Peculiar Velocity & the CMB Dipole

The universe has a speedometer, and you are on it

Stand still on a calm day and you feel no wind. Walk forward and a breeze appears on your face, faster air ahead and a hush behind. The cosmic microwave background — the relic glow of the hot early universe, now cooled to a near-perfect 2.72548 K blackbody — plays exactly the same trick. In a frame at rest with respect to it, the CMB is the same temperature in every direction to a few parts in a hundred thousand. But we are not at rest. The Sun drags the Earth around the Milky Way, the Milky Way falls toward Andromeda, the Local Group streams toward distant superclusters, and the sum of all that motion blows a "microwave wind" across the sky. The radiation we run into looks hotter; the radiation we leave behind looks cooler. That hot-pole/cold-pole pattern is the CMB dipole, and the motion that creates it is our peculiar velocity.

The word "peculiar" here is old usage meaning "particular to this object" — your own motion, as distinct from the universal expansion everyone shares. It is the single most cleanly measured velocity in cosmology: not modelled, not inferred from a ladder of assumptions, but read almost directly off the brightness of the sky.

Hubble flow versus genuine motion

Every galaxy's observed redshift mixes two completely different things. The first is the Hubble flow: space itself stretches, so distant galaxies recede at

v_rec = H₀ d        (H₀ ≈ 70 km/s/Mpc)

This is not motion through space — no galaxy is travelling anywhere; the ruler between us is lengthening. The second is the peculiar velocity v_pec, a true motion through the local space, set by the gravitational tug of nearby mass. The total line-of-sight velocity you infer from a spectrum is, to low order,

cz ≈ H₀ d + v_pec        (peculiar velocity along the line of sight)

Close to home the peculiar term wins. The Andromeda Galaxy (M31), 0.78 Mpc away, has a tiny Hubble velocity (~55 km/s) but is blueshifted — it is falling toward us at about 110 km/s, so its net motion is inward. Far away the Hubble term dominates and peculiar velocities become a small percentage scatter on the smooth expansion. Disentangling the two is the central problem of peculiar-velocity cosmology, because v_pec is sourced by gravity and therefore traces all the mass — visible and dark — out to hundreds of megaparsecs.

Why our motion paints a cosine on the sky

Take the CMB to be a perfect blackbody at temperature T₀ in its own rest frame. An observer moving at speed v sees each photon Doppler-shifted by an amount that depends on the angle θ between the line of sight and the direction of motion. A blackbody stays a blackbody under a boost — it just changes temperature direction by direction:

T(θ) = T₀ √(1 − β²) / (1 − β cos θ),     β = v/c

Expanded for small β:
T(θ) ≈ T₀ [ 1 + β cos θ + β² (cos²θ − 1/2) + … ]

The leading term, β cos θ, is a pure dipole — maximum hot exactly ahead (θ = 0), maximum cold exactly behind (θ = 180°), and a smooth cosine in between. Its amplitude is

ΔT_dipole = (v/c) T₀
          = (369.8 / 299792) × 2.72548 K
          ≈ 3.36 × 10⁻³ K  =  3.36 mK

So a 370 km/s motion — about 0.123% of the speed of light — tilts the whole microwave sky by 3.36 millikelvin. That is enormous compared to the intrinsic primordial fluctuations, which are only ~18 µK in the dipole-subtracted map: the kinematic dipole is roughly 100 times larger than the cosmological signal cosmologists actually want, which is why it is always the very first thing subtracted from any CMB map. The next term, of order β², is the tiny kinematic quadrupole (a few µK) plus a frequency-dependent modulation of the small-scale anisotropies that Planck used to confirm the dipole is genuinely kinematic and not a leftover primordial feature.

The measured numbers

The dipole was first detected in the late 1970s by balloon and U-2 aircraft experiments (Conklin 1969; Smoot, Gorenstein & Muller 1977). COBE pinned it down in the 1990s, WMAP refined it, and Planck delivered the current gold-standard values.

Quantity	Value	Meaning
CMB monopole T₀	2.72548 ± 0.00057 K	Mean temperature (FIRAS)
Dipole amplitude ΔT	3.3621 ± 0.0010 mK	Planck 2018
Solar System speed v☉	369.82 ± 0.11 km/s	From β = ΔT/T₀
Apex direction (l, b)	(264.021°, 48.253°)	Galactic coords, in Leo/Crater
ΔT / T₀	1.2336 × 10⁻³	Our speed in units of c
Local Group speed	620 ± 15 km/s	Toward (l, b) ≈ (276°, 30°)
Primordial anisotropy	~18 µK (rms)	The signal left after subtraction

The chain from "what the sky shows" to "how fast the Local Group moves" is built by removing the velocities we already know. The Sun orbits the Galactic centre at ~233 km/s; the Sun also moves ~18 km/s relative to the Local Standard of Rest; the Milky Way moves within the Local Group. Subtract all those vectors from the 369.8 km/s heliocentric CMB velocity and what remains is the Local Group's bulk peculiar velocity of about 620 km/s — a single number that encodes the gravitational pull of everything around us.

How peculiar velocities are measured for other galaxies

The CMB dipole gives our peculiar velocity for free, but mapping the velocity field of other galaxies requires a redshift-independent distance. The recipe is always the same: measure a distance some way that does not use redshift, predict the Hubble-flow velocity, and call the leftover the peculiar velocity.

v_pec = cz − H₀ d        (line-of-sight, low-redshift approximation)

• Tully-Fisher relation. A spiral galaxy's luminosity scales with its rotation speed (L ∝ v_rot⁴). Measure the rotation from the 21 cm line width, read off the intrinsic luminosity, compare to the apparent brightness for a distance. Workhorse of surveys like Cosmicflows.
• Fundamental Plane. The elliptical-galaxy analogue, linking size, surface brightness, and velocity dispersion. Used by the 6dF and SDSS peculiar-velocity samples.
• Type Ia supernovae. Standardizable candles good to ~5–7% in distance — the most precise single-object peculiar-velocity probes, though rare.
• Surface-brightness fluctuations. The graininess of an unresolved stellar population, calibrated as a distance indicator for nearby ellipticals.

Because the error on v_pec is roughly the fractional distance error times cz, peculiar-velocity surveys are only useful nearby: a 5% distance error at cz = 6000 km/s already injects ±300 km/s of noise, comparable to the signal. This is why direct velocity-field mapping is confined to within ~150–200 Mpc, while statistical "bulk flow" estimates push a little further.

Where we are actually being pulled

The Local Group's 620 km/s is not random — it points at concentrations of mass. The story unfolded over decades:

• The Great Attractor. In the 1980s the "Seven Samurai" team found galaxies across a huge patch of sky streaming coherently toward a region in Norma/Centaurus, about 65 Mpc (≈ 210 million light-years) away, partly hidden behind the Milky Way's "Zone of Avoidance." This Great Attractor — anchored by the Norma Cluster (Abell 3627) — accounts for a large share of the motion.
• The Shapley Supercluster. Look further along the same direction and you find the densest concentration of galaxies in the nearby universe, ~200 Mpc (≈ 650 million light-years) away. Analyses attribute roughly half of the Local Group's motion to Shapley's distant pull — the Great Attractor was partly a foreground waypoint on the road to Shapley.
• The Dipole Repeller. In 2017 the Cosmicflows team showed our motion is equally shaped by a void — an underdense region in the anti-apex direction that effectively "pushes" us (less pull from behind). Motion in the cosmic web is push and pull together.
• Laniakea. The same velocity-field mapping defined our home supercluster, "Laniakea" (2014), a basin of ~10¹⁷ solar masses spanning 160 Mpc within which all galaxies, including ours, flow toward the Great Attractor.

The deep point is that peculiar velocity is gravity made visible. The bulk flow integrates the gravitational acceleration from the entire surrounding mass distribution — including dark matter — so comparing the measured velocity field to the galaxy density field tests the growth of cosmic structure and the parameter combination β = Ω_m^0.55 / b that links the two.

Peculiar velocity distorts the maps too

Peculiar velocities are not just a thing we measure — they actively warp our three-dimensional maps of the universe, because we infer distance from redshift and redshift is contaminated by v_pec. These redshift-space distortions have two famous signatures:

• The Kaiser effect (large scales): galaxies falling coherently into a forming cluster get their infall added to the Hubble flow, squashing the structure along the line of sight — superclusters look flattened, "pancaked" toward the observer.
• The Fingers of God (small scales): inside a virialized cluster, random orbital velocities of hundreds to ~1000 km/s smear the galaxies into long radial streaks pointing straight at Earth, as if every cluster has spikes aimed at us.

Both are pure peculiar-velocity artefacts, and cosmologists turn the problem into a tool: the amplitude of the Kaiser squashing measures the growth rate of structure, fσ₈, one of the sharpest tests of gravity on cosmological scales.

The radio dipole anomaly

If the CMB dipole is purely kinematic, then the same 369.8 km/s motion should produce a matching dipole in the counts of distant radio galaxies and quasars — sources so far away they share the cosmic rest frame. The expected amplitude follows the Ellis & Baldwin (1984) relation, scaling as a few times β.

Several studies (notably Secrest et al. 2021, using ~1.4 million quasars from WISE) find a number-count dipole pointing in roughly the CMB direction but with about twice the expected amplitude, a discrepancy at the 4–5σ level. If real, it would mean either we are moving faster than the CMB dipole implies, or the universe has a genuine large-scale anisotropy violating the Cosmological Principle. The result is hotly debated — selection effects, source evolution, and local structure are all candidate culprits — but it is one of the live tensions in modern cosmology, sitting alongside the Hubble tension as a possible crack in the standard model.

Common misconceptions and edge cases

• "The dipole tells us the early universe was lopsided." No — it is a kinematic effect of our own motion. The primordial CMB is dipole-free to the precision we can test; the 3.36 mK pattern is entirely ours. That is exactly why it is the first thing subtracted from any CMB analysis.
• "Recession velocity is peculiar velocity at large distance." They are physically distinct. Recession is the stretching of space (no motion through space, no Doppler shift in the local sense); peculiar velocity is real motion that does Doppler-shift light. Conflating them breaks the bookkeeping — the famous "galaxies receding faster than light" only makes sense for recession, never for peculiar motion.
• "The CMB rest frame violates relativity." It does not. Special relativity forbids a preferred frame for the laws of physics; it never said the universe's contents can't pick out a convenient frame. The CMB frame is just the one in which the radiation is isotropic — a cosmic still-air frame.
• "Peculiar velocities are negligible." Locally they are decisive. Within ~30 Mpc, peculiar velocities of several hundred km/s rival the Hubble flow and are a leading systematic in measuring H₀ — which is precisely why local-distance-ladder teams correct for the velocity field before quoting a Hubble constant.
• "Relativistic Doppler and classical Doppler give the same dipole." They agree to first order, but the classical formula misses the β² quadrupole and the relativistic aberration that slightly shifts where structures appear — corrections Planck actually measures.