$ is one set of matrices, $\gamma_0^2=I_4$ and $\gamma_\eta^2=-I_4$. The frame belongs to the coordinate, not to the array. The gammas are a frame of $\mathrm{Cl}_{1,3}(\mathbb{C})=M_4(\mathbb{C})$, not a different algebra. Here $L_a$ and $R_b$ are the left and right multiplications, $L_a\tilde{Ξ}=a\tilde{Ξ}$ and $R_b\tilde{Ξ}=\tilde{Ξ} b$, and $c(M)$ is the coefficient matrix of $M$ in the basis of the sixteen products. $$
\gamma_0=iL_{\mathrm{pau},e_t}R_{\mathrm{pau},\vec e_x},\quad \gamma_1=iL_{\mathrm{pau},\vec e_x}R_{\mathrm{pau},\vec e_y},\quad \gamma_2=iL_{\mathrm{pau},\vec e_y}R_{\mathrm{pau},\vec e_y},\quad \gamma_3=iL_{\mathrm{pau},\vec e_z}R_{\mathrm{pau},\vec e_y},\quad \gamma_5=-iL_{\mathrm{pau},e_t}R_{\mathrm{pau},\vec e_z} .
$$ The same in the axis-free display: $$
\gamma_0=iL_{\mathrm{axf},e_t}R_{\mathrm{axf},\vec e_x},\quad \gamma_1=iL_{\mathrm{axf},\vec e_x}R_{\mathrm{axf},\vec e_y},\quad \gamma_2=iL_{\mathrm{axf},\vec e_y}R_{\mathrm{axf},\vec e_y},\quad \gamma_3=iL_{\mathrm{axf},\vec e_z}R_{\mathrm{axf},\vec e_y},\quad \gamma_5=-iL_{\mathrm{axf},e_t}R_{\mathrm{axf},\vec e_z} .
$$ A single product is a pure tensor; the products form the seven-dimensional cone $\operatorname{rank}c(M)\le1$, and a general element is a sum of at most four, with $\ell(M)=\operatorname{rank}c(M)$. $\gamma_5$ is the reference-frame basis operator of the chirality $\Gamma$; the central grading $\chi=i\tilde{Ζ}_{\mathrm{ad}}$ — the difference of the two central idempotents of the complexified algebra, so $\chi^2=e_t$ — is a different object, represented by $\gamma_5$ on the Dirac module. The dictionary is complex-linear; over $\mathbb{R}$ the products span $\mathrm{Cl}_{1,3}(\mathbb{R})\cong M_2(\mathbb{H})$ and the basis operators span $\mathrm{Cl}_{3,1}(\mathbb{R})\cong M_4(\mathbb{R})$. The body's momentum action then reads, in the gamma display, $$
\slashed p=\gamma_0p_0-\sum_\eta\gamma_kp_\eta ,
$$ the spatial signs being those of $\varepsilon=(1,-1,-1,-1)$. The frame satisfies the Clifford relation $$
\{\gamma_\mu,\gamma_\nu\}=2\varepsilon_\mu\delta_{\mu\nu}\mathbf{1},
$$ which is the anticommutator of the body, in its named form: $\gamma_\mu$ is the basis operator $iL_{\mathrm{pau},e_\mu}R_{\mathrm{pau},\vec e_{\nu(\mu)}}$, so this is the same statement. It is not an extra input. For $\mu\neq\nu$ the two products $\gamma_\mu\gamma_\nu$ and $\gamma_\nu\gamma_\mu$ differ by a single transposition — on the left when one index is $t$, on the right when both are spatial — and a transposition reverses a quaternion sign, so they cancel; for $\mu=\nu$ each square is $\pm\mathbf{1}$, the sign coming from $e_t^2=e_t$ and $\vec e_\eta^2=-e_t$. The display. Every array printed in this section is given twice: in the Peirce display first, labelled $\mathrm{pau}$, whose arrays are the ones the literature prints, and in the axis-free display after it, labelled $\mathrm{axf}$, which is the display of the body and of the corpus. The two are carried into one another by $L_{\mathrm{axf},e_\mu}=C_{\Pi}L_{\mathrm{pau},e_\mu}C_{\Pi}^{-1}$, and a conjugation by $C_{\Pi}$ changes no product, no anticommutator, no trace and no determinant, so every identity of the section holds in both. An array is a statement about one display; a basis operator named in the running text keeps the Peirce label, and its axis-free array is the conjugate. The label names the array and not the operator. The Feynman propagator of the biquaternion Dirac field is the time-ordered two-point action $$
S_F(x-y):\ \tilde Q\longmapsto\big\langle 0\big|\,T\,\tilde{Ξ}(x)\,\big(\tilde{Ξ}^*(y)\,i\,\tilde Q\,\vec e_x\big)\,\big|0\big\rangle .
$$ Its content is the contour prescription, and the prescription is fixed by time ordering together with the fermionic anticommutation sign: $$
S_F(x-y):\ \tilde Q\longmapsto -i\int\!\frac{d^4p}{(2\pi\hbar)^4}\,\exp(-i\theta_{\mathrm{pwp}})\frac{\mathcal{A}_p(\tilde Q)+mc\tilde Q}{N(\tilde Q_{\Phi})+m^{2}\,c^{2}-i\epsilon},
\qquad
\mathcal{A}_p:\ \tilde Q\longmapsto i p_0\tilde Q\,\vec e_x-i\sum_{\eta=1}^{3} p_\eta\,\vec e_\eta\,\tilde Q\,\vec e_y .
$$ The $p_0$ integral has poles at $p_0=\pm\big(E_{\vec p}-i\epsilon/2E_{\vec p}\big)$: the positive pole below the real axis and the negative pole above it. Closing below for $x_0-y_0>0$ and above for $x_0-y_0<0$ gives the two branches with the same positive weight $1/(2E_{\vec p})$, the sign of the half-plane cancelling the sign inside the residue and the $2\pi$ cancelling the measure; the fermionic relative minus is carried by the numerator: after $\vec p\to-\vec p$ the negative-frequency numerator obeys $\mathcal{A}_{(-E_{\vec p},-\vec p)}(\tilde Q)+mc\tilde Q=-(\mathcal{A}_{(E_{\vec p},\vec p)}(\tilde Q)-mc\tilde Q)$, which is the two-point form of the anticommutation. Writing the four-momentum biquaternion $\tilde P_{\Phi}=i(E/c)\,e_t+\vec p$, so that $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}=N$, the propagator amplitude becomes $$
S_F(p):\ \tilde Q\longmapsto-\frac{i\big(\mathcal{A}_p(\tilde Q)+mc\tilde Q\big)}{\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}+m^{2}\,c^{2}-i\epsilon},
$$ and the prescription is a deformation of the mass-shell scalar. The added term $-i\epsilon\,e_t$ is central and lies along the $i r'_t$ axis of the material sector $\mathbb{B}_{\Phi}$; read instead as a displacement of $\tilde P_{\Phi}$, the same shift would be Hermitian, so what is fixed is the place of the term, not a direction of the momentum. The Feynman, retarded, and advanced propagators differ only in whether the two pole displacements are opposite or common; the algebra provides the complex plane and the axis of the added term but not its orientation, which is a boundary condition on the distributional inverse. The algebra therefore adds notation and a natural home for the $i\epsilon$, not the selection of the Feynman contour.Relation to the Thermal Two-Point Function
What the Algebra Adds and What It Does Not
The Legacy Pauli Gamma Formulation
Summary
Summary of Notation
Symbol
Meaning
$\mathbb{B}=\mathbb{C}\otimes_\mathbb{R}\mathbb{H}$
Biquaternion algebra
$e_t,\vec e_x,\vec e_y,\vec e_z$
The quaternion units (Algebra and Element Conventions)
$i$
The central imaginary unit (Algebra and Element Conventions)
$\mathbb{B}_{\Phi},\mathbb{B}_{\Psi}$
Material (anti-Hermitian) and informational (Hermitian) sectors
$\mathbb{B}_{C}$
The centre of $\mathbb{B}$ (Algebra and Element Conventions)
$\mathrm{id}$
The identity, as an action on $\mathbb{B}$
$G_t$, $G_\eta$
The four frame basis operators written as actions, $\tilde Q\mapsto i\tilde Q\vec e_x$ and $\tilde Q\mapsto i\vec e_\eta\tilde Q\vec e_y$; $G_t^2=\mathrm{id}$, $G_\eta^2=-\mathrm{id}$, $\{G_\mu,G_\nu\}=2\varepsilon_\mu\delta_{\mu\nu}\mathrm{id}$
$\varepsilon$
The plain weight of the four directions, the square of the units (Frame and Metric Conventions)
$\mathcal{A}_p$
The momentum action, $p_0G_t-\sum_\eta p_kG_\eta$, written out as $\tilde Q\mapsto ip_0\tilde Q\vec e_x-i\sum_\eta p_\eta\vec e_\eta\tilde Q\vec e_y$; $\mathcal{A}_p^2=-N\mathrm{id}$
$\tilde\nabla_\Phi$
Biquaternionic gradient (Operator and Calculus Conventions); $Q_t$ is the time coordinate, $i r'_t$ on $\mathbb{B}_{\Phi}$
$\Box_\Phi$
The d'Alembertian, the gradient squared through its conjugate (Operator and Calculus Conventions)
$\tilde\nabla_\Phi\tilde{Ξ}_{R}=k_m\tilde{Ξ}_{L}$, $(\tilde\nabla_\Phi)^{\natural}\tilde{Ξ}_{L}=k_m\tilde{Ξ}_{R}$
The pair, as two first-order lines; each line composed with its partner is $\Box_\Phi$
$\mathcal{N}=-(\tilde\nabla_\Phi)^{\natural}-k_mi\,\overline{\,\cdot}\,\vec e_z$
Two-sided Dirac operator of The Dirac Equation in Biquaternionic Form
$D_3$
Its spatial part, defined in that article
$\tilde{Ξ}=\tilde{Ξ}_{L}+\tilde{Ξ}_{R}$
The biquaternion Dirac field, an element of $\mathbb{B}$, split into the two parts named by the subscript
$\tilde{Ξ}^*$, $\tilde{Ξ}^*i\vec e_x$
The Hermitian conjugate of the field, and the adjoint it forms with the fixed timelike basis operator
$\tilde w^{(r)}(\vec p)$, $\tilde v^{(r)}(\vec p)$
The plane-wave elements of the two frequency branches
$\mathcal{A}_p+mc$, $\mathcal{A}_p-mc$
The two numerators, on shell: the map form of the two sums over the solutions
$p=(E,\vec p)$, $p_\pm=(\pm E_{\vec p},\vec p)$, $N=-E^{2}/c^{2}+\vec p^{\,2}$
Four-momentum, the two pole momenta, and the Lorentz square
$\tilde{Ξ}_{\mathrm{fop}}$, $a_{\mathrm{fop},r}(\vec p)$, $b_{\mathrm{fop},r}(\vec p)$
Quantized field; particle and antiparticle modes
$W_\pm(x-y)$
Positive- and negative-frequency two-point actions
$S_F(x-y)$
Feynman propagator, an action on $\mathbb{B}$
$\tilde P_{\Phi}=i(E/c)\,e_t+\vec p$
Four-momentum biquaternion, $\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}=N$
$\tilde P_{\Phi}\tilde P_{\Phi}^{\natural}= -m^{2}\,c^{2}$
Biquaternion mass-shell condition
$\mathcal{M}(\tilde Q_{\Phi})=N(\tilde Q_{\Phi})-N(\tilde P_{\Phi})=N(\tilde Q_{\Phi})+m^{2}\,c^{2}$
Mass-shell operator; Feynman shift $\mathcal{M}(\tilde Q_{\Phi})-i\epsilon\,e_t$
$\beta=1/(k_BT)$, $\omega_\beta$, $\alpha_{r'_{t}}$
Inverse temperature, thermal state, Heisenberg evolution
$F_{AB}(r'_t+i\beta)=F_{BA}(-r'_t)$
KMS condition
$\omega_n=(2n+1)\pi/\beta$
Fermionic Matsubara frequencies
Further Reading