A radio wave crossing ionised gas travels a little slower than light in vacuum, and the shortfall grows as the frequency falls. Summed along the path, the extra travel time is the dispersion measure, the number of free electrons in a column of unit cross-section, times a constant, divided by the square of the frequency. This chapter derives that delay from the motion of free electrons, shows why it depends only on the column and not on where the electrons sit, and splits the total a telescope measures into the Milky Way, the space between galaxies and the host galaxy.
You need the wave relation ω=ck for light in vacuum, the idea of a group velocity, and the meaning of redshift as a stretch of wavelength by 1+z.
Waves in a thin plasma
In a plasma the free electrons are pushed back and forth by the wave's electric field. Their current changes the wave relation to
ω2=ωp2+c2k2,ωp2=ε0menee2,
where ne is the free-electron density and ωp the plasma frequency. A pulse carries its energy at the group velocity,
vg=dkdω=c1−ω2ωp2.
Waves below the plasma frequency cannot travel at all. Waves far above it travel almost at c. Petroff et al., 2022, Section 4.1 treats the resulting dispersion measure as the basic observable of the whole field.
Worked example: how far above the plasma frequency are we?Show the answer
The plasma frequency in hertz is fp=8.98kHzne/cm−3. Warm gas in the Milky Way disc has ne≈0.03cm−3, so fp≈1.6kHz. If all the ordinary matter in the universe were ionised and spread evenly, today's mean density would be 2.19×10−7cm−3 and fp≈4Hz. A burst observed at 1GHz is at least 105 times above both, so ωp2/ω2 is below 10−10 and the first-order expansion below is excellent.
From group velocity to the dispersion measure
The travel time over a path of length L is ∫dl/vg. Expanding the square root to first order in ωp2/ω2,
The first term is the vacuum travel time. The second is the extra delay. It depends on the path only through the integral
DM=∫0Lnedl,
the dispersion measure. Astronomers quote it in pccm−3. With the constants evaluated,
Δt=4.149ms(pccm−3DM)(GHzν)−2.
Only differences between frequencies can be measured, because nobody knows when the burst left its source. A telescope records the sweep across its band and fits the ν−2 curve. Macquart et al., 2020, text before eq. 1 writes the dispersion measure of a distant burst in exactly this way, with each piece of path weighted by 1/(1+z).
Worked example: the sweep of FRB 180924Show the answer
FRB 180924 has DM=361.42pccm−3. Between 1.5 and 1.2GHz the delay difference is 4.149ms×361.42×(1.2−2−1.5−2)=1499.5ms×0.25=375ms. The burst itself lasts about a millisecond, so the sweep is hundreds of times longer than the pulse.
Reading a dispersion measure as a column of electrons
Since one parsec is 3.086×1018cm, a dispersion measure of 1pccm−3 means 3.086×1018 electrons in a tube of 1cm2 cross-section, or 3.086×1022 per square metre. The delay does not know where along the tube those electrons sit. A dense knot and a long thin stretch with the same column give the same sweep.
Two quick scales help. One kiloparsec of Milky Way disc at 0.03cm−3 gives 30pccm−3. One thousand megaparsecs at the evenly spread cosmic density of 2.19×10−7cm−3, ignoring expansion, gives 219pccm−3. The space between galaxies is a million times thinner than the disc, but the path is a million times longer. Prochaska and Zheng, 2019, Section 2
Splitting what the telescope measures
The dispersion measure of a burst at redshift z is a sum of four parts,
The first two are the Milky Way's disc gas and its hot halo. The third is everything between the galaxies, including the gas in any halo the sightline happens to cross. The last is the host galaxy, measured in its own rest frame. Macquart et al., 2020, eq. 1 and the text after it
The factor 1/(1+z) follows from the delay formula. In the host's frame the delay is proportional to DMhost/νe2, where the emitted frequency is νe=(1+z)ν. Cosmic expansion also stretches every time interval by 1+z on its way to us. The observed delay is therefore (1+z)DMhost/[(1+z)2ν2]=DMhost/[(1+z)ν2]. Electrons far away count for less than the same electrons nearby.
Worked example: a host galaxy at redshift one halfShow the answer
A host that would contribute 100pccm−3 to an observer sitting inside it adds only 100/1.5=66.7pccm−3 to the burst we see from z=0.5. Forgetting this factor makes distant hosts look heavier than they are.
Lab task
In the FRB baryon-census Lab, keep the default preset and compare the total a telescope measures with the cosmic part on the current sightline. The difference is the Milky Way's 80pccm−3 plus the host term. Then raise the typical host dispersion and watch only the total move.
Exercises
A burst with DM=500pccm−3 is recorded from 800 down to 400MHz. How long does the sweep last?
How many electrons per square metre lie along a column of 80pccm−3, the Milky Way total assumed by Macquart and colleagues?
A host galaxy contributes 200pccm−3 in its own frame. What does it add to a burst seen from z=1?
Why can a single dispersion measure not tell whether its electrons lie near us or near the source?
AnswersShow the answer
Δt=4.149ms×500×(0.4−2−0.8−2)=2074.5ms×4.6875=9.72s. Low-frequency telescopes see sweeps lasting seconds.
80×3.086×1022=2.47×1024 electrons per square metre.
200/(1+1)=100pccm−3.
The delay depends only on the integral of ne along the path. Position enters only through the redshift factor of each contribution, which the measurement cannot separate from the density. Splitting the total needs other information: a model of the Milky Way, an estimate of the host and the host's redshift.