Relativistic Doppler and the Photon Four-Momentum

A light wave sweeps past you and you count crests. Two hundred and seventeen go by while your watch runs from one mark to the next. Now let someone in a rocket, tearing through at $0.9c$ , count the crests that sweep past her between the same two events — the same flashes, on the same watch. She will disagree about the frequency, the wavelength, the direction the wave came from, the energy each photon carries, the interval between your watch's ticks. She will disagree about very nearly everything. She will not disagree about the number of crests. Crests are things. Counting them is not a matter of opinion. That one stubborn fact — the phase of a wave is a Lorentz scalar — forces $(\omega/c,\ \mathbf{k})$ to be a four-vector. And once it is, the entire optics of moving bodies collapses into a single Lorentz transformation. Doppler shift is its time component. Aberration is its space component. Beaming is its solid angle. The photon's masslessness is its norm. There is one object in this article, and every named effect in the chapter is a way of looking at it.

Key takeaways

Signature $(+,-,-,-)$ ; $ds^2 = c^2dt^2 - d\mathbf{x}^2$ . Angle convention (band-locked): $\theta$ is the lab-frame angle between the source's velocity and the photon's propagation direction.
Phase invariance forces the four-vector. $\phi = \omega t - \mathbf{k}\cdot\mathbf{x}$ is a scalar and linear in $x^\mu$ , so it must be a contraction: $\boxed{\phi = +k_\mu x^\mu}$ with $k^\mu = (\omega/c,\ \mathbf{k})$ . The plus sign is the signature. (The base prints a minus — a $(-,+,+,+)$ artefact that contradicts its own next equation. Defect D1.)
Nullity: $k_\mu k^\mu = \omega^2/c^2 - |\mathbf{k}|^2 = 0$ is a Lorentz scalar, so light moves at $c$ in every frame. With $p^\mu = \hbar k^\mu$ this becomes $p_\mu p^\mu = 0$ : $E = |\mathbf{p}|c$ , $m = 0$ .
The master formula, from the invariant $k_\mu u^\mu$ : $\boxed{\omega_{\text{src}} = \gamma\,\omega_{\text{obs}}(1 - \beta\cos\theta)}$ , so $\mathcal{D} = 1/[\gamma(1-\beta\cos\theta)]$ . Every named effect is one value of $\theta$ — longitudinal ( $0, \pi$ ), transverse ( $\pi/2$ ), the beaming cone ( $\cos\theta = \beta$ ) — and aberration is the same boost read in its spatial part.
Owned result: $\mathcal{D}$ is a boost EIGENVALUE. The null directions are the eigenvectors of a boost, with eigenvalues $e^{\pm\varphi}$ — so the longitudinal Doppler factor is $e^{\pm\varphi}$ , and aberration is the same eigenvalue acting on angles: $\tan(\theta'/2) = e^{\varphi}\tan(\theta/2)$ .
Key idea: stop learning Doppler formulas. Learn one null four-vector and boost it. The frequency shift is the time component, the aberration is the space component, and they were never two phenomena.

The core insight

The optics of moving bodies has so many named effects because it was discovered piecemeal, one component at a time. Package frequency and wave vector into $k^\mu = (\omega/c, \mathbf{k})$ and every one becomes a projection of a single Lorentz transformation. Boost $k^\mu$ : the time slot is the Doppler shift; the direction of the space slot is aberration; the solid angle it sweeps is beaming; its norm is the invariance of $c$ ; multiply by $\hbar$ and you have the photon.

Where this lives today

Particle physics simply is this article. Every collider event is bookkeeping on null and timelike four-vectors, and $p_\mu p^\mu = m^2c^2$ turns a spray of detector hits into a particle: the mass of a decayed parent is $M^2c^4 = (\Sigma E)^2 - (\Sigma\mathbf{p}c)^2$ , an invariant contraction computed in the lab and meaningful in every frame. That it works for photons ( $m = 0$ , $E = pc$ ) and massive daughters with no change of formula is the point developed here — they are all four-vectors, and the norm does not care what it is the norm of.

The CMB dipole is the Doppler formula at five digits. The $3.362\ \mathrm{mK}$ dipole is $T_0\beta$ , and its agreement with Planck's $3.36208 \pm 0.00099\ \mathrm{mK}$ measures our own velocity through the universe to a part in a few thousand, using nothing but equation (4).

Laser cooling runs on the master formula. An atom moving towards a red-detuned laser sees the light blueshifted into resonance and absorbs; one moving away sees it pushed further out of resonance and does not. That velocity-selectivity is the $\theta = 0$ versus $\theta = \pi$ asymmetry of $\mathcal{D}$ , and the momentum kick is $\hbar\mathbf{k}$ : both halves of $p^\mu = \hbar k^\mu$ at work.

And the celestial sphere. The half-angle map says a boost acts on the sky conformally: angles between nearby stars are preserved while the sky is dilated towards the direction of motion. Accelerate a camera to $\beta \to 1$ and the stars do not smear — they crowd forward, locally undistorted.

The physical picture

That $k^\mu = (\omega/c, \mathbf{k})$ is a four-vector — that it boosts exactly as $x^\mu = (ct, \mathbf{x})$ does — is not a definition and not an analogy. It is a theorem, and it rests on one physical observation: the phase of a plane wave labels crests, and "how many crests have passed this point in spacetime" is a question about coincidences. If a crest arrives at the event where the rocket's nose passes your fingertip, it arrives there for both of you. The phase is a Lorentz scalar. But $\phi = \omega t - \mathbf{k}\cdot\mathbf{x}$ is also linear in the coordinates, and there is only one way to build a Lorentz scalar linearly out of a four-vector: contract it with another four-vector. The coefficients of $x^\mu$ therefore have to assemble into a four-vector — there is nowhere else for them to live. Those coefficients are $\omega/c$ and $\mathbf{k}$ . Pause on how cheap that was: no Maxwell equations, no field transformation. Crest-counting is frame-independent, and the four-vector fell out.

The core equations

k^\mu = \left(\frac{\omega}{c},\ \mathbf{k}\right), \qquad\qquad \boxed{\;\phi = k_\mu x^\mu = \omega t - \mathbf{k}\cdot\mathbf{x}\;} \qquad (1)

where: $x^\mu = (ct, \mathbf{x})$ , $\eta_{\mu\nu} = \operatorname{diag}(+,-,-,-)$ . Physical reading: the sign is not a matter of taste. Here $k_\mu x^\mu = (\omega/c)(ct) - \mathbf{k}\cdot\mathbf{x} = \omega t - \mathbf{k}\cdot\mathbf{x} = +\phi$ . The AllQFT base prints $\phi = -k_\mu x^\mu$ (defect D1) — correct only in $(-,+,+,+)$ — while its next equation, $k_\mu k^\mu = \omega^2/c^2 - |\mathbf{k}|^2$ , is unmistakably $(+,-,-,-)$ . The base contradicts itself in consecutive lines. We ship the plus.

k_\mu k^\mu = \frac{\omega^2}{c^2} - |\mathbf{k}|^2 = 0 \qquad\Longleftrightarrow\qquad \frac{\omega}{|\mathbf{k}|} = c \ \ \text{in every frame} \qquad (2)

where: the contraction is a Lorentz scalar. Physical reading: this is the second postulate, restated as geometry. A scalar that vanishes in one frame vanishes in all, so "light moves at $c$ " is not an extra assumption bolted on — it says $k^\mu$ lies on the light cone, and boosts preserve the light cone.

p^\mu = \hbar k^\mu = \left(\frac{E}{c},\ \mathbf{p}\right) \qquad\Longrightarrow\qquad p_\mu p^\mu = \frac{E^2}{c^2} - |\mathbf{p}|^2 = 0 \qquad\Longrightarrow\qquad \boxed{E = |\mathbf{p}|c} \qquad (3)

where: $E = \hbar\omega$ and $\mathbf{p} = \hbar\mathbf{k}$ . Physical reading: the photon's masslessness is not an independent fact about particles — it is the nullity of the wave four-vector, multiplied by $\hbar$ . Wave optics and photon kinematics are the same subject in two costumes, and the constant that changes costume is Planck's.

\boxed{\;\omega_{\text{src}} = k_\mu u^\mu = \gamma\,\omega_{\text{obs}}\,(1 - \beta\cos\theta)\;} \qquad\Longrightarrow\qquad \mathcal{D} \equiv \frac{\omega_{\text{obs}}}{\omega_{\text{src}}} = \frac{1}{\gamma(1-\beta\cos\theta)} \qquad (4)

where: $u^\mu = \gamma(c, \mathbf{v})$ is the source's four-velocity and $\theta$ is the lab-frame angle between $\mathbf{v}$ and the photon's propagation direction. Physical reading: this is the master formula, and it is a contraction of two four-vectors — therefore an invariant. The source's own emitted frequency is a scalar you can compute in any frame, from lab data. Every result in the chapter is one value of $\theta$ substituted into it.

Where it all comes from

Derivation AThe wave four-vector, from phase invariance alone

This is the article's reason to exist. Most treatments assert that $(\omega/c, \mathbf{k})$ transforms as a four-vector, by pointing at a plane-wave solution of Maxwell's equations and noticing that the exponent must be invariant. That makes the four-vector look like a fact about electromagnetism. It is not. It is a fact about waves, and the only physical input is that counting crests is not a matter of opinion.

Step 1. The phase is a scalar. Whether an event lies on the $n$ -th crest is a question about coincidence — did the crest arrive here? — and coincidences are absolute. So $\phi(\mathcal{E})$ is the same number for everyone. Why: this is the whole physical content of the derivation, and it is almost embarrassingly weak. Everything below is linear algebra.

Step 2. $\phi$ is linear in the coordinates: $\phi = \omega t - \mathbf{k}\cdot\mathbf{x} = a_\mu x^\mu$ for some four coefficients $a_\mu$ . Why: linearity is what "plane wave" means — surfaces of constant phase are parallel hyperplanes.

Step 3. Invoke the quotient rule. If $a_\mu x^\mu$ is a scalar for every four-vector $x^\mu$ , then $a_\mu$ must itself be a four-vector. Why: there is nowhere else for the coefficients to live. Given that $x^\mu$ is a four-vector and the output is a scalar, the input is forced.

Step 4. Match components, in the correct signature. With $\eta = \operatorname{diag}(+,-,-,-)$ , a_\mu x^\mu = a^0(ct) - \mathbf{a}\cdot\mathbf{x} \;\stackrel{!}{=}\; \omega t - \mathbf{k}\cdot\mathbf{x}, so $a^0 = \omega/c$ and $\mathbf{a} = \mathbf{k}$ . Hence \boxed{k^\mu = \left(\frac{\omega}{c},\ \mathbf{k}\right) \quad\text{and}\quad \phi = +\,k_\mu x^\mu.} Why: the sign is the signature, and the base gets it wrong (defect D1). The base prints $\phi = -k_\mu x^\mu$ — right in $(-,+,+,+)$ , wrong here — while its own next equation, $k_\mu k^\mu = \omega^2/c^2 - |\mathbf{k}|^2$ , is $(+,-,-,-)$ . A base that contradicts itself on consecutive lines is why we never inherit a signature.

Step 5. Collect the payoff. Because $k^\mu$ is a four-vector, its transformation law is fixed: the Lorentz boost, the same matrix that acts on $(ct, \mathbf{x})$ . Nothing further need be assumed about light, waves, or fields, and every Doppler and aberration result below is a component of $\Lambda^\mu{}_\nu k^\nu$ . Why: we bought the entire optics of moving bodies with one observation about crests. That is what a good four-vector does. *Verified in verify_aq-sr-doppler-four-momentum.py: in $\eta = \operatorname{diag}(1,-1,-1,-1)$ , $k_\mu x^\mu$ evaluates symbolically to $\omega t - \mathbf{k}\cdot\mathbf{x}$ (the D1 correction, $\phi = +k_\mu x^\mu$ ); and the phase is invariant under $\Lambda(\varphi)$ acting on both $k^\mu$ and $x^\mu$ .*

Go deeper · Derivation B · 2 minNullity, the invariance of c, and the massless photon

Derivation BNullity, the invariance of c, and the massless photon

Three facts usually taught as three separate lessons are one line of algebra apart: the dispersion relation of light, the second postulate, and the masslessness of the photon are the same statement about the same object — in wave language, geometric language, and particle language. This card says it once, then translates.

Step 1. Compute the norm. For light in vacuum, $\omega = c|\mathbf{k}|$ , so k_\mu k^\mu = \frac{\omega^2}{c^2} - |\mathbf{k}|^2 = \boxed{0}. The wave four-vector is null: it lies on the light cone, not inside it. Why: a timelike four-vector has a rest frame; a null one does not. This is already the statement that light has no rest frame, before any boost is performed.

Step 2. Use the fact that it is a scalar. $k_\mu k^\mu$ is invariant, so if it vanishes in one frame it vanishes in every frame — and $k'_\mu k'^\mu = 0$ says $\omega' = c|\mathbf{k}'|$ , the same dispersion relation, the same phase speed. \text{light moves at } c \text{ in every frame} \iff k_\mu k^\mu = 0 \text{ is invariant.} Why: this is the second postulate, derived rather than assumed, given that boosts are the isometries of $\eta$ . The invariance of $c$ says Lorentz transformations preserve the light cone — which is what they were built to do.

Step 3. Multiply by $\hbar$ . Einstein's $E = \hbar\omega$ and de Broglie's $\mathbf{p} = \hbar\mathbf{k}$ package into $p^\mu = \hbar k^\mu = (E/c, \mathbf{p})$ . Since $\hbar$ is a scalar, $p^\mu$ is a four-vector and p_\mu p^\mu = \hbar^2 k_\mu k^\mu = 0. Why: a four-vector times an invariant scalar is a four-vector. Nothing else is needed to promote the wave to a particle.

Step 4. Read off the photon. The general invariant is $p_\mu p^\mu = m^2c^2$ , so m^2c^2 = \frac{E^2}{c^2} - |\mathbf{p}|^2 = 0 \qquad\Longrightarrow\qquad \boxed{m = 0, \qquad E = |\mathbf{p}|c.} Why: the photon is massless because the wave four-vector is null, which is because light travels at $c$ . The chain runs in that direction; calling $m = 0$ an independent property of photons gets the logic backwards.

Step 5. Check the reduction. $E^2 = (pc)^2 + (mc^2)^2$ collapses to $E = pc$ at $m = 0$ , and to $E \approx mc^2 + p^2/2m$ at $p \ll mc$ . Why: the null case is not a degenerate exception — it is the $m \to 0$ limit of the same hyperbola, flattening onto its asymptote, which is the light cone. *Verified in verify_aq-sr-doppler-four-momentum.py: $k_\mu k^\mu = 0$ under $\omega = c|\mathbf{k}|$ ; the norm is boost-invariant on a grid of $\varphi$ and propagation angles; $p_\mu p^\mu = 0 \Rightarrow E = |\mathbf{p}|c$ ; and $E^2 = (pc)^2 + (mc^2)^2$ reduces correctly at $m = 0$ .*

Go deeper · Derivation C · 2 minOne invariant, every Doppler effect

Derivation COne invariant, every Doppler effect

The master formula is not obtained by boosting components and grinding through algebra. It is obtained by writing down a scalar — the contraction of the photon's wave four-vector with the source's four-velocity — and evaluating it in two frames. The two evaluations must agree, and that agreement is the Doppler formula. Every named effect in the chapter then falls out by choosing an angle.

Step 1. Build the invariant. Take $k^\mu = (\omega_{\text{obs}}/c)(1, \hat{\mathbf{n}})$ — the photon in the lab, propagating along $\hat{\mathbf{n}}$ — and $u^\mu = \gamma(c, \mathbf{v})$ , the source's four-velocity. Their contraction is a Lorentz scalar. Why: this is the whole trick. A scalar can be computed in whichever frame is easiest and quoted in any other.

Step 2. Evaluate it in the source's rest frame, where it is trivial: $u'^\mu = (c, \mathbf{0})$ , so k_\mu u^\mu = \frac{\omega_{\text{src}}}{c}\cdot c - 0 = \omega_{\text{src}}. Why: the invariant IS the source's proper frequency — the frequency the emitter thinks it is emitting, as a frame-independent object. That is the content of the whole construction.

Step 3. Evaluate the same scalar in the lab. k_\mu u^\mu = \frac{\omega_{\text{obs}}}{c}(\gamma c) - \left(\frac{\omega_{\text{obs}}}{c}\hat{\mathbf{n}}\right)\cdot(\gamma\mathbf{v}) = \gamma\,\omega_{\text{obs}}(1 - \beta\cos\theta), using $|\mathbf{k}| = \omega/c$ and $\hat{\mathbf{n}}\cdot\mathbf{v} = \beta c\cos\theta$ . Why: $\theta$ enters exactly here, and this line is what pins the convention. It is the lab-frame angle between the source's velocity and the photon's direction of flight — no other reading of $\theta$ makes the line true.

Step 4. Equate. \boxed{\;\omega_{\text{src}} = \gamma\,\omega_{\text{obs}}(1 - \beta\cos\theta) \qquad\Longleftrightarrow\qquad \mathcal{D} = \frac{\omega_{\text{obs}}}{\omega_{\text{src}}} = \frac{1}{\gamma(1-\beta\cos\theta)}\;} Why: one line, no matrix, no components — and manifestly a statement about two four-vectors, which is why it is convention-proof once $\theta$ is tied to the ray.

Step 5. Harvest the corollaries — by recall, not re-derivation. Each is owned elsewhere and quoted as a one-line special case. $\theta = 0 \Rightarrow \mathcal{D} = \gamma(1+\beta)$ and $\theta = \pi \Rightarrow \mathcal{D} = \gamma(1-\beta)$ : the longitudinal shifts (aq-sr-longitudinal-doppler-effect). $\theta = \pi/2 \Rightarrow \mathcal{D} = 1/\gamma$ : the transverse shift (aq-sr-transverse-doppler-effect). $\cos\theta = \beta \Rightarrow \mathcal{D} = \gamma$ : the ray emitted sideways in the source frame, and also the beaming half-angle (aq-sr-relativistic-beaming). Why: four articles, one substitution each. That is what "unification" means operationally.

Step 6. Get aberration from the space part, free. The same boost gives $\cos\theta' = (\cos\theta - \beta)/(1 - \beta\cos\theta)$ and $\sin\theta' = \sin\theta/[\gamma(1-\beta\cos\theta)]$ — the aberration map, owned by aq-sr-relativistic-aberration. Note the denominator: the same $\gamma(1-\beta\cos\theta) = 1/\mathcal{D}$ . Why: Doppler and aberration share a denominator because they are the time and space components of one boosted vector. They were never two effects. *Verified in verify_aq-sr-doppler-four-momentum.py: $k_\mu u^\mu$ evaluates symbolically to $\omega_{\text{src}}$ in the rest frame and $\gamma\omega_{\text{obs}}(1-\beta\cos\theta)$ in the lab; $\theta = 0, \pi/2, \pi$ and $\cos\theta = \beta$ give $\gamma(1+\beta)$ , $1/\gamma$ , $\gamma(1-\beta)$ , $\gamma$ ; and the aberration denominator equals $1/\mathcal{D}$ identically.*

Go deeper · Derivation D · 2 minThe Doppler factor is a boost eigenvalue

Derivation DThe Doppler factor is a boost eigenvalue

This card replaces a defective one: the base's "inverse map" $\omega' = \mathcal{D}^{-1}\gamma^2(1-\beta^2)\,\omega$ is vacuous, since $\gamma^2(1-\beta^2) \equiv 1$ (defect D7). The structural fact that deserves the slot: $\mathcal{D}$ is not merely a number appearing in Doppler formulas. It is an eigenvalue of the Lorentz boost.

Step 1. Write the boost as a hyperbolic rotation. In the $(ct, x)$ plane, with $\varphi = \operatorname{artanh}\beta$ , \Lambda(\varphi) = \begin{pmatrix}\cosh\varphi & -\sinh\varphi\\ -\sinh\varphi & \cosh\varphi\end{pmatrix}. (Owned by aq-sr-lorentz-boosts-hyperbolic-rotations; recalled.) Why: in this form the eigenvalue problem is two lines; in $\gamma$ and $\beta$ it is a mess.

Step 2. Find its eigenvectors. Try the null directions $n_\pm = (1, \pm1)$ : \Lambda(\varphi)\,n_+ = (\cosh\varphi - \sinh\varphi)\,n_+ = e^{-\varphi}\,n_+, \qquad \Lambda(\varphi)\,n_- = (\cosh\varphi + \sinh\varphi)\,n_- = e^{+\varphi}\,n_-. The light cone's two generators are the eigenvectors of every boost along that axis, with eigenvalues $e^{\mp\varphi}$ . (Recalled from aq-sr-the-light-cone.) Why: a boost cannot tilt a null ray off the cone — it can only slide it along itself, rescaling it. That rescaling is the eigenvalue, and it is the only thing a boost can do to a light ray.

Step 3. Recognise the Doppler factor. A photon along $+\hat{x}$ has $k^\mu = (\omega_{\text{obs}}/c)\,n_+$ . Boosting to the source's frame multiplies it by $e^{-\varphi}$ , so $\omega_{\text{src}} = e^{-\varphi}\omega_{\text{obs}}$ : \boxed{\;\mathcal{D} = \frac{\omega_{\text{obs}}}{\omega_{\text{src}}} = e^{+\varphi} \quad (\theta = 0), \qquad \mathcal{D} = e^{-\varphi} \quad (\theta = \pi).\;} And indeed $e^{\pm\varphi} = \cosh\varphi \pm \sinh\varphi = \gamma(1 \pm \beta)$ — exactly the longitudinal Doppler factors. Why: this is why Doppler shifts multiply while rapidities add. Eigenvalues of composed maps multiply; the logarithm of an eigenvalue is a rapidity. The multiplicativity of redshifts, $(1+z_{\text{tot}}) = (1+z_1)(1+z_2)$ , becomes a triviality about eigenvalues rather than an algebraic accident.

Step 4. Extend the eigenvalue to angles. Aberration, in the half-angle variable, is \boxed{\;\tan\frac{\theta'}{2} = \sqrt{\frac{1+\beta}{1-\beta}}\;\tan\frac{\theta}{2} = e^{\varphi}\,\tan\frac{\theta}{2}\;} — the same eigenvalue, now multiplying $\tan(\theta/2)$ . (Half-angle aberration is owned by `aq-sr-relativistic-aberration` — recalled, not re-derived; owned here is that the eigenvalue is the same.) Why: $\tan(\theta/2)$ is the stereographic coordinate on the sphere of directions, so a boost acts on the celestial sphere as a pure dilation by $e^{\varphi}$ . This is the angular twin of the frequency statement: one eigenvalue, two ways of being multiplied by it. It is also the cleanest form of the aberration map — monotone, singularity-free, and it makes the $\beta\to1$ collapse of the sky obvious.

Step 5. Check it. At $\beta = 0.6$ , $\varphi = \ln 2$ , so $e^{\varphi} = 2$ exactly: $\mathcal{D}(0) = 2$ and $\mathcal{D}(\pi) = 1/2$ ✓. A ray at $\theta = 90^\circ$ in the lab has $\tan(\theta'/2) = 2\tan(45^\circ) = 2$ , so $\theta' = 2\arctan 2 = 126.87^\circ$ — precisely what $\cos\theta' = (0 - 0.6)/(1 - 0) = -0.6$ gives ✓. Why: two independent forms of aberration agreeing to eight digits is the check that the half-angle form is an identity, not a mnemonic. *Verified in verify_aq-sr-doppler-four-momentum.py: $n_\pm$ are eigenvectors of $\Lambda(\varphi)$ with eigenvalues $e^{\mp\varphi}$ ; $e^{\pm\varphi} = \gamma(1\pm\beta)$ ; $\tan(\theta'/2) = e^{\varphi}\tan(\theta/2)$ is derived from the aberration map symbolically and cross-checked numerically on a grid; and the D7 identity $\gamma^2(1-\beta^2) = 1$ is asserted.*

:::svg[PUNCHLINE — one formula, five articles. The Doppler factor 𝒟(θ) = 1/[γ(1 − β cos θ)] at β = 0.9, plotted against the LAB angle between the source's velocity and the photon's direction of flight. Every named effect in this chapter is ONE POINT on this curve: θ = 0 the longitudinal blueshift; cos θ = β the emitter's sideways ray and the beaming half-angle cone, where 𝒟 = γ; θ = 51.19° the one direction with NO shift at all; θ = 90° the transverse shift 1/γ; θ = 180° the longitudinal redshift. There are not five effects. There is one four-vector]

Figure

Why phase invariance forces the four-vector. The crests of a plane wave are HYPERPLANES in spacetime — physical loci of events. Two observers slice spacetime differently and disagree about ω, about k, about which events are simultaneous. They cannot disagree about WHICH CREST passes through a given event: that is a coincidence, and coincidences are absolute. So φ is a scalar — and a scalar LINEAR in x^μ must be a contraction with a four-vector, which forces k^μ = (ω/c, k)

Limiting cases

Case	Result	Owned by
$\theta = 0$	$\mathcal{D} = \gamma(1+\beta) = e^{\varphi}$	blueshift — aq-sr-longitudinal-doppler-effect
$\theta = \pi$	$\mathcal{D} = \gamma(1-\beta) = e^{-\varphi}$	redshift — same
$\theta = \pi/2$	$\mathcal{D} = 1/\gamma$	aq-sr-transverse-doppler-effect
$\cos\theta = \beta$	$\mathcal{D} = \gamma$	emitter's right angle; beaming cone — aq-sr-relativistic-beaming
$\cos\theta = \frac{\gamma-1}{\gamma\beta}$	$\mathcal{D} = 1$	the unshifted ray
space part of the boost	$\cos\theta' = \frac{\cos\theta-\beta}{1-\beta\cos\theta}$	aq-sr-relativistic-aberration
$\beta \to 0$	$\mathcal{D} \to 1 + \beta\cos\theta$	the classical first-order shift
$m \to 0$ in $p_\mu p^\mu = m^2c^2$	$E = pc$	the photon

Common trapNever inherit a signature — and the base contradicts itself

Base defect D1. The AllQFT export prints the phase as $\phi = -k_\mu x^\mu$ — correct in $(-,+,+,+)$ , wrong in ours, where $k_\mu x^\mu = (\omega/c)(ct) - \mathbf{k}\cdot\mathbf{x} = +\phi$ . What makes it more than a convention clash is that the very next equation in the same base, $k_\mu k^\mu = \omega^2/c^2 - |\mathbf{k}|^2$ , is $(+,-,-,-)$ on its face. The base is internally signature-split. Hence the absolute rule: never inherit a signature; re-derive it. One dropped sign here propagates into every Doppler and aberration formula downstream.

Worked examples

1 · The cosmic microwave background dipole — the Doppler formula, measured to five digits

The CMB is a blackbody at $T_0 = 2.7255\ \mathrm{K}$ in its own rest frame. The Solar System moves through it at $v = 369.82\ \mathrm{km\,s^{-1}}$ . Predict the observed temperature pattern.

Step 1 — identify. A boosted blackbody stays a blackbody, its temperature multiplied by the Doppler factor: $T(\alpha) = \mathcal{D}\,T_0$ , with $\alpha$ the angle between our velocity and our line of sight. (From the invariance of $I_\nu/\nu^3$ , recalled from aq-sr-relativistic-beaming.) So the master formula (4) is the temperature map: T(\alpha) = \frac{T_0}{\gamma(1-\beta\cos\alpha)}.

Step 2 — apply. $\beta = v/c = 369.82/299\,792.458 = 1.233587\times10^{-3}$ . To second order, T(\alpha) = T_0\left[1 + \beta\cos\alpha + \beta^2\!\left(\cos^2\alpha - \tfrac12\right) + O(\beta^3)\right].

Step 3 — the boundary number. The first-order term is a pure dipole of amplitude \Delta T = T_0\,\beta = (2.7255\ \mathrm{K})(1.233587\times10^{-3}) = 3.362\ \mathrm{mK}. Planck's measured dipole is $3.36208 \pm 0.00099\ \mathrm{mK}$ . One contraction of two four-vectors, and we agree to four significant figures.

Step 4 — the second order. With $\cos^2\alpha - \tfrac12 = \tfrac23 P_2(\cos\alpha) - \tfrac16$ , the $\beta^2$ term carries a quadrupole of coefficient \tfrac23\beta^2 T_0 = \tfrac23(1.233587\times10^{-3})^2(2.7255\ \mathrm{K}) = 2.77\ \mu\mathrm{K}, plus an unobservable monopole shift. This kinematic quadrupole contaminates the cosmological signal at the microkelvin level and must be subtracted before the primordial quadrupole can be read.

Step 5 — read the structure. The whole sky's temperature pattern, to the precision of a space mission, is $\mathcal{D}$ evaluated in every direction. The largest anisotropy in the microwave sky is not cosmology at all — it is the Doppler effect, and it is us.

2 · Every named effect, at beta = 0.6

Take a source at $\beta = 0.6$ ( $\gamma = 1.25$ , $\varphi = \ln 2$ , $e^{\varphi} = 2$ ) and recover the chapter from equation (4).

Step 1 — identify. One formula: $\mathcal{D} = 1/[\gamma(1-\beta\cos\theta)]$ . Choose $\theta$ .

Step 2 — apply, five times. $\theta = 0$ : $\mathcal{D} = 1/[1.25(0.4)] = 2 = e^{\varphi}$ . $\theta = \pi$ : $\mathcal{D} = 1/[1.25(1.6)] = 0.5 = e^{-\varphi}$ . $\theta = \pi/2$ : $\mathcal{D} = 1/1.25 = 0.8 = 1/\gamma$ . $\cos\theta = \beta = 0.6$ , i.e. $\theta = 53.13^\circ$ : $\mathcal{D} = 1/[1.25(0.64)] = 1.25 = \gamma$ . And $\mathcal{D} = 1$ at $\cos\theta = (\gamma-1)/(\gamma\beta) = 0.25/0.75 = 1/3$ , i.e. $\theta = 70.53^\circ$ .

Step 3 — cross-check with the eigenvalue. $\mathcal{D}(0)\cdot\mathcal{D}(\pi) = 2\times0.5 = 1$ exactly, because $e^{\varphi}e^{-\varphi} = 1$ . A boost stretches one null ray and shrinks the other by reciprocal factors; there is nothing else it can do to light.

Step 4 — the angles, via the half-angle map. A lab ray at $\theta = 90^\circ$ has $\tan(\theta'/2) = e^{\varphi}\tan45^\circ = 2$ , so $\theta' = 2\arctan 2 = 126.87^\circ$ — it went out backwards in the source's frame. Check: $\cos\theta' = (0-0.6)/(1-0) = -0.6$ , and $\arccos(-0.6) = 126.87^\circ$ ✓.

Step 5 — the boundary number. Five effects, five substitutions, one formula, zero new physics. Classically $\mathcal{D} = 1$ at $\theta = \pi/2$ ; relativity says $0.8$ . That $20\%$ gap, in the one direction where classical physics is certain nothing happens, is the whole of the transverse Doppler effect — and it arrived as a substitution, not a derivation.

Physics of the portrait

Origins · 2 minFour-wavevector and how this was discovered

Four-wavevector

1905–1923 · Bern · Cologne · Berlin · St Louis

The four-wavevector has no discoverer: it was assembled from parts that arrived out of order. Einstein wrote its two halves in the same year, in different papers, for different reasons. In March 1905, in Über einen die Erzeugung und Verwandlung des Lichtes betreffenden heuristischen Gesichtspunkt, he proposed that light carries energy in quanta $E = h\nu$ ; and in §7 of Zur Elektrodynamik bewegter Körper he derived the Doppler and aberration formulas together, by demanding that the phase of a plane wave be the same for all observers — the very argument used here, three years before there was a language for it. The language arrived in 1908, when Hermann Minkowski, in his Cologne lecture Raum und Zeit, announced that "space by itself, and time by itself, are doomed to fade away into mere shadows" and gave physics the four-vector; $(\omega/c, \mathbf{k})$ became an object rather than a bookkeeping accident. The last piece was momentum: in 1917, in Zur Quantentheorie der Strahlung, Einstein argued that a radiation quantum must carry directed momentum $h\nu/c$ — light does not merely deposit energy, it pushes, in a definite direction. Arthur Compton proved it in 1923 by scattering X-rays off electrons and watching the wavelength shift by exactly the amount a massless particle of momentum $h\nu/c$ would produce. Only then was the four-vector complete: $p^\mu = \hbar k^\mu$ , a wave and a particle, the same null arrow seen from two disciplines.

History

1905

— Einstein proposes light quanta of energy E equal to h nu, in Über einen die Erzeugung und Verwandlung des Lichtes betreffenden heuristischen Gesichtspunkt.

1905

— Einstein derives the Doppler and aberration formulas together, from the invariance of the phase of a plane wave, in section 7 of Zur Elektrodynamik bewegter Körper.

1908

— Minkowski, in Raum und Zeit, gives physics the four-vector, and the pair of frequency and wave vector becomes a single geometric object.

1917

— Einstein argues in Zur Quantentheorie der Strahlung that a radiation quantum carries directed momentum h nu over c.

1923

— Arthur Compton measures the momentum of the photon by scattering X-rays from electrons, confirming the null four-momentum.

Going further

Texts. Rindler, Relativity: Special, General and Cosmological, 2nd ed., §§4.3–4.5 · Jackson, Classical Electrodynamics, 3rd ed., §11.3 · Griffiths, Introduction to Electrodynamics, 4th ed., §12.2 · Penrose & Rindler, Spinors and Space-Time I, ch. 1 (the boost as a Möbius map of the celestial sphere) · Einstein, Ann. Phys. 17, 891 (1905), §7.

Provenance. Signature $(+,-,-,-)$ , re-derived, never inherited. Angle convention band-locked. All numbers recomputed in verify_aq-sr-doppler-four-momentum.py. Base defects repaired. D1: the base prints $\phi = -k_\mu x^\mu$ — a $(-,+,+,+)$ artefact — while its own next equation is $(+,-,-,-)$ ; the base is internally signature-split, and the correct phase here is $\phi = +k_\mu x^\mu$ . D7: the base's "inverse map" $\omega' = \mathcal{D}^{-1}\gamma^2(1-\beta^2)\,\omega$ is vacuous, since $\gamma^2(1-\beta^2) \equiv 1$ ; it has been dropped and replaced by the boost-eigenvalue result, this article's owned structural payload.

Scope. This article owns the four-vector, the nullity, and the master formula. It does not re-derive what falls out of it: the longitudinal shifts belong to aq-sr-longitudinal-doppler-effect, the transverse shift to aq-sr-transverse-doppler-effect, beaming to aq-sr-relativistic-beaming, the aberration map to aq-sr-relativistic-aberration. Each is recalled as a one-line corollary, by design.

Builds on. The Lorentz Transformation Equations · Four-Vectors · Velocity Composition and the Invariance of Light. Leads to. Longitudinal Relativistic Doppler Effect · Transverse Doppler Effect · Relativistic Aberration of Light · Relativistic Headlight (Beaming) Effect · Doppler-Shift Resolution of the Twin Paradox.

Why does the phase have to be invariant? Could two observers not simply disagree?

Not without disagreeing about a coincidence, which relativity does not permit. "The third crest arrived at the event where the rocket's nose passed your fingertip" is a statement about two things happening at the same event, and no coordinate transformation can undo a coincidence. Since the phase at an event is exactly the crest count there, it is the same number for everyone — and everything else in the article is bought with that one observation.

Is the plus sign in the phase really important, or is it a convention?

It is a convention of signature, and once the signature is fixed it is a fact. In $(+,-,-,-)$ , $k_\mu x^\mu = \omega t - \mathbf{k}\cdot\mathbf{x} = +\phi$ ; in $(-,+,+,+)$ the contraction flips and $\phi = -k_\mu x^\mu$ . What is not a convention is consistency: the AllQFT base prints the minus and then, in the very next equation, uses the opposite signature. That is not a choice, it is a contradiction, and it propagates a sign error into every downstream formula.

The Doppler factor is called an eigenvalue. Eigenvalue of what, exactly?

Of the Lorentz boost, acting on the two null directions. Write the boost as a hyperbolic rotation of rapidity $\varphi$ ; the vectors $(1,1)$ and $(1,-1)$ in the $(ct,x)$ plane are its eigenvectors, with eigenvalues $e^{-\varphi}$ and $e^{+\varphi}$ . A photon's four-vector is one of those null directions, so a boost can do exactly one thing to it: rescale it. That rescaling is the Doppler shift. Hence Doppler factors multiply while rapidities add, and $\mathcal{D}_{\max}\mathcal{D}_{\min} = 1$ — the eigenvalues are reciprocal because the boost has determinant one.

Do I need quantum mechanics to talk about the photon four-momentum?

Only for the value of $\hbar$ . The four-vector $k^\mu$ and its nullity are classical wave optics plus relativity — no quanta anywhere. The single quantum input is $p^\mu = \hbar k^\mu$ , and because $\hbar$ is a Lorentz scalar it cannot spoil anything's four-vector character. The photon's masslessness, its $E = pc$ , and its Doppler shift are all inherited unchanged from the classical wave. Quantum theory supplies a constant, not a new geometry.

How does one formula give five different named effects?

Because they were named before anyone knew they were the same thing. $\mathcal{D} = 1/[\gamma(1-\beta\cos\theta)]$ is a function of one angle: evaluate at $\theta = 0$ and $\pi$ for the longitudinal shifts, at $\pi/2$ for the transverse shift, at $\cos\theta = \beta$ for the emitter's right angle and the beaming half-angle. Take the space components of the same boosted four-vector and you have aberration; count the solid angle it sweeps and you have beaming. Five names, five projections, one null arrow.

What does an equation that is a tautology look like, and why does it matter?

Like defect D7: $\omega' = \mathcal{D}^{-1}\gamma^2(1-\beta^2)\,\omega$ . Dimensionally correct, algebraically true, beautifully typeset — and since $\gamma^2(1-\beta^2) = 1$ identically, it says $\omega' = \mathcal{D}^{-1}\omega$ , which was already on the page. It contains no information. This is a specific failure mode of generated physics: not a wrong statement but an empty one dressed as a result. The check is cheap — substitute the definitions and see whether anything survives.