Transcript X-ray Emission from Massive Stars

X-ray Emission from
Massive Stars
David Cohen
Dept. of Physics & Astronomy
Swarthmore College
X-ray Emission from
Massive Stars
David Cohen
Dept. of Physics & Astronomy
Swarthmore College
The work discussed here is with collaborators:
Stan Owocki and Rich Townsend (U. Delaware), Asif ud-Doula
(U. Delaware and Swarthmore), Maurice Leutenegger
(Columbia), & Marc Gagne (West Chester)
Students: Roban Kramer (’03), Kevin Grizzard (St. John’s
College, ’06), Casey Reed (’05), Stephanie Tonnesen (’03),
Steve St. Vincent (’07)
X-ray Emission from
Massive Stars
O and early B stars: M > 8Msun;
Teff > 20,000 K; term “massive
stars” used interchangeably with
“hot stars”
OUTLINE
1. Introduction
a. Solar x-ray emission…vs. massive
star x-ray emission
b. Massive stars and their winds
2. The wind-shock paradigm
3. Chandra spectroscopy of  Puppis and
 Orionis: wind shocks
4. Chandra spectroscopy of 1 Orionis C:
signatures of a magnetized wind
5. Conclusions
OUTLINE
1. Introduction
a. Solar x-ray emission…vs. massive
star x-ray emission
b. Massive stars and their winds
2. The wind-shock paradigm
3. Chandra spectroscopy of  Puppis and
 Orionis: wind shocks
4. Chandra spectroscopy of 1 Orionis C:
signatures of a magnetized wind
5. Conclusions
X-rays from the Sun
Remember - for thermal radiation - the frequency of light (the
energy of each photon) is proportional to the temperature of the
emitter:
Human body = 300 K  10 microns, or 100,000 Å (infrared)
Sun, light bulb filament = 6000 K  5000 Å (visible, yellow)
Hot star’s surface = 40,000 K  750 Å (far ultraviolet)
Really hot plasma = 5,000,000 K  6 Å (X-ray)
*don’t forget that thermal emitters give off photons with a range
of wavelengths; those listed above represent the peak of the
distribution or the characteristic wavelength.
Note: an Angstrom unit (Å) is equivalent to 0.1 nanometers (nm)
The Sun is a strong source of X-rays
(10-5 of the total energy it emits)
It must have ~million degree plasma on it
The hot plasma is generally confined in magnetic structures
above – but near - the surface of the Sun.
Visible solar spectrum:
continuum, from surface
X-ray/EUV solar spectrum:
emission lines from hot, thin
plasma above the surface
We can use spectroscopy - in our study of massive stars (where
spatial structure can’t be imaged) - to diagnose plasma kinematics
(via Doppler-broadened line shapes) and plasma location with
respect to the stellar surface (via UV-sensitive line ratios)
Theme: spectroscopy as a proxy for imaging.
X-ray/EUV solar spectrum:
emission lines from hot, thin
plasma above the surface
This hot plasma is related to magnetic fields on the
Sun: confinement, spatial structure, conduits of
energy flow, heating
More magnetic structures on the Sun:
x-ray image from TRACE
The Sun’s magnetic dynamo requires
rotation + convection to regenerate and
amplify the magnetic field
Sunspots over several days:
rotation
Note granulation, from
convection
TRACE composite
OK, so the Sun emits x-rays - quite
beautifully - and they’re associated with its
magnetic activity, related to convection and
rotation…
But what of hot, massive stars?
OUTLINE
1. Introduction
a. Solar x-ray emission…vs. massive
star x-ray emission
b. Massive stars and their winds
2. The wind-shock paradigm
3. Chandra spectroscopy of  Puppis and
 Orionis: wind shocks
4. Chandra spectroscopy of 1 Orionis C:
signatures of a magnetized wind
5. Conclusions
Hot, Massive Stars
Stars range in (surface) temperature from about
3500 K to 50,000 K
Their temperatures correlate with mass and
luminosity (massive stars are hot and very bright):
a 50,000 K star has a million times the luminosity of
the Sun (Tsun = 6000 K)
Stars hotter than about 8000 K do not have
convective outer layers - no convection - no dynamo
- no hot corona…
…no X-rays ?
Our Sun is a somewhat wimpy star…
 Puppis:
42,000 K vs. 6000 K
106 Lsun
50 Msun
Optical image of the constellation Orion
Note: many of the brightest stars are blue (i.e. hot, also massive)
In 1979 the Einstein Observatory made the surprising discovery
that many O stars (the hottest, most massive stars) are strong
X-ray sources
1 Ori C: a 45,000 K
O-type star
Chandra X-ray image of the
Orion star forming region
Note: X-rays don’t penetrate the Earth’s atmosphere, so X-ray telescopes must be in space
So, we’ve got a good scientific mystery: how
do massive stars make X-rays?
Could we have been wrong about the lack of a
magnetic dynamo - might massive star X-rays be
similar to solar X-rays?
Before we address this directly, we need to know
about one very important property of massive
stars (that might provide an alternate explanation
for their X-rays)…
Massive stars have very strong radiationdriven stellar winds
What is a stellar wind?
It is the steady loss of mass
from the surface of a star
into interstellar space
The Sun has a wind (the
“solar wind”) but the winds
of hot stars can be a billion
times as strong as the Sun’s
Hubble Space Telescope image
of h Car; an extreme example
of a hot-star wind
How do we know these hot-star winds exist?
Spectroscopy!
rest wavelength(s) – this N V line is a doublet
Absorption comes
exclusively from region F it’s all blue-sifted
You can read the terminal
blue wavelength
red
velocity (in km/s) right off the
blue edge of the absorption line
Why do hot star winds exist?
The solar wind is actually driven by the gas pressure
of the hot corona
But hot-star winds are driven by radiation pressure
Remember, photons have momentum as well as energy:
E h h
p 

c
c

And Newton tells us that a change in momentum is a force:

dv dp
F  ma  m 
dt dt
So, if matter (an atom) absorbs light (a
photon) momentum is transferred to the
matter
Light can force atoms to move!
The flux of light, F
(ergs s-1 cm-2)
dp Fs
Ls T


dt
c
4cR 2

arad
LT

4cR 2
re, the radius of an
electron, giving a cross
section, sT (cm2)
The rate at which momentum is
absorbed by the electron
By replacing the cross section of a single
electron with the opacity (cm2 g-1), the
combined cross section of a gram of plasma,
we get the acceleration due to radiation
For a (very luminous) hot star, this can compete
with gravity…but note the 1/R2 dependence, if arad
> agrav, a star would blow itself completely apart.
And free electron opacity, and the associated
Thompson scattering, can be significantly
augmented by absorption of photons in spectral
lines –
atoms act like a resonance chamber for
electrons: a bound electron can be ‘driven’ much
more efficiently by light than a free one can (i.e.
it has a much larger cross section), but it can
only be driven by light with a very specific
frequency.
Radiation driving in spectral lines not only boosts the radiation
force, it also solves the problem of the star potentially
blowing itself apart:
As the radiation-driven material starts to move off the surface
of the star, it is Doppler-shifted, making a previously narrow
line broader, and increasing its ability to absorb light.
cont.
0
Optically thick line – from stationary plasma (left);
moving plasma (right) broadens the line and
increases the overall opacity.
The Doppler desaturation of optically thick (opaque) lines allows a
hot-star wind to bootstrap itself into existence!
And causes the radiation force to deviate from strictly 1/R2
behavior: the radiation force on lines can be less than gravity
inside the star but more than gravity above the star’s surface.
OUTLINE
1. Introduction
a. Solar x-ray emission…vs. massive
star x-ray emission
b. Massive stars and their winds
2. The wind-shock paradigm
3. Chandra spectroscopy of  Puppis and
 Orionis: wind shocks
4. Chandra spectroscopy of 1 Orionis C:
signatures of a magnetized wind
5. Conclusions
X-rays from shock-heating in linedriven winds
The Doppler desaturation that’s so helpful
in driving a flow via momentum transfer in
spectral lines is inherently unstable
The line-driven instability (LDI) arises when a parcel of wind
material is accelerated above the local flow speed, which moves it
out of the “Doppler shadow” of the material behind it, exposing it
to more photospheric radiation, and accelerating it further…
Numerical modeling of the hydrodynamics show lots of structure:
turbulence, shock waves, collisions between “clouds”
This chaotic behavior is predicted to produce X-rays through
shock-heating of some small fraction of the wind.
A snapshot at a single time from the same simulation. Note the
discontinuities in velocity. These are shock fronts, compressing and
heating the wind, producing x-rays.
There are dense inter-shock regions, though, in which cold material
provides a source of photoelectric absorption
Even in these instability shock models, most
of the wind is cold and is a source of x-ray
continuum opacity - x-rays emitted by the
shock-heated gas can be absorbed by the
cold gas in the rest of the wind
Keep this in mind, because it will allow us to
learn things about the physical properties of
a shocked wind via spectroscopy
X-ray line profiles can provide the
most direct observational constraints
on the x-ray production mechanism in
hot stars
Wind-shocks : broad lines
Magnetic dynamo : narrow lines
The Doppler effect will make the x-ray emission
lines in the wind-shock scenario broad, compared to
the x-ray emission lines expected in the
coronal/dynamo (solar-like) scenario
OUTLINE
1. Introduction
a. Solar x-ray emission…vs. massive
star x-ray emission
b. Massive stars and their winds
2. The wind-shock paradigm
3. Chandra spectroscopy of  Puppis and
 Orionis: wind shocks
4. Chandra spectroscopy of 1 Orionis C:
signatures of a magnetized wind
5. Conclusions
So, this wind-shock model - based on the lineforce instability - is a plausible alternative to the
idea that hot star x-rays are produced by a
magnetic dynamo
This basic conflict is easily resolved if we can
measure the x-ray spectrum of a hot star at high
enough resolution…
In 1999 this became
possible with the
launch of the Chandra
X-ray Observatory
Now, for some data
Si XIV
 Pup
Mg XII
Ne X
(O4 I)
Ne IX Fe XVII
O VIII
10 Å
O VII
20 Å
NV
Focus in on a characteristic portion of the spectrum
12 Å
15 Å
 Pup
(O4 I)
Capella - a
cooler star:
coronal/dynamo
source
Ne X
Ne IX
Fe XVII
Differences in the line shapes become apparent
when we look at a single line (here Ne X Lya)
lab/rest wavelength
Pup
(O4 I)
Capella
(G2 III)
The x-ray emission
lines in the hot star
 Pup are broad -the wind shock
scenario is looking
good!
But note, the line
isn’t just broad, it’s
also blueshifted and
asymmetric…
We can go beyond simply wind-shock vs.
coronal:
We can use the line profile shapes to
learn about the velocity distribution of the
shock-heated gas and even its spatial
distribution within the wind, as well as
learning something about the amount of
cold wind absorption (and thus the
amount of cold wind).
What Line Profiles Can Tell Us
The wavelength of an emitted photon is proportional to the lineof-sight velocity:
Line shape maps emission at each velocity/wavelength
interval
Continuum absorption by the cold stellar wind affects the line
shape
Correlation between line-of-sight velocity and absorption
optical depth will cause asymmetries in emission lines
The shapes of lines, if they’re broad, tells us about
the distribution and velocity of the hot plasma in
the wind -- maybe discriminate among specific
wind shock models/mechanisms
We will now build up a physical – but flexible
– empirical x-ray emission line profile model:
Accounting for the kinematics of the emitting
plasma (and the associated Doppler
shifting/broadening);
Radiation transport (attenuation of the line
photons via bound-free absorption in the cold
wind component).
Note that our line-profile model, while
physical, is agnostic regarding the heating
mechanism.
Emission Profiles from a Spherically
Symmetric, Expanding Medium
A uniform shell
gives a
rectangular
profile.
A spherically-symmetric, x-ray
emitting wind can be built up
from a series of concentric shells.
Occultation by the
star removes red
photons, making
the profile
asymmetric
Continuum Absorption Acts Like Occultation
Contours of constant optical depth
(observer is on the left)
blue
red
wavelength
Red photons are preferentially absorbed, making
the line asymmetric: The peak is shifted to the
blue, and the red wing becomes much less steep.
The model has four parameters:
Ro=1.5
 : v(r)  (1 R /r) 
Ro,q : j   2 rq
for r>Ro

dz'
  :  ( p  0;z)     z

where   
1 
r' (1 )
r'
Ro=3
2
M
4 Rv
The line profile is calculated from:
L  8

2
 
1

1
R
je r 2 drd
Increasing Ro makes lines
broader; increasing *
makes them more
blueshifted and skewed.
Ro=10
=1,2,4
A wide variety of windshock properties can be
modeled
Line profiles
change in
characteristic ways
with * and Ro,
becoming broader
and more skewed
with increasing *
and broader and
more flat-topped
with increasing Ro.
=1,2,4
Ro=1.5
Ro=3
Ro=10
In addition to
the wind-shock
model,
our empirical line-profile model can also
describe a corona
With most of the
emission
concentrated near
the photosphere
and with very little
acceleration, the
resulting line
profiles are very
narrow.
We fit all the unblended strong lines in the
Chandra spectrum of  Pup: all the fits are
statistically good
Ne X
12.13 Å
Fe XVII
17.05 Å
Fe XVII
15.01 Å
O VIII
18.97 Å
Fe XVII
16.78 Å
N VII
24.78 Å
We place uncertainties on the derived model
parameters
lowest *
best *
highest *
Here we show the best-fit model to the O VIII line and two models
that are marginally (at the 95% limit) consistent with the data; they
are the models with the highest and lowest * values possible.
Lines are well fit by our three parameter model: 
Pup’s x-ray lines are consistent with a spatially
distributed, spherically symmetric, radially
accelerating wind scenario, with reasonable
parameters:
*~1
:4 to 15 times less than predicted
Ro~1.5
q~0
But, the level of wind absorption is significantly
below what’s expected.
And, there’s no significant wavelength dependence
of the optical depth (or any parameters).
The results for  Pup were published
several years ago, with Roban Kramer
(Swarthmore 2003) as the lead author.
However, it’s generally been considered
that other massive stars’ x-ray spectra
were not consistent with the wind-shock
scenario.
Much of the work shown on the next few slides – on
 Ori – was done by Kevin Grizzard (St. John’s
College 2006)
Here’s another way of
looking at the situation:
There are claims in the
literature that the
emission lines of most
massive stars can be fit
by Gaussian profiles.
We fit strong lines in the
Chandra spectra of  Ori
with unshifted Gaussians
(top), shifted Gaussians
(center), and the wind
profile model (bottom).
Rejection probabilities
are shown on the right
of each panel.
94%
73%
54%
*
Fit results for  Ori summarized (to appear in
Monthly Notices of the R.A.S., 2006)
The wind profile model provides statistically good fits to all the
lines. The onset radii (left) are exactly what’s expected from
the standard wind-shock picture. There is evidence for
attenuation by the cold wind (right), but at levels nearly 10
times lower than expected. This is the same result that we
found for  Pup.
Ro of several tenths to one stellar radius is expected based on numerical
simulations of the line-force instability (self-excited on the left; sound
wave perturbations at the base of the wind on the right)
This is consistent with the results of the He-like f/i ratio analysis
So…what’s going on with the much lower
wind optical depths?
The atomic cross sections are quite well
known.
Could the mass-loss rates of massive stars be
overestimated? By an order of magnitude?
There would be very serious evolutionary
implications (for, e.g., supernovae and
chemical enrichment of galaxies).
There are, in fact, other recent papers that
show several independent lines of evidence
that wind mass-loss rates are lower than
previously thought.
Some of these rely on the insight that
clumping will cause density-squared
diagnostics to overestimate mass-loss rates.
Density-squared processes – H-alpha emission
(driven by recombination) and radio free-free
emission – are commonly used to determine
wind mass-loss rates.
Bouret, Lanz, & Hillier (2005): detailed fits to UV spectra;
Puls et al. (2006): H-alpha and radio free-free excess;
Fullerton, Massa, & Prinja (2006): FUSE and Copernicus P V
absorption line fitting
The effect of clumping on density-squared emission
emissivity, j = n2 X Vol
12 X 1=1
12 X 1=1
42 X 1=16
02 X 1=0
12 X 1=1
12 X 1=1
02 X 1=0
02 X 1=0
j=4
j = 16
Clumping’s effect on density-squared emission
is scale-free (only the density contrast
between clumps and the inter-clump medium
matters).
However, we have begun to investigate a
separate effect – porosity – the ability of
photons to more easily escape through lowdensity inter-clump channels.
Winds with porosity length increasing to the right
We have discovered that the key parameter for
describing the reduction in effective opacity
due to porosity is the ratio of the clump size
scale to the volume filling factor.
We dub this quantity the porosity length, h.
It turns out that line
profiles (left) are not
significantly affected until
the porosity length is
comparable to the stellar
radius (unity, in the unitless
formulation of these slides).
This degree of porosity is
not expected from the
line-driven instability. The
clumping in 2-D
simulations (right) is on
quite small scales.
Note: these clumps are spherical.
The line-driven
instability might be
expected to
compress clumps
in the radial
direction:
pancakes, oriented
parallel to the
star’s surface.
We’ve started working on models with
non-isotropic/oblate clumps: the
Venetian-blind model.
There’s one more powerful x-ray spectral
diagnostic that can provide useful
information to test the wind-shock
scenario:
Certain x-ray line ratios provide
information about the location of the x-ray
emitting plasma
Distance from the star via the line ratio’s
sensitivity to the local UV radiation field
Helium-like ions (e.g. O+6, Ne+8, Mg+10, Si+12,
S+14) – schematic energy level diagram
1s2p 1P
10-20 eV
1s2p 3P
1s2s 3S
resonance (r)
1-2 keV
forbidden (f)
intercombination (i)
g.s. 1s2 1S
The upper level of the forbidden line is very long lived
– metastable (the transition is dipole-forbidden)
1s2p 1P
10-20 eV
1s2p 3P
1s2s 3S
resonance (r)
1-2 keV
forbidden (f)
intercombination (i)
g.s. 1s2 1S
While an electron is sitting in the metastable 3S level, an
ultraviolet photon from the star’s photosphere can excite it to
the 3P level – this decreases the intensity of the forbidden
line and increases the intensity of the intercombination line.
UV
1s2p 1P
1s2p 3P
1s2s 3S
resonance (r)
forbidden (f)
intercombination (i)
g.s. 1s2s 1S
The f/i ratio is thus a diagnostic of the strength of the local
UV radiation field.
UV
1s2p 1P
1s2p 3P
1s2s 3S
resonance (r)
forbidden (f)
intercombination (i)
g.s. 1s2s 1S
If you know the UV intensity emitted from the star’s surface,
it thus becomes a diagnostic of the distance that the x-ray
emitting plasma is from the star’s surface.
UV
1s2p 1P
1s2p 3P
1s2s 3S
resonance (r)
forbidden (f)
intercombination (i)
g.s. 1s2s 1S
Si XIII line complex in the Chandra spectrum of a massive star
where the local UV mean intensity is not strong enough to affect
the forbidden-to-intercombination ratio.
r
i
f
Si XIII line complex in the Chandra spectrum of a massive star
where the local UV mean intensity is strong enough to affect the
forbidden-to-intercombination ratio.
r
i
f
Here the f/i ratio
is reduced, due
the effects of UV
photoexcitation…
this occurs
because the x-ray
emitting plasma is
relatively close to
the photosphere.
We have fit line profile models simultaneously to
the f-i-r complexes in four hot stars – and get
consistent fits:
Hot plasma smoothly distributed throughout the wind, above
roughly 1.5 Rstar –
•The f/i line ratios are consistent with this spatial distribution
•The line profile shapes are also consistent with this distribution
(as already was shown for single, unblended lines)
Conclusions for most massive
stars: normal O-type supergiants
Spherically symmetric, standard wind-shock
scenario describes the Chandra data for 
Pup and  Ori (and, it looks like, most other
massive stars too) – x-ray line profiles and
line ratios are consistent with the expected
distribution of hot plasma
There’s evidence for attenuation by the cold
wind component, but the level of continuum
absorption in the wind must be reduced
from expected values by factors of ~10
(mass-loss rate reduction? some porosity?)