The Hydrogen 21-cm Line and Its Applications to Radio Astrophysics

Lulu Liu (Partner: Chris Chronopoulos)

MIT Undergraduate
(Dated: March 9, 2008)
We explore some applications of the 21-cm neutral hydrogen line in astrophysics. From observa-
tions in a narrow band around 21 cm we find estimates for the peak temperature of the chromosphere
of the sun, as well as the angular extent of a portion of the galactic plane around l = 30 . We de-
termine the rotation curve of the Milky Way galaxy (for r < R ) and explore its spiral structure
for galactic longitudes 30 < l < 80 . We find that the speed of rotation of the galaxy v(r) rises
linearly in r from r = 0 to r 2.5kpc at which point it remains a constant up to at least R . This
behavior deviates significantly from Keplerian predictions for large r and the expected rotational
curve generated by all visible matter in the galaxy. It is clear that the current models for the
dynamics of the Milky Way galaxy are insufficient to describe the actual physics of this system. A
possible explanation lies in an extensive dark matter halo.

1. INTRODUCTION feature allows for highly accurate determinations of H1

source velocity by simple measurements of the doppler
It is well known in optics that as the wavelength of shift of the 21 cm line. Using the relativistic doppler
light increases, attenuation through a given material de- shift formula, with = vrec /c,
creases. It is no wonder, then, that radio waves reach s
us from parts of deep space that optical light does not. 1
obs = source (1)
In particular, clouds of interstellar gas that would other- 1+
wise hide optical features in our galaxy can reveal much
about general galactic structure and dynamics if we look we can obtain the recessional velocity, vrec , by solving
in a frequency band around 1420.4 MHz. This is known the equation for .
as the 21 cm line. Emitted by neutral hydrogen atoms,
this line can be seen with varying intensity coming from
all directions in the sky, and due to its extremely sharp 3. EQUIPMENT AND TEMPERATURE
nature (very little dispersion in energy), is used widely CALIBRATION
in astronomy for spectroscopic velocity measurements.
For our experiment we have access to the Small Ra-
dio Telescope on the roof of Building 26 developed by
Haystack Laboratories. A motor and feedback loop con-
2. THEORY OF THE 21-CM LINE
trols the pointing of the telescope. Radio signals received
are fourier transformed and sorted into frequency bins
A neutral hydrogen (H1) atom has one proton and one in a multi-channel accumulated spectrum of counts con-
electron. Both particles have quantized total spin (s) of verted to power (in K) against frequency.
1/2 and a spin in the z-direction (m) of either 1/2 or A temperature calibration is performed internally by
-1/2. Spin-spin coupling is the result. It was discovered the electronics of the system. A noise diode emitting
that what was once thought to be a degeneracy in the black body radiation at 115 K is activated at the vertex
ground state of the hydrogen atom is in fact a very fine of the parabolic dish. The effect is a signal that fills up
energy level split between a more energetic parallel spin the entire antenna and is read by the feed horn. Subse-
state |me = 12 , mp = 21 i and a less energetic anti-parallel quent signals are compared against this calibration and
ground state |me = 12 , mp = 21 i. The perturbing Hamil- converted into an effective temperature reading that is
tonian here is given by H = ~ ~ e . When a hydrogen
p B proportional to the power received. The dish has a half-
atom transitions from the excited state into the ground power beamwidth of approximately 7.5 [5]. We point
state, a photon is emitted carrying away the energy dif- our receiver to a patch of the cold sky (elevation > 50 )
ference. This photon has a wavelength of 21 cm. in order to perform our calibration and obtain a baseline
The spin-flip transition, as it is termed, is a highly for- system noise temperature, Tsys .
bidden process, with a mean-life of approximately 107
years. The Heisenberg Uncertainty relationship then
leads us to expect a very sharp emission line with small 4. TELESCOPE POINTING (25-POINT SCAN)
E energy dispersion, or line width, in frequency. This
We use the sun as a means of correcting pointing errors
on the telescope. An Npoint scan of the sun creates a 5
5 grid around a central point and takes a temperature
Electronic address: lululiu@mit.edu measurement at each point in the grid. 25 temperatures
 2

are recorded in all and can be used to create a tempera-

ture contour plot of the sun, see Figure 1. The telescope
recenters on the point of maximal temperature; the off-
sets are recorded. By using this method, we find a general
repeatable pointing error of approximately 2 .

FIG. 2: Drift scan over the sun. The Bessel diffraction pattern
is smeared out by the finite angular extent of the sun and the
resulting curve is a reasonable gaussian fit. We use 2 on the
gaussian distribution as an approximation for the first minimum of
the Bessel function.

FIG. 1: A 25-point scan of the sky surrounding the sun. Contour

curves are shown.

5. BEAM WIDTH DETERMINATION AND

PEAK TEMPERATURE OF THE SUN

We point the telescope approximately 45 minutes

ahead of the sun. By turning off the tracking mechanism,
we allow the sun to drift through our antenna beam, tak-
ing integrated readings approximately 10 seconds apart.
We then sum over the power recorded in each bin in or-
der to obtain an average power per bin as a function of
time. This manipulation, as well as the integration time,
is factored into the error on each data point. Our results FIG. 3: Drift scan over the galactic plane around l = 30 . A
from the scan is given in Figure 2. similar procedure is used to extract the first minimum.
The Fraunhofer Diffration pattern, intensity as a func-
tion of angle , is described in the following relation in-
volving the Bessel function J1 [4]: and the reduction of a circularly polarized radiation to a
linearly polarized receiver. We find a Tsun of 3.5 104K.
 2
2J1 (kasin)  2
7.5
I() = I0 (2) Tsun = 2 77.5K = (3.5 .3) 104 K (3)
kasin .52
The theoretical first minimum occurs at sin = Here, we use the half-power beamwidth 7.5 instead of
1.22/D. Here, = 21 cm and D = 8.2 f t. We find the full beamwidth in order to account for the non-
it to be 5.87 for our setup. Comparing with the 2 uniform receiving power of the antenna. Since this re-
obtained from our gaussian fit, we find reasonable agree- sult is now far above the known surface temperature
ment, considering 2 is a slight underestimate of the first of the sun ( 6000K), we are led to believe that the
minimum of a corresponding Bessel function. bulk of the 21 cm radiation from the sun has its ori-
From the gaussian fit, we find a peak temperature of gins in its chromosphere as opposed to its surface. In
77.5K, which is far below any reasonable limit for the another trial, we allow a portion of the galactic plane
temperature of the sun. Corrections are applied to ac- around l = 30 to drift through our beam width. See
count for both the fraction of the beam width occupied Figure 3. The 2 width is now 8.7 . We use a simpli-
by the sun (with an angular extent of approximately 0.5 ) fied model and treat both solar and galactic emissions as
 3

step functions to be convolved with the diffracting beam at a given radius:

pattern. We estimate the bulk emission from this part of
the galaxy to be coming from an angular region approx- Z r Z r
imately 8.7 5.48 0.5 = (2.7 0.2) across. M (r) = [(r)](2r)dr = [0 er/b ](2r)dr
0 0
h   i
M (r) = 0 b2 1 er/b rber/b (5)

6. EXPLORING GALACTIC STRUCTURE AND

Here, b is a characteristic radius, taken to be about 3 kpc,
DYNAMICS
and 0 is the maximum surface mass density. From there
we
p can quickly derive a v(r) using the relation v(r) =
Recalling that the sharpness of the 21-cm hydrogen GM (r)/r. We find that the limiting behavior for v(r)

line allows for very precise doppler spectroscopy, we make at large r is also Keplerian (goes as 1/ r). Since in
use of the abundance of neutral hydrogen in the Milky reality, we will be observing clumps of gravitating matter
Way to derive a galactic rotation curve (a v(r) for every as opposed to a single point mass, the v(r) from here on
r). We model the galaxy as a differentially rotating thin will represent an average velocity.
disk with stellar and interstellar material in circular or-
bits around a shared origin. This symmetry assumption
allows the average velocity to be independent of polar

v
angle , = 0. Knowledge of an explicit velocity curve 6.2. Experimental Methods
then allows dramatic insight into the distribution of mass
within the galaxy and eventually leads to determination
of galactic structure. We point our telescope to various galactic longitudes l
and collect doppler shifted 21-cm spectra. However, there
is no way to separate the accumulated received radiation
by source distance. In order to derive a rotation curve, we
will need to be able to associate each measured velocity
6.1. Velocity Curve Predictions
with a non-ambiguous radius. We do this by considering
only the maximum red/blue shift in each direction. We
Using Newtonian mechanics we can calculate the ro- propose that within 0 < l < 90, the maximum recessional
tational velocity of a point mass at a distance r from velocity corresponds to the matter traveling directly away
a gravitational center with a total mass of M . Assum- from us in a circle of radius R that is tangent to our line
ing that M is contained within a radius r0 , the orbital of sight, assuming a relatively flat rotation curve, due to
velocity follows the relation, the maximization of the dot product ~v ~s, where ~s is the
line of sight. This is best understood with a diagram
r (Figure 4).
GM 1
v(r) = , r > r0 (4)
r r

This is known as a Keplerian velocity curve, which is

obeyed by the planets in our solar system. For the Milky
Way galaxy, however, this model is unrealistic, as we well
know the mass is not concentrated at the galactic center
but spread out over many kpc. From observational evi-
dence, we know the Milky Way to consist of two visible
parts: a spherical central bulge with radius of approx-
imately r = 1.5 kpc with a constant mass density B
[3], and an outer extent with luminosity falling off ex-
ponentially with r. Considering first the central bulge,
integrating the constant density we find a total enclosed
mass M (r) proportional to r3 and a v(r) that increases
linearly in r. FIG. 4: The maximum recessional velocity along a line of sight
corresponds to matter moving in a circle with radius R around the
Treating the outer regions is only slightly more in- galactic center.
volved. We begin with the empirically determined sur-
face luminosity density curve for the galactic disk, (r) in
units of [L /kpc2 ] [2]. Using two-dimensional geometry The only visible portion of the galactic plane suitable
now, we assign an average mass-to-light ratio, 0.6, for this analysis has 0 < l < 90. We sweep this range to
to convert to surface mass density in units of [M /kpc2 ]. find maximal vrec in each direction and use trigonometry
Finally, we integrate over r to find total enclosed mass to find its associated radius R = R sin(l).
 4

6.3. Calibration A second correction of importance removes the effect

of the receivers motion relative to the galactic center.
We use a frequency generator to conduct a 3-point cal- This includes earths 30 km/s orbit around the sun and
ibration of our output. This procedure converts each bin the suns 20 km/s motion in the direction of RA 18h
in our accumulated spectrum into a calibrated frequency Dec 30 relative to the local standard of rest (LSR). We
value. We obtain calibration errors through a linear fit. use vector geometry to calculate the component of this
combined motion along our line of sight. See Figure 7.

FIG. 7: VLSR correction geometry.

Finally, the rotational velocity of the source can be

FIG. 5: Calibration of spectrum. Frequency vs. bin number. derived from the corrected recessional velocity and the
galactic longitude using the following relation which takes
into account the orbital motion of the LSR around the
galactic center,
6.4. Corrections - VLSR and Dispersion
v = vrec + vLSR sin(l) (6)
Two key corrections were made to our raw data in
order to obtain the velocity profile. First, in order to ac-
count for the velocity dispersion due to relative motion 6.5. Data and Results
of gas clusters in a given region of the galaxy, we point
our telescopes exactly anti-center, at l = 180. Since we
expect no net red- or blueshift in this direction (average
motion perpendicular to our line of sight), the width of
the spectrum can be used to determine a representative
velocity disperson. See Figure 6. We subtract this ex-
pected dispersion from the maximum redshift along our
line of sight to find the maximum average recessional ve-
locity.

FIG. 8: Spectrum taken at galactic longitude 80 and lattitude

0 . The integration time is approximately 5 minutes.

A sample spectrum is shown in Figure 8. We found the

maximum redshift in this case to be about 22.8 km/s, and
an average recessional velocity of approximately 3 km/s
after dispersion factors have been accounted for.
FIG. 6: 21cm power spectrum obtained from observation along l = A sweep from longitude 80 to longitude 5 was made
180. We find the velocity dispersion to be approximately 20km/s. with spectra taken every 5 degrees. We found there to
 5

expected Keplerian 1/ r decay. In fact, the rotation
curve appears to be a constant for R > 2.5kpc and shows
no indication of falling off.
A possible explanation for this phenomenon is a pro-
posed dark matter halo surrounding the visible matter
in the galaxy with a mass density d (r) that falls off as
1/r2 [1]. Integrating d over a sphere indicates a total
enclosed mass M (r) proportional to r, and a constant
rotational velocity, v(r). It is proposed that this dark
matter halo extends many kpc beyond the orbital radius
of the sun and the rest of the visible galaxy. Beyond this
dark halo, the velocity curve is expected to be Keplerian.

6.7. Spiral Arms of the Milky Way

Revisiting our 21cm spectra we notice that there is

TABLE I: Table of corrected recessional velocities vrec obtained more information contained within the peaks than the
for each radius r. simple maximal recessional velocity which we used in the
previous few sections to derive the rotational curve of
the galaxy. In order to extract information about galac-
be a general smearing of the spectrum as we neared the tic structure, we turn to the relative intensities at dif-
galactic center due in part to the increased noise (de- ferent redshifts. Extrapolating between the data points
creased pointing elevation) and in part to the additional on our rotation curve, we plot the locations of the max-
thickness of the emitting region and increase in velocity imas, assuming these to be the regions of highest H1
dispersion. The derived rotation curve using the meth- concentration, along our line of sight in each direction l.
ods outlined in section 6.2 is given in Table I and Figure Trigonometry is helpful in deriving these relations.
9 with comparison to Keplerian and visible matter pre- We determine geometrically radius r and angle of
dictions plotted with dashed lines. these maxima given the corresponding vrec and l. We
interpolate the given vrec in the following function for r:
hr i

vrec (r) = F (r, v(r)) = sin(l) v(r) vLSR (7)
r
From r, we can now find the angle :
 hr i

(r) = G(r) = l + sin1 sin(l) (8)
r

FIG. 9: Rotation curve for the Milky Way galaxy for R R .

Comparison to expected velocity functions.

6.6. Discussion
FIG. 10: Plot of the sources of maximum intensity 21 cm emission
between 30 < l < 80. We see possible evidence of two distinct
We see clearly non-Keplerian behavior in our rotation density waves (spiral arms) of high H1 concentration.
curve. For small R, the rotational velocity rises linearly
in r as predicted by the bulge model, however for large Performing this analysis for each redshifted peak along
R the observed behavior deviates significantly from the 30 < l < 80, we obtain the following polar plot showing
 6

the locations of highest H1 density in the viewable por- corresponds to the radius of closest approach is obviously
tion of the galactic plane (Figure 10). There is evidence called into question by our velocity curve itself. Wrong
of some spiral structure as indicated by two possible arms attributions of maximal velocity would result in an over-
of high density H1 gas, whose locations are approximated estimate of v(r) at small r. In addition, the assumption
by superimposed curves to aid viewing. The outer curve of perfectly circular rotation around the galactic center
is likely the Orion arm, of which the solar system is a is refuted by evidence of spiral arms of high mass den-
part, and the inner may be a short portion of the Saggi- sity with radial components. To some extent, this will
tarius Arm. affect our velocity curve, but to a greater extent, it will
change the locations of the maxima in our plot of density
maxima and warp our spiral arms.
7. ERRORS We would like to also point out certain limitations
to our method in estimating velocity dispersion. Since
The empirically determined parameters used in this stronger gravitational interactions would likely result in
experiment, as a whole, were a surprisingly large contrib- a wider distribution of velocities, regions of lower stellar
utor to general error. The orbital radius of the sun, R , density would not necessarily suffer the same dispersive
for example, is only known to approximately 6-10% ac- forces as regions of higher stellar density. This would
curacy, and was the largest contributor to uncertainty in result in an underestimate of certain velocities and an
r in our galactic rotation curve. Similarly, vLSR and v overestimate of others.
were known only approximately. Systematic inaccuracy
of our results may have resulted from these uncertainties.
After an initial confusion regarding the dipole noise
generator (the result of a poor connection), very little 8. CONCLUSIONS
error can be attributed to our electronics. Calibration
errors became negligible after all other factors were taken In the earlier portions of our lab, we determined the
into consideration. The pointing error, determined by re- chromospheric temperature of the sun to be approxi-
peated scans of the sun, was approximately 2 and were mately (3.5 .3) 104 K, and the full beamwidth of
corrected occasionally with offsets to eliminate its sys- the telescope to correspond to a the circular diffraction
tematic effects. theorys prediction of 1.22/D. Additionally, we mea-
The need for a radio-quiet environment for high- sured the thickness of the peak 21cm emitting hydrogen
accuracy radio astrophysics became apparent after only a in the region of galactic longitude 30 and found it to be
few trials. Aside from noise within the electronics (tem- approximately (2.7 0.2) in angular extent.
perature calibrations determine Tsys ), radio towers, cell In the latter parts of our experiment we went into some
phones, even satellites distorted our spectra immensely. detail investigating the dynamics and structure of the
Although mostly contributing a noise floor, as we moved Milky Way galaxy. We derived a rotation curve from
to lower galactic longitudes and therefore nearer to the doppler spectroscopy performed on 21-cm H1 emissions
horizon, not only were there occasionally loud signals and found it to be non-Keplerian in nature, indicating
that blotted out the entire spectrum, but the noise floor some misalignment between the expected gravitational
also shifted and began to take on some slope. This effects due to visible matter and the actual dynamics of
made determination of maximum redshift difficult and our galaxy. We offered the existence of a dark matter
although we attempted to correct for this effect by tak- halo as a possible explanation of this phenomenon, sup-
ing a noise background at a similar elevation away from ported by our experimental results. Additional analysis
the galactic plane, contributed significantly to our error of the 21-cm spectra in various directions in the galactic
in the lower longitudes. plane indicated the presence of 2 possible spiral arms.
Non-validity ranges in some of our assumptions is a Due to the large uncertainties in this particular analysis
matter of greater concern. These errors contribute sys- the results are interesting but inconclusive and definitely
tematically to our results and may result in significant warrants further investigation. In the end, we believe we
corrections and shifts. The assumption that v(r) is have demonstrated the 21-cm neutral hydrogen line to be
mostly flat and therefore the highest recessional velocity a powerful tool for exploring our galaxy and beyond.

[1] Drees, M. Dark Matter. Technical University, Munich. nexions. [2005]

[2003] [5] Sewell, 21 cm Radio Astrophysics, 8.14 Course Reader,
[2] Mihos, Chris. Rotation Curves. Case Western Reserve [2007]
University. http://burro.astr.cwru.edu/ [2001] [6] van de Hulst, H.C., Muller, C.A., Oort, J.H. Spiral Struc-
[3] University of Oregon Physics. ture... Derived from Hydrogen Emission at 21cm Wave-
http://physics.uoregon.edu/ jimbrau/ [2001] length. Bulletin of Astronomical Institutes of Nether-
[4] Diffraction from a Circular Aperture. Paul Padley, Con- lands. Volume XII. Number 452. [1954]
 7

[7] Shu, F.H. Our Galaxy: The Milky Way System. The
Physical Universe - Chapter 12. University Science Books.
[1982]

Acknowledgments

All non-linear fits were made with the MATLAB

scripts made available to us on the Junior Lab website:
http://web.mit.edu/8.13/www/jlmatlab.shtml.
We would like to thank the Junior Lab staff for their
assistance and Scott Sewell in particular for fixing the
telescope repeatedly and in many ways so that this ex-
periment can be possible.