polarization with the very large array
TRANSCRIPT
Atacama Large Millimeter/submillimeter Array
Karl G. Jansky Very Large Array
Robert C. Byrd Green Bank Telescope
Very Long Baseline Array
Polarization with the Very Large Array
Rick Perley
NRAO -- Socorro
Jansky VLA
Goals of This Presentation
• Polarimetry (especially interferometric polarimetry) is a
daunting topic.
• In this short presentation, I will introduce the subject, and
explain the ‘why, how, when, and where’ of polarimetry.
• Subjects:
– Why do polarimetry?
– What is polarimetry?
– Definitions : polarization, Stokes parameters, Stokes visibilities
– Short, intuitive explanation of how interferometers do polarimetry.
– Summary of calibration of polarimetry data.
– A nice example of imaging polarimetry (the planet Mars).
ASIAA Workshop on the Very Large Array 2
Jansky VLA Why Measure Polarization? • In short – to access extra physics not available in total intensity alone.
– Processes which generate polarized radiation:
• Synchrotron emission: Up to ~80% linearly polarized, with
no circular polarization. Measurement provides information
on strength and orientation of magnetic fields, level of
turbulence.
• Zeeman line splitting: Presence of B-field splits RCP and
LCP components of spectral lines by amount proportional to
the field strength.
– Processes which modify polarization state:
• Faraday rotation: Magneto-ionic region rotates plane of
linear polarization. Measurement of rotation, combined with
density estimate, gives B-field estimate.
• Free electron scattering: Induces a linear polarization
which can indicate the origin of the scattered radiation.
ASIAA Workshop on the Very Large Array
Jansky VLA
Example: Linear Polarization of Cygnus A • VLA @ 8.5 GHz Synchrotron emission B-vectors Perley & Carilli (1996)
• Field strengths in lobes of tens of mG, fields lie along brightness gradients.
10 kpc
ASIAA Workshop on the Very Large Array 4
Jansky VLA
M51 – magnetic fields in spiral arms
ASIAA Workshop on the Very Large Array 5
• Radio data from VLA + Effelsberg:
4.86 GHz, 15” resolution (700 pc)
• Contours show radio brightness,
dashed lines show magnetic field
orientation.
• Overlaid on optical image from
Hubble
• Clearly shows the magnetic fields
following the spiral pattern.
• From Fletcher et al. (2011)
Jansky VLA Example: Faraday Rotation for Cygnus A
– Color-coded map of the Rotation Measure – the slope of Faraday Rotation vs. l2.
– RM is proportional to the density-weighted longitudinal B-field, imbedded in the
cluster gas surrounding the radio source.
– For simple propagation:
• Dreher, Carilli
and Perley
(1987))
• RMs are very
high – source is
imbedded in a
dense, ionized,
magnetized
medium.
ASIAA Workshop on the Very Large Array 6
lB dnRM
there
here
e 8.0
Jansky VLA
Electron Scattering in S106
ASIAA Workshop on the Very Large Array 7
• A spectacular example of polarized
emission due to electron scattering.
• This is a 2.2m image of the bipolar
HII nebula S106
• Center peak is IR source IRS4.
• Contours show emission intensity
• Lines show the apparent E-field.
• Azimuthal orientation shows the
emission is scattered light from the
central source.
Aspin et al. (1990)
Jansky VLA
ASIAA Workshop on the Very Large Array 8
Example: Zeeman Splitting
Jansky VLA
What Is Polarization?
• EM Radiation is a transverse phenomenon – the E and B fields lie in a plane
perpendicular to the direction of propagation.
• The instantaneous E vector is comprised of two components which are
*completely independent*.
• The two components can be described in terms of orthogonal linear (Ex
and Ey), or opposite circular (Er and El).
• The polarization is defined by the locus of the E-vector as it passes
through a fixed plane, perpendicular to the propagation direction.
• The locus of the resultant E-vector describe an ellipse in this plane.
• For monochromatic radiation, three quantities are sufficient to fully
describe this ellipse:
– The amplitudes of the two components
– The phase difference between them.
ASIAA Workshop on the Very Large Array 9
Jansky VLA The Polarization Ellipse – monochromatic radiation
• We consider the time behavior of the E-field in a fixed perpendicular plane, with the z-axis directed along the direction of propagation towards the observer.
• For a monochromatic wave of frequency n, we can write
• These two equations describe an ellipse in the (X-Y) plane:
• The ellipse is described by three parameters:
– The amplitudes Ax, Ay, and
– The phase difference, d = dy – dx
ASIAA Workshop on the Very Large Array
10
Jansky VLA
Elliptically Polarized Monochromatic Wave
According to Maxwell’s
equations, an EM wave is
elliptically polarized.
In general, three parameters are
needed to describe the
ellipse.
• Ax – X-axis amplitude max
• Ay – Y-axis amplitude max
• a = atan(Ay/Ax) – an angle
describing the orientation
If the E vector is rotating:
• Clockwise, the wave is Left
Elliptically Polarized:
• Anti-clockwise, the wave is
Right Elliptically Polarized.
ASIAA Workshop on the Very Large Array 11
Jansky VLA
Example: Decomposition into Linear components
ASIAA Workshop on the Very Large Array 12
Jansky VLA
13 ASIAA Workshop on the Very Large Array
Form of Ellipse depends on two amplitudes, and
the phase difference:
Ex = Ey, d = 90 Ex = Ey, d = 0 in general:
Circular Linear Elliptical
Jansky VLA
And, in Motion:
• Showing the component (blue, red) and net (black) fields as the wave
passes a fixed (x,y) plane.
• Here, the amplitudes are equal, and the phase difference is 90 degrees.
• As seen from the right, the E-vector is rotating clockwise:
• The wave is 100% LCP.
14 ASIAA Workshop on the Very Large Array
Jansky VLA Alternate Description – The Circular Basis
• We can decompose the E-field into a circular basis, rather than a
cartesian one:
– where AR and AL are the amplitudes of two counter-rotating unit
vectors: eR (rotating counter-clockwise), and eL (rotating clockwise)
• The polarization ellipse is again described by three parameters:
ASIAA Workshop on the Very Large Array
15
Jansky VLA
Circular Basis Example • The polarization ellipse
(black) can be decomposed
into an X-component of
amplitude 2, and a Y-
component of amplitude 1
which lags by ¼ turn.
• It can alternatively be
decomposed into a
counterclockwise (RCP)
rotating vector of length 1.5
(red), and a clockwise
rotating (LCP) vector of
length 0.5 (blue).
ASIAA Workshop on the Very Large Array
Jansky VLA Decomposition into circular components
• The same ellipse, but now decomposed into oppositely-rotating circular
components.
17
ASIAA Workshop on the Very Large Array
Jansky VLA In a more Natural Reference Frame
• A more natural description is in a frame (x,h), rotated so the x-axis lies along the major axis of the ellipse.
• The three parameters of the ellipse are then:
– Ax : the major axis length
– Y: the position angle of this major axis, and
– tan c Ah/Ax : the axial ratio
• It can be shown that:
• The ellipticity c is signed:
c > 0 => LEP (clockwise)
c < 0 => REP (anti-clockwise)
ASIAA Workshop on the Very Large Array
18
Jansky VLA Stokes Parameters
• Three parameters are sufficient to describe the monochromatic EM polarization.
• It is most convenient to have the three parameters share the same units, and have easily grasped physical meanings.
• It is standard in radio astronomy to utilize the parameters defined by George Stokes (1852), and introduced to astronomy by Chandrasekhar (1946):
• Note that
• Thus – a monochromatic wave is 100% polarized.
• Note also that the Stokes parameters have units of brightness – very convenient for imaging instruments!
ASIAA Workshop on the Very Large Array
19
Jansky VLA
Interpreting the Meaning of I, V, Q, and U
• I is the total power.
• V is the difference between the
power in RCP and LCP components.
• Q is the difference between vertical
and horizontal power.
• U is the difference between
orthogonal components in a frame
rotated by 45 degrees.
• If V = 0, the wave is 100% Linearly Polarized
• If Q=U=0, the wave is 100% Circularly Polarized
• In general, the wave is Elliptically Polarized
ASIAA Workshop on the Very Large Array
20
Jansky VLA
Pure Linear Polarization: V = 0 • Presume a linearly polarized wave, of unit power.
• Then, if:
– The wave is vertically polarized:
• Q = 1 U = 0
– The wave is horizontally polarized:
• Q = -1 U = 0
– The wave is polarized at pa = 45 deg:
• Q = 0 U = 1
– The wave is polarized at pa = -45.
• Q = 0 U = -1
ASIAA Workshop on the Very Large Array 21
Jansky VLA Linear Polarization (Q, U)
• In general, we define:
– The Linearly Polarized Intensity:
– The orientation of the plane of polarization:
– Signs of Q and U tell us the orientation of the plane of polarization:
Q > 0
Q < 0 Q < 0
Q > 0
U > 0
U > 0
U < 0
U < 0
ASIAA Workshop on the Very Large Array
22
For this
case:
Q>0
U<0
Jansky VLA
In General: Some Illustrative Examples
Pol Ellipse I V Q U
1 1 0 0
1 -1 0 0
1 0 1 0
1 0 -1 0
1 0 0 -1
1 0 0 1
1 0
1 0
ASIAA Workshop on the Very Large Array
23
Jansky VLA Non-Monochromatic Radiation, and Partial Polarization
• Monochromatic radiation is nonphysical.
• No such condition can exist (although it can be closely
approximated).
• In real life, radiation has a finite bandwidth Dn – the signals are
quasi-sinusoidal, specified by an amplitude and phase for a
limited time t~1/ Dn.
• Because the amplitudes and phases of the received signals are
now variable, there are no unique, stationary values of
amplitude and phase.
• We must then use suitably defined averages of these
quantities to define the polarization characteristics.
ASIAA Workshop on the Very Large Array 24
Jansky VLA Stokes Parameters for wideband radiation
• In the quasi-monochromatic approximation, the incoming EM wave can be described, for a period Dt~1/Dn, by two amplitudes, Ap and Aq, and a phase difference, dpq.
• For the orthogonal linear, and opposite circular bases, we have:
22
22
2222
sin2
sin2cos2
cos2
LRXYYX
RLLRXYYX
RLLRYX
LRYX
AAAAV
AAAAU
AAAAQ
AAAAI
d
dd
d
• The angle brackets <> denote an average over a time much longer
than the coherence time 1/Dn.
• These four real numbers are a complete description of the polarization
state of the incoming radiation.
• They are a function of frequency, position, and time.
ASIAA Workshop on the Very Large Array
25
Jansky VLA
Imaging in All Four Stokes
• The four quantities defined on the last slide provide a
complete description of the polarization state of the
incoming radiation.
• To do astronomy, we want to get images of the sky – one
image for each Stokes quantity.
• How do we get these from a telescope? An
interferometer?
26
ASIAA Workshop on the Very Large Array
Jansky VLA
Antennas are Polarized! • Polarimetry is possible because antennas are polarized – their output
is not a function of I alone.
• It is highly desirable (but not required) that the two outputs be sensitive to two orthogonal modes (i.e. linear, or circular).
• In interferometry, we have two antennas, each with two differently polarized outputs.
• We can then form four complex correlations.
• What is the relation between these four correlations and the four Stokes’ parameters?
Polarizer RCP LCP
A generic antenna
ASIAA Workshop on the Very Large Array 27
Jansky VLA Four Complex Correlations
per Pair of Antennas
• Two antennas, each
with two polarized
outputs, produce four
complex correlations.
• What is the relation
between these four
correlations, and the
four Stokes
parameters?
L1 R1
X X X X
L2 R2
Antenna 1 Antenna 2
RR1R2 RR1L2 RL1R2 RL1L2
(feeds)
(polarizer)
(signal
transmission)
(complex
correlators)
ASIAA Workshop on the Very Large Array 28
Four complex cross-products
Jansky VLA
Stokes Visibilities
• But an interferometer doesn’t measure these values directly.
Recall that an interferometer measures the Visibility, which is
related to the image intensity by a Fourier Transform:
V (u,v) I (l,m) (a Fourier Transform Pair)
• Define the Stokes Visibilities I, Q, U, and V, to be the
Fourier Transforms of the Stokes’ brightesses: I, Q, U, and V:
• Then, the relations between these are:
• I I, Q Q, U U, V V
• Stokes Visibilities are complex functions of (u,v), while the
Stokes Images are real functions of (l,m).
• Our task is now to obtain these Stokes visibilities from the
cross-power measurements provided by an interferometer.
ASIAA Workshop on the Very Large Array 29
Jansky VLA Relating the Products to Stokes’ Visibilities
• Let ER1, EL1, ER2 and EL2 be the complex representation (phasors) of
the RCP and LCP components of the EM wave which arrives at the
two antennas.
• We can then utilize the definitions earlier given to show that the four
complex correlations between these fields are related to the desired
visibilities by (ignoring gain factors):
• So, if each antenna has two outputs whose voltages are faithful replicas
of the EM wave’s RCP and LCP components, then the simple
equations shown are sufficient.
2/)(
2/)(
2/)(
2/)(
*
2121
*
2121
*
2121
*
2121
iUQ
iUQ
VI
VI
RLRL
LRLR
LLLL
RRRR
EER
EER
EER
EER
ASIAA Workshop on the Very Large Array 30
Jansky VLA Solving for Stokes Visibilities
• The solutions are straighforward:
)(2121
2121
2121
2121
RLLR
RLLR
LLRR
LLRR
RRi
RR
RR
RR
U
Q
V
I
• For an unresolved, or partially resolved source, Q, U, and V are
much smaller than I (low polarization).
• Thus, the amplitudes of the cross-hand correlations are often much
less than the parallel hand correlations.
• V is formed from the difference of two large quantities, while Q and
U are formed from the sum and difference of small quantities.
• If calibration errors dominate (and they often do), the circular basis
favors measurements of linear polarization.
ASIAA Workshop on the Very Large Array 31
Jansky VLA For Linearly Polarized Antennas …
• We can go through the same exercise with perfectly linearly
polarized feeds and obtain (presuming they are oriented with the
vertical feed along a line of constant HA, and again ignoring
issues of gain):
• For each example, we have four measured quantities and four
unknowns.
• The solution for the Stokes visibilities is easy.
2/)(
2/)(
2/)(
2/)(
*
2121
*
2121
*
2121
*
2121
iVU
iVU
QI
QI
VHVH
HVHV
HHHH
VVVV
EER
EER
EER
EER
ASIAA Workshop on the Very Large Array
Jansky VLA Stokes’ Visibilities for Pure Linear Antennas
• Again, the solution in straightforward:
)(2121
2121
2121
2121
VHHV
VHHV
HHVV
HHVV
RRi
RR
RR
RR
V
U
Q
I
• Now, the Q parameter is the difference of large quantities –
trouble!
• But, V is now the difference of small quantities – that’s good.
• Bottom line here: No perfect system!
ASIAA Workshop on the Very Large Array 33
Jansky VLA
Summary: Polarimetry in Astronomy
• So, (in principle) the process is easy:
1. Collect all four cross-correlations from your interferometer.
2. Calibrate!
3. Generate the four Stokes visibilities from your four correlations.
4. Fourier transform all four.
5. Deconvolve all four
6. Combine to form polarization and position angle images
7. Think hard.
8. Publish that paper.
ASIAA Workshop on the Very Large Array 34
Jansky VLA
But The Real World is harder …
• The preceding analysis presumes:
1. Antennas fixed to the sky frame (no parallactic angle
rotation)
2. Perfectly polarized antennas
3. Perfectly calibrated data.
• Sadly, none of these things happens in the real world (although
the 1st can be arranged to happen by suitable design).
• Real antennas are imperfectly polarized – some RCP power
emerges from the LCP port, and vice versa.
• In this short introduction, I can give no detail.
• What follows is a summary.
ASIAA Workshop on the Very Large Array 35
Jansky VLA
Imperfectly Polarized Antennas
• Typically, for our wideband systems, there is about a 5%
‘leakage’ of opposite polarization into each output.
• The ‘leakage’ is characterized by a quantity labelled ‘D’, a
complex quantity which describes the amplitude and phase of
the leakage.
• We write: Vr,obs = Vr + DrVl, and a similar expression for Vl,obs
• Then, using the magic of Jones matrix algebra, and the
definitions give before, analysis shows that:
ASIAA Workshop on the Very Large Array 36
Jansky VLA Slightly Imperfect Circularly Polarized Antennas
Y
Y
2/)(
2/)(
2/)(
2/)(
1
1
1
1
2
2
*
21
*
21
*
2
*
211
1
*
21
*
2
*
211
*
2
21
21
21
21
VI
UQ
UQ
VI
ie
ie
DDDD
DDDD
DDDD
DDDD
R
R
R
R
P
P
i
i
RLRLLRRL
LRLRRLRL
LRRLLRRL
LRLRLRLR
LL
RL
LR
RR
L
R
i
LL
i
RR
e
e
D
D
2
2
tan
tan
Where:
• If |D|<<1, we can then ignore D*D products. (Typically, |D| ~ .05)
• Furthermore, as |Q| and |U| << |I|, we can ignore products between
them and the Ds. (This is dangerous for a well resolved objects …)
• And V can be safely assumed to be zero.
• These (generally reasonable) approximations then give us:
ASIAA Workshop on the Very Large Array 37
Jansky VLA ‘Nearly’ Circular Feeds (small D approximation)
• We get:
• Our problem is now clear. The desired cross-hand responses are contaminated by a term proportional to ‘I’.
• Stokes ‘I’ is typically 20 to 100 times the magnitude of ‘Q’ or ‘U’.
• If the ‘D’ terms are of order a few percent (and they are!), we must make allowance for the extra terms.
• To do accurate polarimetry, we must determine these D-terms, and remove their contribution.
• Knowing the D-terms, one can easily modify the Rs to their correct values.
ASIAA Workshop on the Very Large Array
38
Jansky VLA Measuring Cross-Polarization • Correction of the X-hand response for the ‘leakage’ is important, since the
leakage amplitude is comparable to the fractional polarization.
• There are two ways to proceed:
1. Observe a calibrator source of known polarization (preferably zero!)
2. Observe a calibrator of unknown polarization for a ‘long time’.
• First case (with polarization = 0).
2/
2/
2/
2/
*
2121
*
2121
21
21
VHVH
HVHV
HH
VV
DDR
DDR
R
R
I
I
I
I
• Then a single observation should suffice to measure the leakage terms.
• This is not actually correct – because the cross-hand visibility is always
the sum of two terms, the ‘D’ values must be referenced to an assumed
value (DV1 = 0, for example).
ASIAA Workshop on the Very Large Array 39
Jansky VLA Determining Source and Antenna Polarizations
• You can determine both the (relative) D terms and the calibrator
polarizations for an alt-az antenna by observing over a wide range of
parallactic angle, YP. (R. Conway and P. Kronberg invented this)
• As time passes, YP changes in a known way.
• The source polarization term then rotates w.r.t. the antenna
leakage term, allowing a separation.
ASIAA Workshop on the Very Large Array 40
Jansky VLA
Fortunately, It’s all ‘under the hood’.
• At this point, you could be forgiven if you have decided to take a vow to
never do interferometric polarimetry.
• Fortunately, the procedures to determine these antenna polarization
impurities (the ‘D’ terms) are quite straightforward.
• For alt-az systems like the VLA, you need only observe one (or more)
calibrators at least a half dozen times, spanning a parallactic angle change of
at least ~50 degrees.
• Normally, your regular phase calibrator will suffice.
• In addition, a single observation of an object of know position angle is
required to rotate the phase frame.
• Both AIPS and CASA have the necessary software built in.
• Once editing and parallel-hand calibration is done, the polarimetric
calibration is normally ‘a piece of cake’.
ASIAA Workshop on the Very Large Array 41
Jansky VLA
Illustrative Example – Emission from Mars.
• The planet Mars radiates as a blackbody, with a brightness temperature
near 200 K.
• Due to the change in refractive index between the Martian surface and the
atmosphere, the (unpolarized) emerging radiation becomes partially
polarized upon passing through the interface – increasingly so as the angle
increases.
• The observable effect is that emission away from the center of the planet
becomes increasingly linearly polarized, with the direction directed radially
away from the center.
• I show next some I, Q, U data of Mars, taken at 23 GHz.
ASIAA Workshop on the Very Large Array 42
Jansky VLA I,Q,U,V Visibilities
• It’s useful to look at the visibilities which made these images.
I Q
Amplitude
Phase
ASIAA Workshop on the Very Large Array 43
Jansky VLA Illustrative Example – Thermal Emission from Mars
• Mars emits in the radio as a black body.
• Shown are false-color coded I,Q,U,P images from Jan 2006 data at 23.4 GHz.
• V is not shown – all noise – no circular polarization.
• Resolution is 3.5”, Mars’ diameter is ~6”.
• From the Q and U images alone, we can deduce the polarization is radial, around
the limb.
• Position Angle image not usefully viewed in color.
I Q U U P
ASIAA Workshop on the Very Large Array 44
Jansky VLA Mars – A Traditional Representation
• Here, I, Q, and U are
combined to make a more
realizable map of the total
and linearly polarized
emission from Mars.
• The dashes show the
direction of the E-field.
• The dash length is
proportional to the
polarized intensity.
• One could add the V
components, to show little
ellipses to represent the
polarization at every point.
ASIAA Workshop on the Very Large Array 45
Jansky VLA
A Summary
• Polarimetry is a little complicated.
• But, the polarized state of the radiation gives valuable information into the
physics of the emission.
• Well designed systems are stable, and have low cross-polarization, making
correction relatively straightforward.
• Such systems easily allow estimation of polarization to an accuracy of 1
part in 10000.
• Beam-induced polarization can be corrected in software – development is
under way.
ASIAA Workshop on the Very Large Array 46
The National Radio Astronomy Observatory is a facility of the National Science Foundation
operated under cooperative agreement by Associated Universities, Inc.
www.nrao.edu • science.nrao.edu