Transverse Doppler Effect

Ask a classical physicist what happens to a siren's pitch at the exact instant it passes you — the moment it is neither approaching nor receding. He will say: nothing. Classical Doppler is radial; at closest approach the radial velocity is zero; the shift is zero. He is right about sound, and for a good reason: classical Doppler is bookkeeping about distances, and at that instant the distance is not changing. Relativity says he is wrong about light, and it says so by a factor that has nothing to do with distances at all. The source's clock is running slow. It emits fewer crests per second than it thinks it does, and it does so whether it is coming, going, or sailing past. So even in the direction where the geometry contributes exactly nothing, a redshift survives:

\nu_\perp = \frac{\nu_0}{\gamma} = \nu_0\sqrt{1-\beta^2}.

This is time dilation made visible. It has no classical analogue whatsoever — not a small correction to a classical effect, but a shift where the classical theory predicts precisely zero. And it comes with a trap that has snared textbooks for eighty years, because "transverse" is not one condition but two: a right angle in your frame gives a redshift of $1/\gamma$ ; a right angle in the source's frame gives a blueshift of $\gamma$ . Both are true. They are different questions.

Key takeaways

Signature $(+,-,-,-)$ . Angle convention (band-locked): $\theta$ is the lab-frame angle between the source's velocity and the photon's propagation direction; $\mathcal{D} = 1/[\gamma(1-\beta\cos\theta)]$ .
The lab-frame right angle ( $\theta = 90^\circ$ ) kills the classical factor exactly: $1 - \beta\cos 90^\circ = 1$ . What is left is $\boxed{\nu_\perp = \nu_0/\gamma}$ — pure time dilation, with no classical analogue.
The emitter-frame right angle ( $\theta' = 90^\circ$ ) is a different physical ray: it arrives at lab angle $\arccos\beta$ and is blueshifted by $\gamma$ . Both statements are correct.
Owned result: the two right angles are EXACT RECIPROCALS. $\mathcal{D}(\theta = 90^\circ)\cdot\mathcal{D}(\theta' = 90^\circ) = (1/\gamma)(\gamma) = 1$ , for every $\beta$ . (The AllQFT base explicitly denies this. It is wrong.)
The effect is second order: $\nu_\perp/\nu_0 - 1 = 1/\gamma - 1 = -\tfrac12\beta^2 - \tfrac18\beta^4 - \cdots$ . The $\tfrac38\beta^4$ that the base prints belongs to $\gamma - 1$ , not to $1/\gamma - 1$ .
Ives and Stilwell confirmed it in 1938 — without ever observing at $90^\circ$ . They judged the right-angle geometry hopeless and instead measured both longitudinal lines, red and blue, and found the centroid of the pair displaced: $\lambda_{\text{avg}} = \tfrac12(\lambda_\to + \lambda_\leftarrow) = \gamma\lambda_0$ , exactly.
Key idea: strip the geometry away and something is still there. The transverse shift is the moving clock, seen directly — which is why it, and not the longitudinal shift, is the experiment that tests relativity.

The core insight

Every other Doppler shift mixes two effects: the classical crest-bunching that tracks the changing distance, and the relativistic slowing of the source's clock. The transverse direction is the one place where the first vanishes identically — $(1-\beta\cos\theta)$ is exactly $1$ at $\theta = 90^\circ$ — leaving the second naked. That is the point: the transverse Doppler effect is not a Doppler shift with a relativistic correction. It is the relativistic correction, with the Doppler shift removed.

Where this lives today

The Mössbauer rotor. Kündig's 1963 experiment put a $^{57}\mathrm{Fe}$ absorber on the rim of an ultracentrifuge, $9.3\ \mathrm{cm}$ from the axis, with a $^{57}\mathrm{Co}$ source at the centre. The recoil-free gamma line is spectacularly narrow, so a fractional shift of $10^{-12}$ is detectable — and the geometry is exactly transverse by construction, since the rim's velocity is always perpendicular to the radius. Rotation remains the cheapest way to hold a right angle honestly.

Storage rings, and the modern precision. Ion rings turn the experiment inside out: rather than hunting a tiny shift on a fast-decaying beam, you hold ions for hours at $\beta$ of a few tenths and interrogate them with lasers. The saturation-spectroscopy variants compare the forward and backward transitions of the same ions, so the product of the two shifted frequencies gives $\nu_0^2$ directly — the $k\cdot(1/k) = 1$ reciprocity above, deployed as a systematics killer. Deviations from $1/\gamma$ are now bounded at parts in $10^{8}$ .

Every atomic clock that moves. The velocity term in the GPS timing budget — satellite clocks losing about $7\ \mu\mathrm{s}$ per day from orbital speed — is the transverse Doppler effect, at $\beta \approx 1.3\times10^{-5}$ and a shift of order $10^{-10}$ . Optical clocks now see the same $-\tfrac12\beta^2$ at walking pace, and "second order, therefore negligible" has stopped being an acceptable sentence.

And a warning about jets. The transverse value $1/\gamma$ is not the minimum of $\mathcal{D}$ — it is an interior point of $\gamma(1-\beta) \le \mathcal{D} \le \gamma(1+\beta)$ , and a receding source is redshifted far harder. Using $1/\gamma$ as though it bounded $\mathcal{D}$ below is a real error: see aq-sr-relativistic-beaming.

The physical picture

Start from the master formula (owned by aq-sr-doppler-four-momentum, used here):

\mathcal{D} = \frac{\nu_{\text{obs}}}{\nu_0} = \frac{1}{\gamma\,(1 - \beta\cos\theta)},

with $\theta$ the lab-frame angle between the source's velocity and the photon's direction of travel. The formula factorises into two pieces that are conceptually independent:

$(1 - \beta\cos\theta)^{-1}$ — the classical factor. It is pure geometry: how fast the source is closing on, or retreating from, the line the photon flies along. It is the only factor a nineteenth-century physicist had.
$\gamma^{-1}$ — the relativistic factor. It is the source's clock, and it is there in every direction, at every angle, always.

Now set $\theta = 90^\circ$ . Then $\cos\theta = 0$ and the classical factor is identically one — not approximately, not to leading order, but exactly. Everything geometrical has cancelled, and what remains is

\nu_\perp = \frac{\nu_0}{\gamma}.

A redshift, of purely relativistic origin, in the one direction where classical physics guarantees no shift at all.

Which photon is this? The one whose propagation direction is perpendicular to the source's track — and a straight ray is perpendicular to the track exactly when it left from the point of closest approach. So the light you catch arriving from the direction at right angles to the source's path is redshifted by $1/\gamma$ . You catch it later, when the source has already sailed past: it is no longer where its light appears to come from. That gap between the retarded and the apparent position is where most of the confusion in this subject lives.

Interactive Simulation

Relativistic Doppler vs. Emission Angle — set the source speed and emission angle and read the frequency ratio, including the purely relativistic transverse redshift at 90°.

Figure

PUNCHLINE — WHICH 90 DEGREES? Two different rays, two opposite answers, both correct. Source moves at β = 0.5 (γ = 1.1547). LEFT: the photon whose propagation direction is perpendicular to v IN THE LAB (θ = 90°). It left the point of closest approach; in the source frame it went out BACKWARD, at θ′ = arccos(−β) = 120°. Result: REDSHIFT, 𝒟 = 1/γ = 0.8660. RIGHT: the photon emitted perpendicular to v IN THE SOURCE FRAME (θ′ = 90°). Aberration tips it FORWARD to lab angle θ = arccos β = 60°. Result: BLUESHIFT, 𝒟 = γ = 1.1547. And 0.8660 × 1.1547 = 1.0000 — the two shifts are EXACT reciprocals, at every β

The core equations

\boxed{\;\nu_\perp = \frac{\nu_0}{\gamma} = \nu_0\sqrt{1-\beta^2}, \qquad\qquad \lambda_\perp = \gamma\,\lambda_0 \;} \qquad (1)

where: $\theta = 90^\circ$ in the lab. Physical reading: the classical factor $(1-\beta\cos\theta)$ is exactly $1$ here, so the entire shift is the moving clock. In wavelength it is a redshift by $\gamma$ ; in frequency, a redshift by $1/\gamma$ . Same statement.

\mathcal{D} = \frac{1}{\gamma(1-\beta\cos\theta)} \;=\; \gamma\,(1 + \beta\cos\theta') \qquad (2)

where: $\theta$ is the lab angle, $\theta'$ the emission angle in the source's rest frame, and the two are linked by aberration, $\cos\theta = (\cos\theta'+\beta)/(1+\beta\cos\theta')$ (owned by aq-sr-relativistic-aberration). Physical reading: one Doppler factor, two ways to name the ray. The right-hand form makes the emitter-frame right angle trivial: set $\cos\theta' = 0$ and read $\mathcal{D} = \gamma$ — a blueshift.

\theta = 90^\circ \iff \cos\theta' = -\beta, \qquad\qquad \theta' = 90^\circ \iff \cos\theta = +\beta \qquad (3)

where: both follow from the aberration map. Physical reading: the two right angles are different rays. A ray that is perpendicular in the lab went out backwards in the source frame ( $\theta' = \arccos(-\beta) = 120^\circ$ at $\beta = 0.5$ ) — the source was partly receding as it emitted, so it redshifts. A ray perpendicular in the source frame arrives forwards in the lab ( $\theta = \arccos\beta = 60^\circ$ ) — the source was partly approaching, so it blueshifts. And these two shifts are exact reciprocals.

\frac{\Delta\nu}{\nu_0} = \frac{1}{\gamma} - 1 = -\frac{1}{2}\beta^2 - \frac{1}{8}\beta^4 - \frac{1}{16}\beta^6 - \cdots \qquad\text{(NOT } -\tfrac38\beta^4\text{)} \qquad (4)

where: the expansion is of $\sqrt{1-\beta^2} - 1$ . Physical reading: the effect is second order in $\beta$ — which is why it hid from nineteenth-century optics, and why finding it is a real test rather than a formality. The AllQFT base prints $-\tfrac12\beta^2 - \tfrac38\beta^4$ (defect D2); $\tfrac38$ is the fourth-order coefficient of $\gamma - 1 = \tfrac12\beta^2 + \tfrac38\beta^4 + \cdots$ , an entirely different series. At $\beta = 0.5$ the base's version is $10.8\%$ off, in the wrong direction.

Where the transverse shift comes from

Derivation AThe lab right angle is pure time dilation

The result $\nu_\perp = \nu_0/\gamma$ can be got in one line from the master formula, and it is worth getting that way and getting it again from a bare clock argument, because the two routes illuminate different things. The one-liner shows that the transverse case is not special — it is one value of $\theta$ among a continuum. The clock argument shows why it is special: it is the only value of $\theta$ at which relativity is not sharing the stage with Newton.

Step 1. Take the master formula, $\mathcal{D} = 1/[\gamma(1-\beta\cos\theta)]$ , with $\theta$ the lab-frame angle between the source's velocity and the photon's direction of flight. (Derived from the invariant $k_\mu u^\mu$ in aq-sr-doppler-four-momentum; recalled, not re-derived.) Why: the whole family of Doppler results is one formula. Naming the transverse case a separate phenomenon is a convenience, not a physical fact.

Step 2. Impose the transverse condition in the lab: $\theta = \pi/2$ , so $\cos\theta = 0$ . Then \mathcal{D} = \frac{1}{\gamma(1 - 0)} = \frac{1}{\gamma} \qquad\Longrightarrow\qquad \boxed{\nu_\perp = \frac{\nu_0}{\gamma} = \nu_0\sqrt{1-\beta^2}} Why: the classical factor is exactly $1$ , not approximately. Every trace of the source's motion through space has cancelled, and only the clock survives.

Step 3. Re-derive it from the clock alone. The source ticks at proper rate $\nu_0$ ; in the lab its clock runs slow, so it emits at $\nu_0/\gamma$ . Does light-travel add anything? The received ray runs perpendicular to the track, so the component of the source's velocity along that ray is zero: successive crests have the same distance to fly, and the classical bunching factor is unity. Received rate $= \nu_0/\gamma$ . Why: the transverse direction is the one direction in which the light-travel bookkeeping contributes nothing — and so the one direction that measures the clock and nothing else.

Step 4. Confirm there is no classical analogue. Classically the shift is $f_{\text{obs}}/f_0 = 1/(1+\beta_s\cos\psi)$ , which at $\psi = 90^\circ$ is exactly $1$ . Classical physics predicts zero shift, at every order, in the transverse direction. Relativity predicts $-\tfrac12\beta^2$ . Why: not a correction to a classical result, but a signal where the classical prediction is identically null. That is what makes it a clean test — and why Einstein proposed the canal-ray experiment in 1907.

Step 5. Get the size right. $\nu_\perp/\nu_0 - 1 = \sqrt{1-\beta^2} - 1 = -\tfrac12\beta^2 - \tfrac18\beta^4 - \tfrac{1}{16}\beta^6 - \cdots$ . At $\beta = 0.5$ the exact value is $0.866025 - 1 = -0.133975$ ; the two-term series gives $-0.132813$ ( $0.87\%$ low). The base's $-\tfrac12\beta^2 - \tfrac38\beta^4$ gives $-0.148438$ — $10.8\%$ off, and the $\tfrac38$ is stolen from the expansion of $\gamma-1$ . Why: base defect D2, quantified. The two series $\gamma - 1$ and $1/\gamma - 1$ agree at second order and diverge at fourth, which is exactly the sort of thing that survives a casual proof-read. *Verified in verify_aq-sr-transverse-doppler-effect.py: $\mathcal{D}(\theta = \pi/2) = 1/\gamma$ symbolically; $\sqrt{1-\beta^2}-1$ has series coefficients $(-\tfrac12, -\tfrac18, -\tfrac1{16})$ while $\gamma-1$ has $(\tfrac12, \tfrac38, \tfrac5{16})$ — the D2 correction; the classical transverse shift is identically zero; and every number above is recomputed from the article text.*

Go deeper · Derivation B · 2 minThe two right angles are exact reciprocals

Derivation BThe two right angles are exact reciprocals

This is the article's owned structural result, and it settles a question the AllQFT base got explicitly and demonstrably wrong. The base's own derivation card ends with the assertion that "the product of the two shifts equals unity is false". It is not false. It is true, exactly, for every $\beta$ — and once you see why, the whole "which $90^\circ$ " confusion stops being a paradox and becomes a triviality.

Step 1. Compute the lab-right-angle shift. $\theta = 90^\circ \Rightarrow \mathcal{D}_A = 1/[\gamma(1-0)] = 1/\gamma$ . Redshift. Why: established above.

Step 2. Compute the emitter-right-angle shift. Use the second form of (2): $\theta' = 90^\circ \Rightarrow \mathcal{D}_B = \gamma(1 + \beta\cdot 0) = \gamma$ . Blueshift. Why: this ray was emitted sideways in the source's own frame, but aberration tips it forward in the lab — so the observer catches it while the source is still, on balance, approaching. The forward tilt over-compensates the clock, and the net effect is a blueshift.

Step 3. Multiply. \boxed{\;\mathcal{D}_A \cdot \mathcal{D}_B = \frac{1}{\gamma}\cdot\gamma = 1 \quad\text{exactly, for every }\beta\;} Why: the base asserts this is false. It is a two-line calculation and it is true. A defect in a "verified" line of an authoritative source is precisely the kind of thing that propagates.

Step 4. Understand why it is a reciprocal pair, rather than a coincidence. Locate the two rays. From aberration, $\theta = 90^\circ$ requires $\cos\theta' = -\beta$ ; and $\theta' = 90^\circ$ requires $\cos\theta = +\beta$ . The two configurations are related by $\beta \to -\beta$ with the frames exchanged — which is exactly the reciprocity of the Lorentz transformation. So $\mathcal{D}_B(\beta) = 1/\mathcal{D}_A(\beta)$ is forced. Why: the source sees the observer moving at $-\beta$ , and its "sideways" is the observer's "sideways" with the sign of the velocity flipped. Reciprocity of boosts, restated in optics.

Step 5. Put numbers on it at $\beta = 0.5$ , where $\gamma = 1.154701$ . Lab right angle: $\mathcal{D}_A = 0.866025$ , a $13.40\%$ redshift, and the ray went out at $\theta' = \arccos(-0.5) = 120.00^\circ$ — backwards in the source frame. Emitter right angle: $\mathcal{D}_B = 1.154701$ , a $15.47\%$ blueshift, arriving at $\theta = \arccos(0.5) = 60.00^\circ$ — forwards in the lab. Product: $0.866025 \times 1.154701 = 1.000000$ . Why: the asymmetry in the percentages ( $13.40$ against $15.47$ ) is not an error — a factor and its reciprocal do not produce equal-and-opposite percentage changes. That, too, is a trap.

Step 6. Find the ray that really is unshifted — neither of these is. $\mathcal{D} = 1$ requires $\cos\theta = (\gamma-1)/(\gamma\beta)$ ; at $\beta = 0.5$ that is $\theta = 74.46^\circ$ : inside the forward hemisphere, a right angle in neither frame. Why: the sharpest way to see that "transverse" and "unshifted" are different conditions. The null direction is a third ray, in between. *Verified in verify_aq-sr-transverse-doppler-effect.py: $\mathcal{D}(\theta=\pi/2)\cdot\mathcal{D}(\theta'=\pi/2) = 1$ symbolically in $\beta$ ; the aberration map sends $\theta = \pi/2 \mapsto \cos\theta' = -\beta$ and $\theta' = \pi/2 \mapsto \cos\theta = +\beta$ ; the $\beta = 0.5$ numbers ( $0.866025$ , $1.154701$ , $120.00^\circ$ , $60.00^\circ$ , $74.46^\circ$ ) are all recomputed; and the base's "product is not unity" claim is asserted FALSE.*

Go deeper · Derivation C · 2 minIves and Stilwell measured it without ever looking sideways

Derivation CIves and Stilwell measured it without ever looking sideways

The experiment that first confirmed time dilation did not observe at a right angle. Ives and Stilwell judged the transverse geometry hopeless — hold the angle to a fraction of a degree and the enormous first-order longitudinal term leaks in and swamps the second-order effect you are hunting. So they did something better: they observed the beam longitudinally, capturing both the approaching and the receding line at once (the beam directly, and its reflection in a mirror), and they looked at where the centre of gravity of that pair sat relative to the undisplaced line. That centroid displacement is the time-dilation effect, and this is the calculation.

Step 1. Write the two longitudinal lines. Let $k = \sqrt{(1+\beta)/(1-\beta)}$ be the approach factor. Forward-emitted light arrives blueshifted, $\nu_\to = k\,\nu_0$ ; backward-emitted light arrives redshifted, $\nu_\leftarrow = \nu_0/k$ . Why: the two lines straddle the rest frequency, and their first-order shifts are equal and opposite. That is exactly the property to exploit.

Step 2. Average the frequencies. \tfrac12\left(\nu_\to + \nu_\leftarrow\right) = \tfrac12\nu_0\left(k + \tfrac1k\right) = \tfrac12\nu_0\left(\sqrt{\tfrac{1+\beta}{1-\beta}} + \sqrt{\tfrac{1-\beta}{1+\beta}}\right) = \tfrac12\nu_0\cdot\frac{(1+\beta)+(1-\beta)}{\sqrt{1-\beta^2}} = \boxed{\gamma\,\nu_0} Why: $k + 1/k = 2\gamma$ — an exact identity, and the engine of the whole experiment. The first-order terms cancel identically in the sum, leaving $\gamma$ standing alone.

Step 3. Average the wavelengths — and note that you get $\gamma$ again. $\lambda_\to = \lambda_0/k$ , $\lambda_\leftarrow = k\lambda_0$ , so $\tfrac12(\lambda_\to + \lambda_\leftarrow) = \tfrac12\lambda_0(k + 1/k) = \gamma\lambda_0$ . Redward. \boxed{\;\lambda_{\text{avg}} = \tfrac12(\lambda_\to + \lambda_\leftarrow) = \gamma\,\lambda_0\;} Why: the set $\{k, 1/k\}$ is closed under inversion, so the arithmetic mean picks up the same factor $\gamma$ whether you average in $\nu$ or in $\lambda$ . The magnitude is unambiguous. (The direction is not the same word in both variables: the mean frequency is $\gamma$ times higher and the mean wavelength $\gamma$ times longer. Ives and Stilwell measured wavelengths on a photographic plate, so what they reported is a <strong>displacement of the centroid towards the red</strong>.)

Step 4. Correct the base. The AllQFT base prints $\nu_{\text{avg}} \approx \nu_0(1 - \tfrac12\beta^2)$ — that is $\nu_0/\gamma$ , the wrong sign (defect D3). The truth is $\nu_{\text{avg}} = \gamma\nu_0 \approx \nu_0(1 + \tfrac12\beta^2)$ . Why: this is not a cosmetic slip. The sign of the centroid shift <strong>is the experiment's observable</strong>; get it backwards and you have predicted the opposite of what Ives and Stilwell saw.

Step 5. Compute the real number, for the real apparatus. Ives and Stilwell accelerated hydrogen canal rays ( $\mathrm{H_2^+}$ and $\mathrm{H_3^+}$ ) through $6788$ to $18\,350$ volts and watched the $\mathrm{H}\beta$ line at $\lambda_0 = 4861.35\ \text{Å}$ . At the top voltage, an $\mathrm{H_2^+}$ ion carries $18\,350\ \mathrm{eV}$ shared by two nucleons, so each emitting hydrogen atom has $K = 9175\ \mathrm{eV}$ against a rest energy of $938.78\ \mathrm{MeV}$ : \beta = \sqrt{\frac{2K}{mc^2}} = \sqrt{\frac{2(9175)}{9.3878\times10^{8}}} = 4.4212\times10^{-3}, \qquad \gamma - 1 = 9.773\times10^{-6}. Hence \Delta\lambda = (\gamma-1)\lambda_0 = (9.773\times10^{-6})(4861.35\ \text{Å}) = 0.0475\ \text{Å}. Why: the published expected second-order displacement was about $0.05\ \text{Å}$ — roughly five thousandths of a millimetre on the plate. Our independent recomputation lands on $0.0475\ \text{Å}$ . They were measuring a shift a hundred times smaller than the first-order splitting, on a photographic plate, in 1938. *Verified in verify_aq-sr-transverse-doppler-effect.py: $k + 1/k = 2\gamma$ symbolically; $\tfrac12(\nu_\to+\nu_\leftarrow) = \gamma\nu_0$ and $\tfrac12(\lambda_\to+\lambda_\leftarrow) = \gamma\lambda_0$ symbolically (the D3 correction, with the base's sign asserted WRONG); and the Ives–Stilwell chain $18\,350\ \mathrm{V} \to \beta = 4.4212\times10^{-3} \to \Delta\lambda = 0.0475\ \text{Å}$ recomputed from first principles.*

Figure

The size of the effect, and where the base went wrong. Plotted: the EXACT fractional shift 1/γ − 1 (solid red), the correct two-term series −½β² − ⅛β⁴ (dashed red, hugging it), and the AllQFT base's −½β² − (3/8)β⁴ (dashed grey, visibly wrong). The 3/8 is not a typo of ⅛ — it is the fourth-order coefficient of γ − 1 (solid blue, mirrored above), a completely different series. The two series agree at second order and part company at fourth, which is exactly how the defect survived review. At β = 0.5 the exact answer is −0.13397; the base says −0.14844, off by 10.8%

Figure

IVES AND STILWELL, 1938 — how you measure time dilation without ever looking sideways. Hydrogen canal rays at 18 350 V give β = 4.4212 × 10⁻³. Photograph the Hβ line (4861.35 Å) BOTH ways at once: the beam directly (blueshifted, λ₀/k) and its mirror image (redshifted, kλ₀). The first-order splitting is huge — ±21.5 Å — and it cancels exactly in the mean. What does NOT cancel is the centroid: ½(λ→ + λ←) = γλ₀, displaced to the RED by (γ − 1)λ₀ = 0.0475 Å. That is a shift 450 times smaller than the splitting, on a photographic plate, and it is time dilation

Limiting cases

Case	Result	Meaning
$\beta \to 0$	$\nu_\perp/\nu_0 \to 1 - \tfrac12\beta^2$	second order — invisible to nineteenth-century optics
$\theta = 90^\circ$ (lab)	$\mathcal{D} = 1/\gamma$	redshift, pure time dilation
$\theta' = 90^\circ$ (emitter)	$\mathcal{D} = \gamma$	blueshift — a different ray, also correct
product of the two	$= 1$ exactly	the owned result
$\cos\theta = (\gamma-1)/(\gamma\beta)$	$\mathcal{D} = 1$	the actual unshifted ray: $74.46^\circ$ at $\beta = 0.5$
classical, any $\beta_s$	$\mathcal{D}_\perp = 1$ exactly	no classical transverse effect at any order
$\beta \to 1$	$\nu_\perp \to 0$	the transverse redshift becomes total
Ives–Stilwell mean	$\lambda_{\text{avg}} = \gamma\lambda_0$	the centroid moves REDWARD by $(\gamma-1)\lambda_0$

Common trapNever say the transverse shift is a redshift without saying whose right angle

This is the single most common error in the literature on this topic, and the AllQFT base itself stumbles over it. A right angle in the observer's frame gives a redshift, $\mathcal{D} = 1/\gamma$ . A right angle in the emitter's frame gives a blueshift, $\mathcal{D} = \gamma$ . These are not contradictory: they are two different rays, emitted at two different moments, arriving from two different directions. Aberration is what maps one $90^\circ$ into a non-right angle in the other frame ( $\theta = 90^\circ \Leftrightarrow \theta' = 120^\circ$ at $\beta = 0.5$ ). Their product is exactly $1$ . If a source states "the transverse Doppler effect" without naming a frame, it has not said anything.

Common trapThe beta-to-the-fourth coefficient is one eighth, not three eighths

Base defect D2. The expansion of the transverse shift is \frac{1}{\gamma} - 1 = \sqrt{1-\beta^2}-1 = -\tfrac12\beta^2 - \tfrac18\beta^4 - \tfrac1{16}\beta^6 - \cdots The base prints $-\tfrac12\beta^2 - \tfrac38\beta^4$ . The $\tfrac38$ is real, but it is the fourth-order coefficient of the other series, $\gamma - 1 = \tfrac12\beta^2 + \tfrac38\beta^4 + \tfrac5{16}\beta^6 + \cdots$ . The two agree at second order — which is why almost every use of them is unaffected — and diverge at fourth, which is why the error is invisible until precisely the moment you need it. At $\beta = 0.5$ the base's value is $10.8\%$ wrong.

Worked examples

1 · Ives and Stilwell, 1938 — recomputing the shift they hunted

Hydrogen canal rays are accelerated through $18\,350\ \mathrm{V}$ as $\mathrm{H_2^+}$ ions, which then pick up an electron and emit the $\mathrm{H}\beta$ line at $\lambda_0 = 4861.35\ \text{Å}$ . Predict the displacement of the centroid of the forward- and backward-emitted lines.

Step 1 — identify. The observable is the mean wavelength of the red/blue pair, and the theory says $\lambda_{\text{avg}} = \gamma\lambda_0$ . So we need $\gamma - 1$ , which needs $\beta$ , which needs the kinetic energy per emitting atom.

Step 2 — the kinetic energy per nucleon. An $\mathrm{H_2^+}$ ion carrying charge $e$ through $18\,350\ \mathrm{V}$ gains $18\,350\ \mathrm{eV}$ , shared between two hydrogen nuclei. Each emitting hydrogen atom therefore carries $K = 9175\ \mathrm{eV}$ , against a rest energy $mc^2 = 938.78\ \mathrm{MeV}$ .

Step 3 — apply. At these speeds the non-relativistic relation is overwhelmingly good: \beta = \sqrt{\frac{2K}{mc^2}} = \sqrt{\frac{2\times 9175}{9.3878\times 10^{8}}} = \sqrt{1.9546\times10^{-5}} = 4.4212\times10^{-3}. Then $\gamma - 1 = 9.773\times10^{-6}$ (and $\tfrac12\beta^2 = 9.773\times10^{-6}$ — the second-order formula, doing its job).

Step 4 — the boundary number. \Delta\lambda = (\gamma - 1)\lambda_0 = (9.773\times10^{-6})(4861.35\ \text{Å}) = 0.0475\ \text{Å}. The published expected displacement was about $0.05\ \text{Å}$ — some five thousandths of a millimetre on the photographic plate. Our independent recomputation agrees. For scale, the first-order splitting of the two lines is $\pm\beta\lambda_0 \approx \pm21.5\ \text{Å}$ — four hundred and fifty times larger than the effect being hunted, and it cancels exactly in the mean.

Step 5 — read the structure. At the lowest voltage they used, $6788\ \mathrm{V}$ , the same chain gives $\beta = 2.689\times10^{-3}$ and $\Delta\lambda = 0.0176\ \text{Å}$ . The shift scales as $\beta^2 \propto V$ — linearly in the accelerating voltage — and it was that linear dependence, traced across a range of voltages, that made the result convincing rather than merely suggestive.

2 · A stored lithium beam at beta = 0.338

The 2014 Botermann experiment circulated $\mathrm{Li}^+$ ions at $\beta = 0.338$ in a storage ring and used laser spectroscopy to compare the moving ions' clock with a laboratory one. What transverse shift does relativity predict?

Step 1 — identify. We want $1/\gamma$ , the pure time-dilation factor.

Step 2 — apply. $\gamma = 1/\sqrt{1 - 0.338^2} = 1/\sqrt{0.885756} = 1.062534$ , so $\nu_\perp/\nu_0 = 1/\gamma = 0.941146$ .

Step 3 — the boundary number. A $5.89\%$ redshift — enormous by spectroscopic standards, and four thousand times larger than the fractional shift Ives and Stilwell chased. The second-order approximation $1 - \tfrac12\beta^2 = 0.942878$ is already $0.18\%$ off at this speed: at $\beta = 0.338$ you must use the exact $\sqrt{1-\beta^2}$ .

Step 4 — read the structure. This is why storage rings superseded canal rays. The signal grows as $\beta^2$ , so going from $\beta \sim 4\times10^{-3}$ to $\beta \sim 0.34$ multiplies the effect by nearly $6000$ — and the ions can be held and interrogated for as long as you like. Modern tests confirm $1/\gamma$ to a few parts in $10^{8}$ .

Physics of the portrait

Origins · 2 minTransverse Doppler effect and how this was discovered

Transverse Doppler effect

1907–2014 · Bern · New York · Zürich · Darmstadt

The effect was born as a proposal. In 1907, prompted by Johannes Stark's work on canal rays, Einstein published a two-page note in Annalen der Physik — Über die Möglichkeit einer neuen Prüfung des Relativitätsprinzips — pointing out that a fast ion beam ought to show a second-order frequency shift no classical theory could produce, and that this would be a clean test of the relativity principle. It took thirty-one years. In 1938 Herbert Ives and G. R. Stilwell, at Bell Telephone Laboratories, accelerated hydrogen canal rays and photographed the $\mathrm{H}\beta$ line — but they did not attempt the right-angle geometry Einstein had imagined, judging that nobody could hold the angle accurately enough to keep the huge first-order term out. Instead they photographed the forward and backward beams together and measured the displacement of the centre of gravity of the two shifted lines, which is $\gamma\lambda_0$ exactly. It worked, and it stands as the first direct measurement of time dilation. The irony is heavy: Ives was a lifelong opponent of special relativity, and read his own result as confirming Larmor and Lorentz's ether theory rather than Einstein's kinematics — the two predict the same number here, and the experiment could not distinguish them. The right angle itself had to wait: Walter Kündig reached it in 1963 with a Mössbauer absorber on an ultracentrifuge rim, and Hasselkamp, Mondry and Scharmann observed the shift directly at $90^\circ$ in 1979. Modern storage-ring tests — Botermann and colleagues, 2014, $\mathrm{Li}^+$ at $0.338c$ — confirm $1/\gamma$ to parts in $10^{8}$ .

History

1905

— Einstein derives the relativistic Doppler formula; the second-order transverse shift is an immediate corollary.

1907

— Einstein proposes canal rays as a test of the second-order shift, in Über die Möglichkeit einer neuen Prüfung des Relativitätsprinzips.

1938

— Ives and Stilwell measure the displacement of the centroid of the red- and blue-shifted hydrogen lines, confirming the gamma factor without ever observing at a right angle.

1941

— Ives and Stilwell repeat the measurement at higher accelerating voltages.

1963

— Walter Kündig observes the transverse shift with a Mössbauer absorber on an ultracentrifuge rotor.

1979

— Hasselkamp, Mondry and Scharmann observe the transverse shift directly at ninety degrees.

2014

— Botermann and colleagues test time dilation with stored lithium ions at 0.338c.

Going further

Texts. Rindler, Relativity: Special, General and Cosmological, 2nd ed., §4.3 · Jackson, Classical Electrodynamics, 3rd ed., §11.3 · Ives & Stilwell, "An Experimental Study of the Rate of a Moving Atomic Clock", J. Opt. Soc. Am. 28, 215 (1938), and part II, JOSA 31, 369 (1941) · Einstein, "Über die Möglichkeit einer neuen Prüfung des Relativitätsprinzips", Ann. Phys. 23, 197 (1907) · Botermann et al., Phys. Rev. Lett. 113, 120405 (2014).

Provenance. Signature $(+,-,-,-)$ ; angle convention band-locked. All numbers recomputed in verify_aq-sr-transverse-doppler-effect.py. Base defects repaired. D2: the base printed $1/\gamma - 1 \approx -\tfrac12\beta^2 - \tfrac38\beta^4$ ; the correct fourth-order coefficient is $-\tfrac18$ , and the $\tfrac38$ belongs to $\gamma - 1$ . D3: the base printed the Ives–Stilwell mean as $\nu_{\text{avg}} \approx \nu_0(1 - \tfrac12\beta^2)$ — the wrong sign. The truth is $\nu_{\text{avg}} = \gamma\nu_0$ exactly, and $\lambda_{\text{avg}} = \gamma\lambda_0$ : the centroid is displaced to the red, and that displacement is the experiment's observable. Base defect found here: the base's second derivation ends with the assertion that "the product of the two shifts equals unity is false". The product is $(1/\gamma)\cdot\gamma = 1$ , exactly, for every $\beta$ — a two-line calculation, asserted symbolically in the verify script. That reciprocity is this article's owned structural result.

Scope. The master formula and the wave four-vector are owned by aq-sr-doppler-four-momentum; aberration by aq-sr-relativistic-aberration; the longitudinal shifts by aq-sr-longitudinal-doppler-effect; the range of $\mathcal{D}$ by aq-sr-relativistic-beaming. None is re-derived here.

Builds on. Longitudinal Relativistic Doppler Effect · Relativistic Doppler and the Photon Four-Momentum · Relativistic Aberration of Light. Leads to. Relativistic Headlight (Beaming) Effect · The Twin Paradox · Doppler-Shift Resolution of the Twin Paradox.

Is the transverse Doppler effect a redshift or a blueshift?

Both, and the question is incomplete as asked. A right angle in the observer's frame gives a redshift of $1/\gamma$ — this is the one usually meant, and it is the one that is pure time dilation. A right angle in the emitter's frame gives a blueshift of $\gamma$ — because aberration tips that ray forward, so you catch it while the source is still, on balance, approaching, and the forward tilt over-compensates the slow clock. They describe different rays emitted at different moments. Their product is exactly $1$ . Always name the frame.

Why is there no classical transverse Doppler effect at all, not even a small one?

Because classical Doppler is entirely about the rate of change of distance. The classical shift is $1/(1+\beta_s\cos\psi)$ , which at $\psi = 90^\circ$ is exactly $1$ — not approximately, and at every order in $\beta_s$ . The source is neither closing nor retreating along the line the wave travels, so no crest has further to fly than the one before it. The relativistic effect is not a correction to this: it has a completely different cause — the source's clock — which classical physics does not have. That is what makes the transverse measurement a test rather than a check.

Why did Ives and Stilwell not just observe at ninety degrees?

Because at $\beta \approx 4\times10^{-3}$ the first-order longitudinal shift is $\pm 21.5\ \text{Å}$ and the second-order transverse shift is $0.0475\ \text{Å}$ — a ratio of $450$ . To keep the first-order contamination below the signal you would have to hold the observation angle to within about $0.1^\circ$ of a true right angle, on a diffuse ion beam of finite width, in 1938. They judged it impossible and found a better way: photograph both longitudinal lines and average them. The first-order terms cancel identically ( $k$ and $1/k$ are reciprocals) and what survives is exactly $\gamma$ . It is a lovely piece of experimental design and it is why the null geometry was never needed.

What is the significance of k plus one over k equalling two gamma?

It is the whole mechanism of Ives–Stilwell, in one line. With $k = \sqrt{(1+\beta)/(1-\beta)}$ the two longitudinal lines sit at $k\nu_0$ and $\nu_0/k$ ; their mean is $\tfrac12\nu_0(k+1/k)$ , and $k + 1/k = 2\gamma$ identically. So the mean is $\gamma\nu_0$ : the first-order shifts cancel and the second-order factor stands alone. Read in wavelength, the same identity gives $\lambda_{\text{avg}} = \gamma\lambda_0$ . (It is also the Bondi $k$ -factor, and $\gamma = \tfrac12(k+1/k)$ is the identity behind the Doppler resolution of the twin paradox — see aq-sr-twin-paradox-doppler-resolution.)

Which direction genuinely has no Doppler shift at all?

Neither right angle. Setting $\mathcal{D} = 1$ gives $\cos\theta = (\gamma-1)/(\gamma\beta)$ , which at $\beta = 0.5$ is $\theta = 74.46^\circ$ — a third ray, inside the forward hemisphere, at a right angle in no frame. The lab right angle ( $90^\circ$ ) is already in the redshifted region; you have to swing forward of it to find the null. This is the cleanest way to feel that "transverse" and "unshifted" are simply different conditions, and that the transverse redshift is a real, directionally-specific physical effect rather than a bookkeeping residue.

Is the transverse value the smallest the Doppler factor can be?

No — and this is a common, consequential error. $\mathcal{D}$ ranges over $\gamma(1-\beta) \le \mathcal{D} \le \gamma(1+\beta)$ , and $1/\gamma$ sits strictly inside. The floor is $\gamma(1-\beta) \to 1/(2\gamma)$ , attained dead astern, smaller than $1/\gamma$ by a factor $(1+\beta)$ : at $\beta = 0.9$ the floor is $0.2294$ where $1/\gamma$ is $0.4359$ . A receding source is redshifted far harder than a transverse one. See aq-sr-relativistic-beaming.

Transverse Doppler Effect

Where this lives today

The physical picture

The core equations

Where the transverse shift comes from

Limiting cases

Worked examples

1 · Ives and Stilwell, 1938 — recomputing the shift they hunted

2 · A stored lithium beam at beta = 0.338

Physics of the portrait

Going further

Before this

Where this leads