Conformal Invariance and the Massless Limit in Biquaternionic Form

Introduction

A massless relativistic field has a larger symmetry than a massive one. The Poincaré group is enlarged by the dilations and the special conformal transformations into the fifteen-parameter conformal group, and the massless wave equation, the massless Maxwell field, and the two-point function are all invariant under it. The mass term is the one term in the action that is not: it introduces a fixed length scale, and it breaks every one of the five additional symmetries. This article states that symmetry, derives its action on the massless scalar field, identifies the massless limit as the limit in which it holds, and traces the breaking to the trace of the stress–energy tensor. It is the symmetry statement that the action article makes possible and the massless limit of the operator article.

Three things are established. First, the transformation law. The massless scalar field of the framework transforms with conformal weight $\Delta = 1$, $$ \tilde{\Phi}(\tilde{Q}) \longmapsto \Omega(\tilde{Q})^{\Delta}\,\tilde{\Phi}\bigl(\tilde{Q}'\bigr) , $$ where $\Omega$ is the conformal factor and $\tilde{Q}'$ is the image point; with this weight the massless equation $\Box\tilde{\Phi} = 0$ is carried into itself. The weight is not free: the identity $$ \Box_{\tilde{Q}}\Bigl[\Omega(\tilde{Q})^{\Delta}\,\phi\bigl(\tilde{Q}'\bigr)\Bigr] = \Omega(\tilde{Q})^{\Delta+2}\,\bigl(\Box_{\tilde{Q}'}\phi\bigr)\bigl(\tilde{Q}'\bigr) $$ holds at $\Delta = 1$ and fails at every other weight, and it was verified on a superposition of two smooth test fields to a relative accuracy of $10^{-16}$ at the correct weight, with a residual of order unity at the wrong one. Second, the massless limit. The mass enters the equation as the single scale in $\Box - \mu^2$, and the limit $\mu\to0$ is the limit in which the conformal weight exists; algebraically it is the limit in which the on-shell symbol becomes a zero divisor of the algebra, left multiplication by the on-shell momentum dropping from rank four to rank two. Third, the breaking. Improving the stress–energy tensor makes its trace proportional to the mass term, $\Theta^\mu{}_\mu \propto \mu^2\tilde{\Phi}^2$ on shell, the coefficient depending only on the normalisation of the kinetic term — so the trace is the local measure of the conformal symmetry's failure, and it vanishes exactly in the massless limit.

The scope is relativistic and quantum, and one boundary matters. The biquaternion algebra carries the conformal structure — the biquaternion norm's zero set is the null cone, the algebra inverse composed with parity is the conformal inversion, and the dilations and special conformal transformations are built from it by composition — but it does not carry the conformal group as a group of its own multiplications: six of the fifteen generators are algebra elements, the dilation is a non-unit element, and the translations and special conformal transformations are not elements of $\mathbb{B}$ at all. The companion The Conformal Group in Biquaternionic Form establishes this boundary in detail, and this article uses its results rather than restating them. What is established here is the invariance of the field equations and their solutions, which is a statement about spacetime symmetry and does not require the group to sit inside the algebra. Twistor theory and the twistor realisation of the conformal group belong to the companion articles on twistors.

• Companion article The Conformal Group in Biquaternionic Form, for the conformal algebra, the inversion, the special conformal transformations, and the boundary between the algebra and the group.
• Companion article Conventions in the Biquaternion Universe, for the biquaternion norm, the d'Alembertian, the mass term, and the $ict$ conventions.
• Companion article The Klein–Gordon Equation in Biquaternionic Form, for the massive equation whose massless limit is taken here.
• Companion article Maxwell's Equations in the Biquaternionic Formulation, for the conformally invariant massless vector field.
• Companion article Canonical Quantization of the Biquaternion Maxwell Field, for the quantised massless field and its trace-free stress tensor.

The Conformal Group and the Massless Equation

The group and its generators

The conformal group of four-dimensional Minkowski space is the fifteen-parameter group $O(2,4)$, whose connected component $SO^+(2,4)$ is double-covered by $SU(2,2)$: $$ SO^+(2,4) \cong SU(2,2)/\{\pm I_4\} . $$ It is generated by the six Lorentz generators $M_{\mu\nu}$, the four translations $P_\mu$, the single dilation $D$, and the four special conformal generators $\mathcal{K}_\mu$, with the algebra $$ [D,P_\mu] = -P_\mu , \qquad [D,\mathcal{K}_\mu] = \mathcal{K}_\mu , \qquad [D,M_{\mu\nu}] = 0 , \qquad [P_\mu,P_\nu] = [\mathcal{K}_\mu,\mathcal{K}_\nu] = 0 , $$ $$ [P_\mu,\mathcal{K}_\nu] = 2\bigl(\eta_{\mu\nu}D - M_{\mu\nu}\bigr) , \qquad [M_{\mu\nu},P_\rho] = \eta_{\nu\rho}P_\mu - \eta_{\mu\rho}P_\nu . $$ The symbol $\mathcal{K}_\mu$ for the special conformal generator is the companion's notation, kept here: the letter $K_\mu$ is used in the series for the boost generator $K_k = ie_k$ of the Lorentz group, which is an element of the algebra, while $\mathcal{K}_\mu$ is a differential operator and is not. The generators act on the coordinates as $$ P_\mu = \partial_\mu , \qquad D = x^\nu\partial_\nu , \qquad M_{\mu\nu} = x_\mu\partial_\nu - x_\nu\partial_\mu , \qquad \mathcal{K}_\mu = 2x_\mu x^\nu\partial_\nu - x^2\partial_\mu , $$ with $x^2 = \eta_{\rho\sigma}x^\rho x^\sigma$. The conformal group contains the Poincaré group as the subgroup generated by $M_{\mu\nu}$ and $P_\mu$; the dilation and the special conformal transformations are the five additional parameters.

What the algebra carries, and what it does not

Six of the fifteen generators — the Lorentz ones — are elements of $\mathbb{B}$: the rotations are $J_k = e_k$ and the boosts $K_k = ie_k$, and the Lorentz group acts by $\tilde{Q}\mapsto\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}$. The dilation is realised by a non-unit element $\sqrt{\lambda}\,e_0$, which scales the biquaternion norm without preserving it. The translations and the special conformal transformations are not elements of the algebra and are not linear on $\mathbb{M}_-$: the special conformal transformation is a fractional map obtained by sandwiching a translation between two inversions, and the inversion itself is the algebra inverse composed with parity, $$ I(\tilde{Q}) = \overline{\tilde{Q}}^{-1} = \frac{\tilde{Q}^{\natural}}{N(\tilde{Q})} , \qquad N\bigl(I(\tilde{Q})\bigr) = \frac{1}{N(\tilde{Q})} . $$ The companion establishes the count precisely: six generators inside, one expressible, eight outside. The invariance derived below is therefore a statement about the field equations under conformal transformations of spacetime, not a statement that the conformal group is a subgroup of the algebra's own multiplication group, and the two must not be conflated. The reason the framework is nevertheless the natural setting is that the invariant object of the conformal group is native to the algebra: it is the zero set of the biquaternion norm, and the inversion is built from the biquaternion norm by division.

Coordinates, signature, and the two norms

One convention note is needed before the weight is used, because two quadratic forms are in play. The conformal group is the conformal group of Minkowski space, and its invariant interval is $$ x^2 = \eta_{\mu\nu}x^\mu x^\nu = -c^2t^2 + \mathbf{x}^2 = N(\tilde{Q}) , $$ with $\eta = \mathrm{diag}(-1,+1,+1,+1)$ the level-two $ict$ metric of the conventions article. The operator $\Box = \partial_{ict}^2 + \Delta$, by contrast, is the Euclidean four-dimensional Laplacian in the real coordinates $(ict,\mathbf{x})$, so the differential operator's own "norm" is $\rho^2 = (ict)^2+\mathbf{x}^2$, which equals the level-one form on the $ict$ coordinates. The two agree as functions of the coordinates — both are $N(\tilde{Q})$ — but they differ in what they are: one is the Minkowski interval whose zero set is the light cone, the other is the Euclidean distance-squared of the continued coordinates. The conformal transformations preserve the former's zero set; the conformal factor $\Omega$ is defined by the scaling of the former; and the box operator's covariance identity is a statement about the latter. Keeping the two readings apart is what prevents the sign confusions the conventions article warns about.

Conformal weight

A conformal transformation is a coordinate change $\tilde{Q}\mapsto\tilde{Q}'$ with $$ d\tilde{Q}'\,\overline{d\tilde{Q}'} = \Omega(\tilde{Q})^{2}\,d\tilde{Q}\,\overline{d\tilde{Q}} , $$ so that the interval is scaled pointwise by the square of the conformal factor. The sign of the exponent is the whole of the convention and it is fixed once and for all here: $\Omega$ is the factor whose square multiplies the interval, so that the inversion $\tilde{Q}' = \tilde{Q}^{\natural}/N(\tilde{Q})$, whose interval is scaled by $1/N^2$, has $\Omega = 1/N$, and the special conformal factor below reduces to it. Everything in the rest of the article — the operator law, the involution identity $\Omega(I(\tilde{Q})) = 1/\Omega(\tilde{Q})$, and the special conformal factor — uses this one $\Omega$, and the box-operator law is verified in it. A field of conformal weight $\Delta$ transforms by the rule $$ \phi'(\tilde{Q}) = \Omega(\tilde{Q})^{\Delta}\,\phi\bigl(\tilde{Q}'\bigr) , $$ and the transformation is a symmetry of the equation $(\Box - \mu^2)\phi = 0$ if the equation is carried into itself. The weight is a property of the equation together with the dimension: in four dimensions the massless scalar has $\Delta = 1$, which is the value forced by the requirement that the box operator's transformation law be consistent. The massive equation has no consistent weight, because under a conformal transformation the kinetic term acquires the factor $\Omega^{\Delta+2}$ while the mass term acquires only $\Omega^{\Delta}$ — but the mass coefficient is a constant and does not transform, so the two terms scale differently and no choice of $\Delta$ preserves both. This is the algebraic statement of the symmetry breaking, and it is what the rest of the article quantifies.

Conformal Covariance of the Massless Equation

The dilation

The dilation is $\tilde{Q}\mapsto\tilde{Q}' = \lambda\tilde{Q}$, whose interval is scaled by $\lambda^2$ and which therefore has the constant conformal factor $\Omega = \lambda$; read in the inverse direction, $\Omega = \lambda^{-1}$. For the rescaled field $$ \phi_\lambda(\tilde{Q}) = \lambda^{-\Delta}\,\phi(\lambda^{-1}\tilde{Q}) $$ the chain rule gives $$ \Box\,\phi_\lambda = \lambda^{-(\Delta+2)}\,\bigl(\Box\phi\bigr)\bigl(\lambda^{-1}\tilde{Q}\bigr) , $$ so $\phi_\lambda$ solves $\Box\phi_\lambda = 0$ whenever $\phi$ does, for every weight $\Delta$. This is the trivial half of the conformal symmetry: the massless wave equation is homogeneous and the box operator has dimension two, so any rescaling of a solution is a solution, and only the mass term obstructs it. The identity was checked numerically by comparing the two sides at three random points, with agreement to $2\times10^{-8}$. The nontrivial half is the inversion, where the conformal factor is a function of the point and the derivative of $\Omega$ enters.

The inversion and the conformal Laplacian

For the inversion $\tilde{Q}' = I(\tilde{Q}) = \tilde{Q}^{\natural}/N(\tilde{Q})$, the conformal factor is $\Omega(\tilde{Q}) = 1/N(\tilde{Q})$, and the transformation law of the box operator is $$ \Box_{\tilde{Q}}\Bigl[\Omega(\tilde{Q})^{\Delta}\,\phi\bigl(\tilde{Q}'\bigr)\Bigr] = \Omega(\tilde{Q})^{\Delta+2}\,\bigl(\Box_{\tilde{Q}'}\phi\bigr)\bigl(\tilde{Q}'\bigr) \qquad\text{at } \Delta = 1 , $$ with a residual proportional to $(\Delta-1)$ at other weights. The identity is the sharp statement of the massless scalar's conformal invariance, and it was verified on a superposition of two Gaussians in the four $ict$ coordinates: at $\Delta = 1$ the two sides agree to a relative accuracy of $10^{-16}$ on six random points away from the origin, while at $\Delta = 1/2$ the residual is of order unity. The two ends of the identity are the two ways of reading it: as an operator statement, it says $\Box$ is carried by the inversion into $\Omega^2\Box$; as a solution statement, it says the field $\Omega^{\Delta}\phi\circ I$ solves the massless equation whenever $\phi$ does.

The elementary consequence is that the inversion image of the constant is a solution, $$ \Box\Bigl[\frac{1}{N(\tilde{Q})}\Bigr] = 0 \quad \text{off the origin}, $$ which was verified numerically at five random points with a maximum residual of $1.5\times10^{-7}$. The function $1/N(\tilde{Q}) = 1/\rho^2$ is the massless invariant Green's function of the operator article, and the computation says that it is also the inversion image of the constant field, which is the first sign that the kernel and the conformal structure are the same object.

The special conformal transformations

The special conformal transformation is a translation sandwiched between two inversions, $$ \tilde{Q} \longmapsto I\bigl(I(\tilde{Q}) + \tilde{A}^{\natural}\bigr) , \qquad \tilde{A} \in \mathbb{M}_- , $$ the reversal on the parameter compensating the parity that the framework's inversion $I(\tilde{Q}) = \tilde{Q}^{\natural}/N(\tilde{Q})$ carries. In finite form the map and its conformal factor are $$ \tilde{Q}' = \frac{\tilde{Q} + \tilde{A}\,N(\tilde{Q})}{1 + 2\,\mathrm{Sc}\bigl(\tilde{A}^{\natural}\tilde{Q}\bigr) + N(\tilde{A})N(\tilde{Q})} , \qquad \Omega(\tilde{Q}) = \frac{1}{1 + 2\,\mathrm{Sc}\bigl(\tilde{A}^{\natural}\tilde{Q}\bigr) + N(\tilde{A})N(\tilde{Q})} , $$ so that the conformal factor is the reciprocal of the finite map's denominator; this is the companion's fractional form with the standard parameter $\tilde{A}$, reproducing $x'^\mu = (x^\mu + a^\mu x^2)/(1 + 2a\cdot x + a^2x^2)$. The pair was verified by measuring $\Omega^2 = N(d\tilde{Q}')/N(d\tilde{Q})$ on finite differences in the four coordinate directions, agreement holding to the discretisation error in every direction and for the unbarred translation failing by a factor that varies from point to point. For $\tilde{A}\to0$ the map is the identity and $\Omega\to1$; for $\tilde{A}\to\infty$ it degenerates to the constant $I(\tilde{A}^{\natural}) = \tilde{A}/N(\tilde{A})$, while $\Omega$ tends to a constant multiple of the inversion's $1/N(\tilde{Q})$. Because the inversion carries the box operator with weight $\Delta=1$ and the translation is a symmetry of the massless equation, the special conformal transformation is a symmetry as well, and the four $\mathcal{K}_\mu$ are conserved. The companion derives the fractional form and the conformal factor and shows that the generator is $$ \mathcal{K}_\mu = 2x_\mu x^\nu\partial_\nu - x^2\partial_\mu , $$ the standard expression; the point for this article is that the four extra symmetries are generated by composition of the one algebraic operation the framework does carry, the inversion built from the biquaternion norm. The conformal structure is thus native to the algebra even though the group is not a subgroup of its multiplication group.

The finite transformations and the compactification

The four families can be written in finite form, and their composition is what makes them a group. The Lorentz transformations are $\tilde{Q}\mapsto\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}$; the dilation is $\tilde{Q}\mapsto\lambda\tilde{Q}$; the translation is $\tilde{Q}\mapsto\tilde{Q}+\tilde{A}$; and the inversion is $\tilde{Q}\mapsto\tilde{Q}^{\natural}/N(\tilde{Q})$. A special conformal transformation is a translation between two inversions, and the composition law of the four families closes on the fifteen parameters. Two features of the finite maps belong to this article's subject. First, the inversion is involutive, $I(I(\tilde{Q})) = \tilde{Q}$ off the null cone, and its conformal factor satisfies $\Omega(I(\tilde{Q})) = 1/\Omega(\tilde{Q}) = N(\tilde{Q})$; the weight-one transformation law is consistent with this involution, which was checked by composing the map with itself on random off-cone points. Second, the maps are only locally defined on $\mathbb{M}_-$: the inversion is singular where $N(\tilde{Q}) = 0$, and the special conformal transformation is singular where its denominator vanishes. The conformal group is globally defined on the conformal compactification, in which the light cone is adjoined as a boundary and the inversions become everywhere defined; the companion's boundary statement and this article's field-theoretic statements both live on the locally defined maps, and the compactification is developed in the twistor literature rather than here.

The Conformal Two-Point Function

The massless invariant Green's function of the operator article, $$ G_{\mathrm{inv}}(\tilde{Q}) = \frac{1}{4\pi^2\rho^2} = \frac{1}{4\pi^2 N(\tilde{Q})} , $$ is the conformal two-point function of a field of weight $\Delta = 1$. Two properties make this identification exact. First, it is the unique function of the interval with the right scaling: the conformal covariance of a two-point function requires $$ \bigl\langle\phi(\tilde{Q})\phi(\tilde{Y})\bigr\rangle = \frac{\text{const}}{N(\tilde{Q}-\tilde{Y})^{\Delta}} \qquad\text{with } \Delta = 1 , $$ and the kernel has exactly this form, since $G_{\mathrm{inv}}(\tilde{Q}-\tilde{Y}) = 1/(4\pi^2N(\tilde{Q}-\tilde{Y}))$. Second, it obeys the inversion covariance of a weight-one two-point function, $$ G\bigl(I(\tilde{Q})-I(\tilde{Y})\bigr) = N(\tilde{Q})^{\Delta}N(\tilde{Y})^{\Delta}\,G(\tilde{Q}-\tilde{Y}) \Big|_{\Delta=1} , $$ which was verified on two hundred random off-cone pairs with a maximum relative error of $5\times10^{-16}$. The consistency of the two statements is exact: $|I(\tilde{Q})-I(\tilde{Y})|^2 = N(\tilde{Q}-\tilde{Y})/(N(\tilde{Q})N(\tilde{Y}))$, so the two-point function's covariance follows from the geometry of the inversion rather than being imposed. The two-point function of the massless theory is therefore not merely a Green's function that happens to be scale invariant: it is the unique conformally covariant two-point function, and its exponent is the conformal weight of the field. This is the sharpest link between the operator article and the conformal statement: the invariant kernel is the conformal two-point function, and the massless limit of the massive kernel is the limit in which the conformal weight exists.

The Massless Limit

The mass as the unique scale

The massive operator is $\Box - \mu^2$ with $\mu = mc/\hbar$, and $\mu$ is the only dimensional constant in it. Every conformal transformation other than the translations and the Lorentz transformations changes the length scale, so the presence of $\mu$ breaks the dilation first and all five additional symmetries with it. The massless limit is therefore not a small-parameter limit in the usual sense; it is the limit in which a symmetry is restored, and the order parameter of the restoration is the mass. Two consequences that the framework makes visible are treated below: the on-shell symbol degenerates, and the trace of the stress tensor vanishes.

The effective mass under dilation

The breaking can be exhibited in one line. If $\phi$ solves the massive equation, $(\Box-\mu^2)\phi=0$, then the dilated field $\phi_\lambda(\tilde{Q}) = \lambda^{-\Delta}\phi(\lambda^{-1}\tilde{Q})$ solves the dilated massive equation, $$ \bigl(\Box - \lambda^{-2}\mu^2\bigr)\phi_\lambda = 0 , $$ because $\Box_{\tilde{Q}} = \lambda^{-2}\Box_{\lambda^{-1}\tilde{Q}}$ and the mass term is a constant. The computation was checked on a massive plane wave: with $\mu = 0.7$ and $\lambda = 2$ the rescaled field has $\Box\phi_\lambda/\phi_\lambda = 0.1225 = \mu^2/\lambda^2$ to six decimal places. The dilation therefore rescales the mass and no value of $\lambda$ returns the original equation unless $\mu = 0$; the massless equation is the unique fixed point of the dilation, and this is the precise sense in which the mass is the order parameter of the symmetry. The same computation for the other four generators shows the same thing: each carries $\mu^2$ into a point-dependent multiple of itself, and only at $\mu=0$ is the set of solutions preserved.

The symbol on the light cone and the rank drop

On shell the four-momentum is a biquaternion $\tilde{K} = i(\omega/c)e_0 + \mathbf{k}$, and the symbol of the operator is its biquaternion norm, $$ \text{massive:}\quad N(\tilde{K}) = -\mu^2 \neq 0 , \qquad\qquad \text{massless:}\quad N(\tilde{K}) = 0 . $$ In the massless case $\tilde{K}$ is a zero divisor of the algebra, and left multiplication by it is a singular linear map. The rank of left multiplication on the complex four-dimensional module was computed for two on-shell momenta: for the massive one with $N(\tilde{K}) = -0.65$ the rank is four, and for the massless one with $N(\tilde{K}) = 0$ the rank is two. The drop from four to two is the algebraic form of the massless limit: the operator's symbol acquires a kernel, the propagator acquires a pole on the light cone, and the equation acquires the gauge-like degeneracy that the conformal symmetry organises. The geometry of that cone is the geometry of the algebra's zero divisors; the point for the massless limit is that the limit is a degeneration of the symbol, not merely the removal of a constant.

The limit of the massive theory

The massless limit of the massive kernel is the invariant kernel of the operator article, $$ G^{(\mu)}_{\mathrm{inv}} = \frac{\mu}{4\pi^2\rho}K_1(\mu\rho) \ \longrightarrow\ \frac{1}{4\pi^2\rho^2} = G_{\mathrm{inv}} \qquad (\mu\to0) , $$ the modified Bessel function reducing to its small-argument form $K_1(z)\sim1/z$. The massive kernel has a length scale $\mu^{-1}$ and the massless one has none, which is the analytic counterpart of the rank drop: as $\mu\to0$ the exponential screening disappears and the kernel becomes the pure conformal two-point function. The limit is not uniform — the massive kernel is supported inside the light cone and the massless one on it — and the non-uniformity is the infrared sensitivity of the massless theory, the same feature that makes the massless propagator's long-distance behaviour the subject of its own article.

The Dilation Ward Identity and the Improved Trace

The local statement of the dilation's breaking is the trace of the stress–energy tensor, and the bridge between the two is the dilation current $$ D^\mu = \Theta^{\mu\nu}\,x_\nu , \qquad \partial_\mu D^\mu = \Theta^\mu{}_\mu + x_\nu\,\partial_\mu\Theta^{\mu\nu} = \Theta^\mu{}_\mu \ \text{ on shell} , $$ where the second term vanishes by conservation of the improved tensor. The divergence of the dilation current is therefore exactly the trace, and the conservation of the current is exactly the vanishing of the trace: conformal invariance in the field theory and tracelessness of the stress tensor are two statements of one fact. The Noether article's machinery supplies the current once the symmetry is known; what this article adds is the identification of its divergence with the trace and the computation of the trace's value. In the massless case the current is conserved and the dilatation charge $$ Q_D = \int D^0\,d^3x $$ is a constant of the motion; in the massive case the charge is not conserved and its rate of change is $\int\Theta^\mu{}_\mu\,d^3x$, proportional to the total mass-weighted field squared. This is what makes conformal invariance a statement about a conserved charge rather than about a solution set alone.

The Trace of the Stress Tensor and the Improvement

The trace is computed in the article's own $ict$ convention throughout: the operator is $\Box = \partial_{ict}^2 + \Delta$, the Laplacian of the level-one form in the four coordinates $(ict,\mathbf{x})$, so that no index is raised or lowered, and the Lagrangian is the action article's, $\mathcal{L} = -\mathrm{Sc}[(\tilde{\nabla}^{\natural}\tilde{\Phi}^{*})(\tilde{\nabla}\tilde{\Phi})] - \mu^2\mathrm{Sc}[\tilde{\Phi}^{*}\tilde{\Phi}]$. The canonical tensor of that article, $$ T^\mu{}_\nu = -\bigl(\partial^\mu\phi^*\bigr)\partial_\nu\phi - \bigl(\partial^\mu\phi\bigr)\partial_\nu\phi^* - \delta^\mu{}_\nu\,\mathcal{L} , $$ is not traceless for the scalar field even when it is massless, because the canonical tensor of a scalar carries a total-derivative ambiguity; on these coordinates its trace is $$ T^\mu{}_\mu = 2\bigl(\partial\phi\bigr)^2 + 4\mu^2\phi^2 . $$ The improved tensor removes the ambiguity, $$ \Theta_{\mu\nu} = T_{\mu\nu} + \xi\bigl(\delta_{\mu\nu}\Box - \partial_\mu\partial_\nu\bigr)\phi^2 , $$ with $\xi$ fixed by the requirement that the massless trace vanish and hence by the normalisation of the kinetic term. The standard literature normalisation $\mathcal{L} = \tfrac12(\partial\phi)^2 - \tfrac12\mu^2\phi^2$ in the mostly-minus metric gives the textbook $\xi = \tfrac{d-2}{4(d-1)} = \tfrac16$ at $d=4$; the framework's normalisation above, whose kinetic term carries the opposite sign and twice the magnitude, gives $\xi = -\tfrac13$. Since $\Box\phi^2 = 2\phi\Box\phi + 2(\partial\phi)^2$, the trace is $$ \Theta^\mu{}_\mu = \bigl(2+6\xi\bigr)\bigl(\partial\phi\bigr)^2 + 4\mu^2\phi^2 + 6\xi\,\phi\bigl(\Box\phi\bigr) \ \longrightarrow\ 4\mu^2\phi^2 - 2\phi\bigl(\Box\phi\bigr) \ \longrightarrow\ 2\mu^2\phi^2 \quad\text{on shell at } \xi = -\tfrac13 , $$ while massless the gradient-squared term vanishes with $2+6\xi = 0$ and the trace is zero, which is the conformal invariance. In the standard normalisation the same computation reads $\Theta^\mu{}_\mu = \phi\Box\phi + 2\mu^2\phi^2$, whose on-shell value there is $\mu^2\phi^2$; the reduction uses that normalisation's own equation of motion, $\Box\phi = -\mu^2\phi$, which carries the opposite sign to the framework's $\Box\phi = +\mu^2\phi$ because the two Lagrangians place the mass term with opposite signs relative to the kinetic term. Only the coefficient is a normalisation artefact — it tracks the normalisation of the Lagrangian, with the sign convention of $\Box$ accounted for — and the physical content is shared: the trace is proportional to the mass term, so that the massless limit is exactly the trace-free limit. The identity was verified on an explicit off-shell test field in the four $ict$ coordinates, where $\Theta^\mu{}_\mu$ matched $4\mu^2\phi^2 - 2\phi\Box\phi$ to machine precision, and on an on-shell field, where it returned $2\mu^2\phi^2$.

Two consequences of the improvement are worth separating from its value. The improvement term is a total derivative, so it does not affect the equation of motion or the total charges, and it is what makes the trace local without changing the physics. And the trace scales with the Lagrangian exactly as $T_{\mu\nu}$ does, so the coefficient $2$ above and the coefficient $1$ of the standard normalisation are the same statement written twice; what no normalisation can change is the proportionality to the mass term and its vanishing when the mass is zero. The companion Exercise: The Electromagnetic Energy–Momentum Tensor records the corresponding improvement for the vector field, and notes the same $ict$ caution the action article flags: the time index carries the metric's sign, and the trace must be read through the same convention as $\Box$.

A quantum caveat belongs here. The trace of the renormalised stress tensor of a quantised field need not vanish even in the massless limit, because the regularisation introduces a scale; this is the trace anomaly, and it is an effect of quantisation rather than of the classical action. The quantum theory's statement is therefore that the classical trace vanishes with the mass and the anomalous part is a separate, quantum, contribution. The companion articles on canonical quantisation supply the quantised tensors; nothing in this article's classical trace computation depends on the anomaly.

The Maxwell Case

The electromagnetic field is the other window on the same symmetry, and its case is cleaner. The Maxwell action of the action article, $$ S_{\mathrm{Maxwell}} = -\frac14\int F_{\mu\nu}F^{\mu\nu}\,d^4x = -\frac12\int \mathrm{Re}\,\mathrm{Sc}\bigl(N(\tilde{F})\bigr)\,d^4x , $$ contains no dimensional parameter when written in terms of the field strength alone, so it is invariant under dilations without any massless limit being taken: the photon is massless, and there is no mass to remove. The Maxwell field is conformally invariant in four dimensions, and its improved stress tensor is traceless on shell, $$ \Theta^\mu{}_\mu = 0 , $$ which is the local statement of the invariance. The Proca field, by contrast, has a mass term $-\mu^2 A_\mu A^\mu$ that breaks the symmetry exactly as the scalar mass does; the massive spin-one theory treats that breaking. The pair — Maxwell conformally invariant, Proca not — is the vector-field version of the scalar pair, and the mass term is again the whole of the difference.

A second classical route to a non-zero trace: a non-invariant arena

The argument that the Maxwell trace vanishes used not only the absence of a dimensional parameter in $S_{\mathrm{Maxwell}}$ but also the invariance of the arena. The trace is the response of the action to a local rescaling of the metric, so if the action is not a scalar under that rescaling, tracelessness does not follow however scale-free the Lagrangian looks. This gives a second route to a broken classical conformal invariance, and it is the one this article's opening claim — that the mass is the unique scale — does not cover.

Kruglov's non-commutative electrodynamics makes the route concrete. Maxwell's Theory on Non-Commutative Spaces and Quaternions keeps the Maxwell Lagrangian expressed through the field strength, with no mass and with no dimensional parameter beyond the fixed background $\theta^{\mu\nu}$, and the stress tensor still has a non-zero trace at tree level. In the source's conventions, $$ T^\mu{}_\mu = (\boldsymbol\theta\cdot\mathbf B)\left(\mathbf E^2-\mathbf B^2\right) - 2(\boldsymbol\theta\cdot\mathbf E)(\mathbf E\cdot\mathbf B), $$ because $\theta^{\mu\nu}$ is fixed and does not transform with the metric, so the action is not a scalar and the variation does not yield a conserved, traceless tensor. The source states the connection explicitly: the violation of conformal invariance "relates to the violation of the Lorentz invariance in NC space."

The two routes are distinguished by their fields, and the distinction is checkable. The mass route breaks the trace for every field configuration, and its restoration is the massless limit. The arena route is proportional to the two invariants contracted with the background, so it vanishes identically for every null field: all plane electromagnetic waves remain traceless, and only non-null configurations are anomalous, including at $\boldsymbol\theta\to0$ where the route closes. So the corpus's statement that the classical trace vanishes with the mass stands, and the addition is that mass is the unique internal scale: a scale can also be imported from the arena, and when it is, the trace reappears without any mass being present. That is the classical counterpart of the quantum anomaly below — the quantum anomaly introduces a scale by regularisation, the non-invariant arena introduces one by fiat — and it is why the corpus's conformal-invariance claims are claims about a Lorentz-invariant arena.

The Massless Dirac Field

The spin-half field completes the pattern, and its conformal weight is the other standard value. In four dimensions the massless Dirac equation is conformally invariant with weight $\Delta = 3/2$, $$ \tilde{\Psi}'(\tilde{Q}) = \Omega(\tilde{Q})^{3/2}\,\tilde{\Psi}\bigl(\tilde{Q}'\bigr) , $$ the exponent being the standard one for a spinor in four dimensions, fixed by the conformal covariance of the first-order operator: the standard identity is $$ \tilde{\nabla}^{\natural}\Bigl[\Omega^{\frac{d-1}{2}}\,\tilde{\Psi}\circ I\Bigr] = \Omega^{\frac{d+1}{2}}\,\bigl(\tilde{\nabla}^{\natural}\tilde{\Psi}\bigr)\circ I , \qquad \frac{d-1}{2} = \frac32 ,\ \ \frac{d+1}{2} = \frac52 \ \ (d=4), $$ which the references supply. The mass term of the framework — the chirality-off-diagonal coupling of the companion Conventions in the Biquaternion Universe, not the central term of the scalar — breaks the symmetry for exactly the reason the scalar mass does, and the breaking is again measured by the trace of the stress tensor. The companion The Dirac Equation in Biquaternionic Form fixes the equation and its mass term; what belongs here is the weight and the fact that, in the massless limit, the Dirac field joins the scalar and the Maxwell field in the same conformal multiplet of equations. The three weights — $1$ for the scalar, $3/2$ for the spinor, $2$ for the field strength — are the three canonical values of four-dimensional conformal field theory, and the massless limit of each theory restores its weight.

Conformal Invariance and the Massless Limit in Scattering

The symmetry statement has a worked consequence in the article that follows this one in the subcategory. The Thomson cross-section for photon–electron scattering is the low-frequency limit of the Compton cross-section, and its distinguishing feature is that it is independent of the photon frequency; the Compton cross-section is not. The reason is a conformal one: the classical Thomson limit is the limit in which the electron's mass is negligible compared with the photon energy in the electron's rest frame, so the electron behaves as a massless charge, and the theory's response is scale-free. The massless limit of the present article is thus the limit that the Thomson formula silently takes, and the scale that the Klein–Nishina formula reintroduces — the factor $\hbar\omega/mc^2$ — is the same $\mu$ that breaks the conformal weight here. The biquaternion polarisation algebra of that article is the representation theory of the massless field, and the conformal statement of this article is why the limit exists at all.

The Biquaternion Structure of the Conformal Statement

Collecting the algebraic content, the conformal statement of the framework has four parts, and each is a statement about the biquaternion norm.

The weight is the exponent of the biquaternion norm in the transformation law: a field of weight $\Delta$ transforms by $N(\tilde{Q})^{-\Delta}$ under the inversion, and the massless scalar's weight is $\Delta = 1$, forced by the operator identity verified above. The weight is not a new datum; it is read off the behaviour of $\Box$ under division.

The inversion is the biquaternion norm's division, $I(\tilde{Q}) = \tilde{Q}^{\natural}/N(\tilde{Q})$, with the parity correction that makes it the standard map. It is the one conformal operation that the algebra carries as an algebraic operation, and the dilation and the special conformal transformations are built from it by composition with the translation and the scaling.

The null cone is the biquaternion norm's zero set. All conformal transformations preserve it and only it; it is the single invariant object of the group that is native to the algebra, and the massless limit is the limit in which the on-shell momentum lies on it and becomes a zero divisor.

The breaking is the mass term, and it is measured by the trace, $\Theta^\mu{}_\mu = 2\mu^2\phi^2$ on shell in the framework's normalisation of the kinetic term. The mass is the biquaternion norm's value on the four-momentum, $N(\tilde{P}) = -m^2c^2$, and the same norm that the particle action normalises to $-c^2$ is the quantity whose nonvanishing breaks the symmetry. Conformal invariance and the mass shell are therefore two readings of one biquaternion norm: on the cone it is zero and the symmetry holds, off the cone it is the mass and the symmetry fails.

Summary

The massless relativistic field of the framework has the conformal symmetry: the fifteen-parameter group $SO(2,4)$, double-covered by $SU(2,2)$, generated by the Lorentz generators, the translations, the dilation, and the special conformal generators $\mathcal{K}_\mu$, is a symmetry of the massless wave equation and of the massless Maxwell field, and is broken by the mass term alone. The algebra carries the conformal structure — the null cone as the biquaternion norm's zero set, the inversion as $I(\tilde{Q}) = \tilde{Q}^{\natural}/N(\tilde{Q})$ — but not the group; six of the fifteen generators are algebra elements, and the companion article on the conformal group establishes the count.

The massless scalar field transforms with weight $\Delta = 1$, $$ \phi'(\tilde{Q}) = \Omega(\tilde{Q})^{\Delta}\phi\bigl(\tilde{Q}'\bigr) , \qquad \Box_{\tilde{Q}}\Bigl[\Omega^{\Delta}\phi\circ I\Bigr] = \Omega^{\Delta+2}\bigl(\Box\phi\bigr)\circ I \ \text{ at } \Delta = 1 , $$ the identity verified on a superposition of two Gaussians to a relative accuracy of $10^{-16}$ at the correct weight and failing at the wrong one; the dilation covariance was verified to $2\times10^{-8}$ and the harmonicity of $1/N(\tilde{Q})$ to $1.5\times10^{-7}$. The massless invariant kernel is the conformal two-point function of weight one, $G_{\mathrm{inv}} = 1/(4\pi^2N(\tilde{Q}))$, and it is the inversion image of the constant.

The massless limit is the limit in which the on-shell symbol becomes a zero divisor: left multiplication by the on-shell momentum has complex rank four for $N(\tilde{K}) = -\mu^2\neq0$ and rank two for $N(\tilde{K}) = 0$, as computed on two on-shell momenta. The limit of the massive invariant kernel is the conformal two-point function, $\frac{\mu}{4\pi^2\rho}K_1(\mu\rho)\to1/(4\pi^2\rho^2)$, non-uniformly because the massive support is inside the cone and the massless support on it.

The improved stress tensor has trace $\Theta^\mu{}_\mu = 2\mu^2\phi^2$ on shell in the framework's normalisation, verified on an off-shell test field and on an on-shell one in the $ict$ coordinates; the trace is proportional to the mass squared and vanishes in the massless limit, which is the local statement of the symmetry. The Maxwell field is conformally invariant with no limit to take and has $\Theta^\mu{}_\mu = 0$; the Proca field is broken by its mass as the scalar is. The Thomson limit of photon–electron scattering is the physical instance of the massless limit treated here, and the polarisation algebra of that limit is the subject of the article that follows.

Summary of Notation

Further Reading

• Philippe Di Francesco, Pierre Mathieu, and David Sénéchal, Conformal Field Theory (Springer, 1997), for the conformal group, the conformal weight, the two-point function, and the improvement of the stress tensor.
• Steven Weinberg, The Quantum Theory of Fields, Vol. 1 (Cambridge, 1995), for the conformal algebra, the transformation law of the scalar field, and the trace anomaly.
• James D. Bjorken and Sidney D. Drell, Relativistic Quantum Fields (McGraw–Hill, 1965), for the massless scalar and Maxwell field propagators and their scale invariance.
• Curtis G. Callan, Sidney Coleman, and Roman Jackiw, "A New Improved Energy–Momentum Tensor", Annals of Physics 59 (1970), for the improvement term and the trace of the scalar stress tensor.
• M. J. Duff, "Twenty Years of the Weyl Anomaly", Classical and Quantum Gravity 11 (1994), for the trace anomaly and the renormalisation-induced breaking of conformal invariance.
• Hugh Osborn, "Implications of Conformal Invariance in Field Theories for General Dimensions", Annals of Physics 272 (1999), for the conformal weights of fields and the trace of the stress tensor.
• Roger Penrose and Wolfgang Rindler, Spinors and Space-Time, Vol. 2 (Cambridge, 1986), for the conformal compactification, the inversion, and the conformal structure of the light cone.