A&A 481, 117-122 (2008)
DOI: 10.1051/0004-6361:20078906
K. Heng1,2 - R. A. Sunyaev1,3
1 - Max Planck Institut für Astrophysik,
Karl-Schwarzschild-Strae 1, 85740 Garching, Germany
2 -
Max Planck Institut für extraterrestrische Physik,
Giessenbachstrae, 85478 Garching, Germany
3 -
Space Research Institute, Russian Academy of Sciences,
Profsoyuznaya 84/32, 117997 Moscow, Russia
Received 23 October 2007 / Accepted 18 December 2007
Abstract
Context. Charge transfer (or exchange) reactions between hydrogen atoms and protons in collisionless shocks of supernova remnants (SNRs) are a natural way of producing broad Balmer, Lyman, and other lines of hydrogen.
Aims. We wish to quantify the importance of shock-induced, non-thermal hydrogen emission from SNRs in young galaxies.
Methods. We present a method estimating the luminosity of broad (1000 km s-1) Ly
,
Ly
,
Ly
,
H
and P
lines, as well as the broad and narrow luminosities of the two-photon (2
)
continuum, from existing measurements of the H
flux. We consider cases of
and 1, where
is the ratio of electron-to-proton temperatures. We examine a modest sample of 8 proximate, Balmer-dominated SNRs from our Galaxy and the Large Magellanic Cloud. The expected broad Ly
luminosity per object is at most
1036 erg s-1. The 2
continuum luminosities are comparable to the broad H
and Ly
ones. We restrict our analysis to homogenous and static media.
Results. Differences in the Ly/H
and Ly
/H
luminosity ratios between the
and 1 cases are factors
2 for shock velocities
km s-1, thereby providing a direct and unique way to measure
.
In principle, broad, ``non-radiative'' Ly
from SNRs in young galaxies can be directly observed in the optical range of wavelengths. However, by taking the different rates between core collapse and thermonuclear supernovae into consideration, as well as the duration we expect to observe such Ly
emission from SNRs, we expect their contribution to the total Ly
luminosity from
to 5 galaxies to be negligibly small (
0.001%), compared to the radiative shock mechanism described by Shull & Silk (1979). Although broad, non-thermal Ly
emission has never been observed, these photons are produced in SNRs. Hence, the non-radiative Ly
luminosity is a part of the intrinsic Ly
spectrum of young galaxies.
Key words: ISM: supernova remnants - atomic processes - radiation mechanisms: general - galaxies: general
Observations of galaxies at high redshifts have revealed a broad class
of Ly-emitting galaxies at
to 5 (e.g., Tapken et al. 2007). The Ly
emission from these objects is reaching us
as light in the visible spectral band, enabling their study using
large, ground-based optical telescopes, which in turn permits detailed spectroscopic studies of these galaxies. Observations of quasars at
(e.g., Fan et al. 2006) have revealed heavy elemental abundances exceeding solar values. We know that at least some of the galaxies at
to 5 have high abundances of heavy elements, facilitating the formation of dust. In homogenous and static media, the dust particles impede the escape of Ly
emission from gas-rich galaxies, because of the small mean free paths of the photons, low temperatures of the gas, and ultimately high probabilities of absorption. In clumpy media, dust can enhance the escape of Ly
photons relative to the continuum (Neufeld 1991; Hansen & Oh 2006). Broadening of Ly
lines due to multiple scatterings is a slow process requiring a long diffusion time (though velocity fields in the interstellar medium may broaden the Ly
lines and reduce the diffusion time). Hence, there is special interest in the physical processes that are able to naturally produce extremely broad wings in Ly
lines, which may permit the photons to leave the host galaxy without requiring many scatterings (but see Sect. 5).
Among obvious mechanisms is the one at work in the unique massive
binary SS433 (for a recent review, see Fabrika 2004), with strongly blue- and redshifted
H
and H
lines result from the cooling and recombination of
hydrogen in the baryon-dominated, precessing jet moving with velocity
0.26c. Such objects are very rare - SS433 is the only such
example in our Galaxy. More well-known Galactic sources of H
emission with broad line wings are the supernova remnants (SNRs) of
type Ia, emitting because of charge transfer (or ``charge exchange'')
reactions between hydrogen atoms and protons in the blast wave
penetrating the low-density (
1 cm-3), ambient gas. The
widths of the H
lines correspond to Doppler broadening with
velocities up to
5000 km s-1. The same process should
produce not only H
emission, but also photons in the Lyman series
of hydrogen. Recently, some of these SNRs have been observed in Ly
using the FUSE spacecraft (Korreck et al. 2004; Ghavamian et al. 2007b, hereafter G07).
Knowledge of the cross sections of charge transfers to excited levels
and excitation of the fast-moving hydrogen atoms permit us to find
simple formulae relating the luminosities of SNRs in the broad
H
and Ly
lines. The Ly
line should have a similar spectral distribution to the observed H
one in the broad wings, because the optical depth of the SNR for broad photons is negligibly small and the optical depth for coherent scattering (in the distant Lorentzian wings) in interstellar gas is low.
Table 1:
H
& Ly
observations of SNRs.
We compiled the existing data for core collapse and thermonuclear SNRs,
including SNR 1987A (where the reverse shock is bright in the broad
H
line), and present their theoretically expected broad
Ly
and Ly
luminosities. For two objects, we present
their expected broad Ly
,
H
,
and P
luminosities.
Taking into account the supernova (SN) rates, the luminosities of the
SNRs in H
and the duration of their active phase (for the
charge transfer mechanism described), we find that - even without
discussing the cosmological evolution of the SN rates - the expected
broad Ly
is several orders of magnitude lower than the
estimate of Shull & Silk (1979), who treated fully radiative SNRs with low
metallicities and velocities (20 to 120 km s-1). We come to the conclusion that the contribution of both core collapse and thermonuclear SNRs to the Ly
luminosity of young galaxies is negligibly small. In Sect. 2, we gather a modest sample of 8 Galactic and Large Magellanic Cloud (LMC) remnants and use them as a template for estimating the expected Ly
,
Ly
,
Ly
,
H
,
and P
production. In Sect. 3, we compute the Ly
/H
,
Ly
/Ly
,
Ly
/H
,
Ly
/H
,
H
/H
,
and P
/H
luminosity ratios. We present our results in Sect. 4 and discuss their implications in Sect. 5.
SNRs are the result of the interaction of SN ejecta with ambient matter. The nature of the interaction can be approximately categorized into several stages (Truelove & McKee 1999, hereafter TM99, and references therein): the ejecta-dominated (ED) or freely-streaming stage, the Sedov-Taylor (ST) or self-similar stage, the pressure-driven snowplow (PDS) stage, and a possible momentum-conserving snowplow stage (Cioffi et al. 1988). Many of the well-studied, young SNRs like Kepler, Tycho, and SN 1006 are intermediate between the ED and ST stages. This has been corroborated by the numerical studies of TM99, who show that there is no sharp transition between the two stages. The transition from the ED to ST stage occurs on a timescale
,
with the characteristic timescale being
![]() |
(2) |
In the ED and ST stages, the emission from some SNRs is
``non-radiative'', meaning the timescale for thermal, radiative losses from the
interacting gases is much longer than
.
When the blast
wave of the SNR slams into ambient gas consisting predominantly of
hydrogen atoms, it emits in Balmer and Lyman lines consisting of a
broad (
1000 km s-1) and a narrow (
10 km s-1)
component (Chevalier & Raymond 1978; Bychkov & Lebedev 1979; Chevalier et al. 1980; Heng
& McCray 2007, hereafter HM07; Heng et al. 2007, hereafter H07; G07; and references therein). These
objects are known as ``Balmer-dominated'' SNRs. Positive detections
of the line components are so far only from Galactic and LMC SNRs.
Even though narrow Ly
emission is produced, it is not seen due
to interstellar absorption, and broad Ly
should be observed
. Non-thermal H
and Ly
emission has not been observed in studies of local starburst galaxies (e.g., Kunth et al. 2003).
The narrow Balmer and Lyman lines are produced when the fast-moving
ejecta directly excite stationary hydrogen atoms in the
surrounding material. The broad lines are produced when the
post-shock protons and atoms engage in charge transfer reactions,
creating a population of post-shock atoms in broad velocity
distributions known as ``broad neutrals'' (HM07; H07). In the frame
of the observer, these broad neutrals move at a velocity
,
where
is the shock velocity (of the blast
wave). For
km s-1, the broad neutrals can
produce Ly
that is blue or redshifted out of resonance with
the stationary atoms, thereby providing an escape route for the photons.
The ratio of broad-to-narrow H
(and Ly
)
emission is a
function of the shock velocity (HM07; H07); it also depends on factors
like the pre-shock neutral density and the degree to which the
temperatures of the electrons and ions are equilibrated. The
contribution from the broad H
line dominates when the shock
velocity is
3000 km s-1 and when the narrow
H
line assumes Case A conditions (HM07). Existing
observations of H
and Ly
emission from 8 Balmer-dominated SNRs are catalogued in Table 1. At
least 5 of these SNRs are believed to have resulted from type Ia
explosions. Only SNR 1987A has a clear core collapse origin, and it is the youngest SNR in the sample.
To convert H
line fluxes to broad Ly
luminosities, we use
![]() |
Figure 1:
Luminosity ratios of Ly![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
The ratio of Ly
to H
luminosities (Fig. 1) is computed using the methods developed by HM07:
For Ly,
we consider charge transfers (with protons) and
excitations (by electrons and protons) to the sub-levels 2p, 3s, and
3d. For H
,
we consider the same processes, but for the
sub-levels 3s, 3p, and 3d. Hence, we compute
for
and
,
where the factor of
is the fraction of radiative decays from 3p that result in H
,
with
the remainder going to Ly
.
For
,
we
consider instead
.
Cascade
contributions from higher levels are
effects. For
example, contributions to H
from n=4 are at most
;
other contributions from
4p, 4d, and 4f are at the
level.
![]() |
Figure 2:
Luminosity ratios of Ly![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
One can calculate the luminosity ratios for Ly/H
,
H
/H
,
and P
/H
as well. However, the
cross sections for impact excitation of hydrogen atoms by protons
to the sub-levels 4s, 4p, 4d, and 4f are unavailable at the time of
writing
. The cross sections for charge transfers to these excited
states, however, are available. At
km s-1,
,
and we may obtain luminosity ratios for
Ly
/H
,
H
/H
,
and P
/H
to
within a factor of 2 (Fig. 2). A list of the
relevant radiative decay fractions,
,
is given in Table 3 (see Appendix A for details). In principle, if the charge transfer and excitation cross sections are known to higher levels in the relevant velocity range, one can calculate the luminosity ratios for other lines in the Balmer, Lyman, Paschen, and other series of hydrogen.
We use the atomic cross sections of Balança, Lin & Feautrier
(1998), Barnett et al. (1990), Belkíc et al. (1992),
Harel et al. (1998), & Janev & Smith (1993), as well as
those found in the NIST Electron-Impact Cross-Section Database.
Details concerning the cross sections are given in Appendix B, where we provide fitting functions to them. We consider a pure hydrogen gas and include charge transfer, excitation, and ionization events between hydrogen atoms, electrons, and protons. We employ the thin-shock approximation, such that the relative velocity between atoms and ions is
,
which has been shown by H07 to be an excellent approximation. At the shock velocities considered,
km s-1, the significance of impact excitation by electrons is comparable to that by protons and cannot be neglected. We do not consider broad emission from within the shock front (see Appendix C).
The luminosity ratios
,
,
and
are shown in
Fig. 1. In the shock velocity range
km s-1, the differences in
and
between the
and 1
cases are factors
2 and are due to the sensitivity to
temperature of impact
excitation and ionization of hydrogen atoms by electrons. This may present a direct and unique opportunity to measure
.
We emphasize that our calculations are only valid for the broad lines; the narrow lines have optical
depths
and Lyman line trapping is a non-negligible
effect (Ghavamian et al. 2001, 2002). For example, narrow Ly
photons may be converted into narrow H
photons and a two-photon
(2
)
continuum. In addition, narrow Ly
cannot
propagate easily through the interstellar gas.
![]() |
Figure 3:
Left: Expected Ly![]() ![]() ![]() ![]() |
Open with DEXTER |
We used the data in Table 1 to compute the expected
luminosity of Ly,
(Fig. 3).
In estimating a range for
,
we only considered the
observational error bars in
(if available) and
allowed for a generous range in temperature equilibration between
electrons and protons,
,
where
.
Hence, the displayed error bars for
are not formal ones. We are aware of the recent
work by Ghavamian et al. (2007b), who show that there is
an empirical correlation between
and
- namely,
for
km s-1 and
for
km s-1. For the LMC remnants
detected in Ly
by G07, we computed the range in
by considering both the H
and Ly
fluxes. We note that the computed
erg s-1value for broad Ly
in SNR 1987A is comparable to the
erg s-1 figure predicted by Michael et al. (2003). Note
that the condition
is not true
in general. This is because the cross section for charge transfers to
the level 3p falls below that of 3s at a relative velocity
2000 km s-1 (Fig. B.2).
Table 2:
Predicted broad Ly,
Ly
,
Ly
,
H
,
P
,
and broad/narrow two-photon luminosities (erg s-1).
In Table 2, we make some predictions for the Ly,
Ly
,
H
,
and P
luminosities. It is puzzling that the theoretically expected Ly
luminosities are about 10 to 20 times higher than those inferred from the observations of G07. In other words, the observed H
fluxes in the LMC SNRs are comparable to the observed Ly
ones. We are not certain why this is the case, but note that Ly
is more susceptible to absorption by interstellar dust than H
and suspect this effect plays at least some part in the discrepancy. Moreover, the H
and Ly
observations were taken at different epochs (Tuohy et al. 1982 versus Ghavamian et al. 2007b). As described in Sect. 3, we are only able to provide rough predictions for Ly
,
H
,
and P
,
and only in the cases of 0509-67.5 and SNR 1987A, as these SNRs have shock velocities
5000 km s-1.
![]() |
Figure 4:
Luminosity ratios of the 2![]() ![]() ![]() |
Open with DEXTER |
We can make some estimates for the expected 2
continuum as
well, which is produced in the
transition. In the
case of an optically-thin plasma, the
transition
is negligible as collisions are unimportant. In Table 2, we make conservative predictions for the
2
continuum luminosity from both broad and narrow atoms, but consider only charge transfers
and excitations to the 2s level. Additional contributions from
n=3 range from
(Case A) to
(Case B); those from n=4 are
.
We only wish to make the point that
is comparable to
and
,
so the 2
transitions are a potentially observable source of continuum. In the
case of Galactic and LMC SNRs, the Galaxy Evolution Explorer (GALEX)
should be able to measure the low-frequency wing of 2
decay, using its 135-175 and 175-280 nm channels. By comparing
H
and 2
emission, it will be possible to estimate emission directly from the SNR shock due to broad Ly
(and the
contribution of narrow Ly
that cannot reach us). This is an
additional, unique source of information on the detailed physical
processes in shocks.
Several sources of uncertainty can affect the predicted luminosities. These include uncertainties in the age of the
SNR,
,
the distance to it, d, the measured non-radiative component of the H
flux, the temperature
equilibration between electrons and ions, and the atomic cross
sections used. Uncertainties in the cross sections are typically
10%. For charge transfer to excited states, the uncertainty
can be as much as 30% (Janev 2007, private communication). The
predicted luminosities have not been corrected for reddening by dust.
SNR 1987A is a unique example of a Balmer-dominated SNR. By virtue of
adiabatic expansion cooling, the SN ejecta comprises mostly neutral
hydrogen and rushes out at velocities 12 000 km s-1(Michael et al. 2003; Heng et al. 2006). The non-radiative H
and Ly
result from the interaction of the ejecta with the reverse shock and not the blast wave (Heng 2007). As SNR 1987A has a
type II origin, it is possible to produce Balmer and Lyman lines
via this mechanism, which is obviously not possible with type Ia's.
Smith et al. (2005) predict that the H
and Ly
emission from the reverse shock of SNR 1987A is shortlived (
2012 to 2014) and will be extinguished by the increasing flux of
extreme ultraviolet (EUV) and X-ray photons traveling into the
pre-shock region and ionizing the atoms - pre-ionization. This is marginal evidence that broad Ly
from SNRs of a core collapse origin will be shortlived, i.e.,
100 years. In general, for this scenario to work, some interaction of the blast wave with the ambient material is needed, but if it is too strong the pre-shock gas becomes ionized (Chevalier
2007, private communication).
To further investigate the viability of the short-lived, non-radiative Ly
hypothesis, we examined the sample of optically identified SNRs by Matonick & Fesen (1997), who studied an ensemble of SNR samples from 12 different galaxies, including the Small Magellanic Cloud (SMC), LMC, M31, and M33, with distances up to 7 Mpc. In galaxies like NGC 2403, M81 and M101, the SNRs are associated with starforming regions and most of them probably have a Type Ib/c origin. In most cases, the measured H
flux is
erg cm-2 s-1 and the inferred luminosity is
erg s-1. Since Matonick & Fesen (1997) did not provide H
line profiles, it is impossible to estimate the proportion of the H
emission that is non-radiative. Furthermore, their selection criterion is based on picking out objects with [S II]/H
,
which will not detect SNRs with predominantly non-radiative H
emission.
Shull & Silk (1979) computed the temporally-averaged Ly
luminosity from radiative shocks of a population of type II SNRs, assuming low metallicities, to be
![]() |
(5) |
Table 3: Radiative decay fractions.
A very conservative upper limit on the broad Ly
from the
Matonick & Fesen (1997) samples can be obtained if one generously
allows for all of the H
to be broad, for the shock velocities
to be low (
500 km s-1) such that
,
and for the
non-radiative emission to last
years. Even in this very
unlikely scenario,
erg s-1 is
only about
.
Hence, our charge transfer mechanism
is not energetically competitive. There is the possibility a SNR can produce both radiative and
non-radiative components of H
.
Well-known examples are Kepler
(Fesen et al. 1989; Blair et al. 1991) and RCW 86 (Long &
Blair 1990; Smith 1997). There is also the possibility that the
non-radiative emission from the SNR is inhibited. For example, Foster (2005) observed and studied the Galactic SNR 3C 434.1 (
yr;
kpc; possible Type Ib/c), which formed inside the eastern portion of a pre-existing stellar-wind bubble of interior density
0.1 cm-3. Strong H
emission (
erg s-1) is measured from the eastern side that is believed to be from a radiative shock. Being farther away from the western wall of the bubble, the shock on the western side is essentially still in free expansion and produces no measurable, non-radiative H
.
Our SNR sample and the considerations of SNR 1987A lead us to believe that if the short-lived emission contribution from type Ib/c and type II SNRs in young galaxies exists, it has a luminosity of
![]() |
(6) |
![]() |
(7) |
We have restricted our analysis to homogeneous and static media. Though broad, non-thermal Ly
emission has never been observed, these photons are produced in SNRs, so the non-radiative Ly
luminosity is a part of the intrinsic Ly
spectrum of young galaxies. The optical depth for a broad photon in the line wings is (Verhamme et al. 2006)
![]() |
(8) |
Acknowledgements
K.H. is grateful to: Ratko Janev, C. D. Lin, and Fernando Martín for invaluable advice regarding atomic cross sections; Dick McCray, Roger Chevalier, Rob Fesen, Bob Kirshner, and Bryan Gaensler for engaging discussions; Christian Balança for providing atomic cross sections in an electronic form; John Raymond and Mike Shull for helpful suggestions following their careful reading of the manuscript. He acknowledges the Max Planck Institutes for Astrophysics (MPA) and Extraterrestrial Physics (MPE) for their generous support and kind hospitality during the months of June to October 2007, where he was a visiting postdoctoral scientist. He is indebted to the tranquil Bavarian countryside for necessary moments of academic solitude, and to his wife, Stefanie, for her steadfast support.
To compute the rate coefficients for Ly,
Ly
,
Ly
,
H
,
H
and P
,
one needs to calculate the ratio of Einstein A-coefficients. The Einstein A-coefficient for hydrogen,
,
is the radiative decay rate (s-1) from the levels nl to
.
The radiative decay fraction is
![]() |
(A.1) |
The Einstein A-coefficients are proportional to the square of the
magnitude of the radial integrals,
.
(See Appendix A2 of HM07 for details on how to calculate them
analytically.) As a check, we have compared our computed values of
to the ones tabulated by Green et al. (1957), and find they agree.
Cross sections for interactions between protons and hydrogen atoms
(charge transfer and excitation) to the sub-levels 3s, 3p and 3d are
computed in Balança et al. (1998) and kindly provided to us by
C. Balança (2007, priv. comm.). We note that these calculations utilize a two-center atomic-orbital (TCAO) close-coupling method with an asymmetric (TCAO-A) basis set of 26 states, so as to avoid the spurious numerical oscillations
caused by using a traditional, symmetric set (TCAO-S). We fit these cross
sections, as well as those from Belkíc et al. (1992),
Harel et al. (1998) and the NIST Electron-Impact Cross
Section Database using the function:
![]() |
(B.1) |
![]() |
(B.2) | ||
![]() |
(B.3) | ||
![]() |
(B.4) | ||
![]() |
(B.5) | ||
![]() |
(B.6) | ||
![]() |
(B.7) | ||
![]() |
(B.8) | ||
![]() |
(B.9) |
![]() |
(B.11) |
![]() |
Figure B.1: Cross sections for charge transfers between hydrogen atoms and protons, to the sub-levels 2s and 2p, taken from Belkíc et al. (1992) and Harel et al. (1998). Also shown are the fitting functions of Barnett et al. (1990), denoted by ``B90''. |
Open with DEXTER |
![]() |
Figure B.2: Cross sections for charge transfers between hydrogen atoms and protons, to the sub-levels 3s, 3p, and 3d, taken from Belkíc et al. (1992) and Harel et al. (1998). Shown for comparison are the calculations of Balança et al. (1998), denoted by ``BLF98''. |
Open with DEXTER |
![]() |
Figure B.3: Cross sections for charge transfers between hydrogen atoms and protons, to the sub-levels 4s, 4p, 4d, and 4f, taken from Belkíc et al. (1992) and Harel et al. (1998). |
Open with DEXTER |
![]() |
Figure B.4: Cross sections for impact excitation of hydrogen atoms by protons, from the two-center atomic-orbital (TCAO) close-coupling calculations of Balança et al. (1998). Shown are the fits to the TCAO-A calculations. |
Open with DEXTER |
![]() |
Figure B.5: Cross sections for impact excitation of hydrogen atoms by electrons, to the sub-levels 2p and 3p, from the NIST Electron-Impact Cross Section Database. |
Open with DEXTER |
Table B.1: Fitting parameters for various cross sections.
We have considered only the case of hydrogen atoms crossing the shock front (in the frame of the front) and interacting with protons. In this appendix, we examine the possibility of Ly
being created within the shock front. The width of the shock front in collisional shocks is on the order of an atomic mean free path,
,
assuming a pure hydrogen gas. Zel'dovich & Raizer (1966) have shown that for weak shocks, the collisional shock width is
![]() |
(C.1) |
The question is: how robust is the assumption of shocks in Balmer-dominated SNRs being collisionless? This occurs when the electron and proton gyroradii -
and
,
respectively - are much smaller than
.
The typical value of the magnetic fields in SNRs is
B = B-4 10-4 G. For example, Strom & Duin (1973) find
and
G for Tycho and Cas A, respectively. The electron gyroradius is
![]() |
(C.2) |
About
of the hydrogen atoms cross the shock front without being ionized, where
is the width of the shock. In collisional shocks,
.
In collisionless shocks,
,
and virtually all of the atoms pass through. We thus conclude that broad Ly
is probably not produced in a significant amount within the shock front, consistent with the findings of H07.