As of June 17 2024
Blog by Roger F. Malina
On the 14th of July 1974
My best friend the late Arthur Davidsen and I
Spent a pointless night observing a strange phenomenon
On,
Or in,
The sky.
All night.
Several nights.
At the Lick Observatory
Above Silicon valley
Which had just been named in 1971
It didn’t go away on the second or third nights, so it had to be real.
Silicon valley that is.
Or our discovery was AI-generated.
A.I did exist then .
( The field of AI research was founded at a workshop held on the campus of Dartmouth College in the U.S. during the summer of 1956. )
Observations of the optical candidate were made with the Robinson-Wampler image-tube scanner (ITS).
It was attached to the 3 m telescope of Lick Observatory
On 1974 July 14 and 1975 June 13 and August 12.
I think we used photographic film and developed it in the Lick dark room.
We added our PhD advisor to the author list,
Even though he had no involvement
Except funding the trip
( yes that’s a false authorship anxiety).
He certainly never checked our conclusions
And I can’t check whether we used film or not,
We didn’t make it clear in the paper.
But the thrill was intellectual hedonism
(and maybe eroticism)
I went on to work in astrophysics for several decades.
Arthur died at the age of 57
His father was a commercial fisherman and his mother was a housekeeper
Go figure how he became an astrophysicist,
Hunting for fishy x-ray stars
My first friend to die.
So Arthur, wherever you are
Happy discovery
Strange discovery
Yeah
We got 175 citations so far
So it must be true
But you were a true friend.
And the poem is over
So now to the strange facts:
Conclusion:
The observations presented here indicate that the
proposed counterpart of GX 1+4 has an extremely
unusual spectrum and is, perhaps for that reason alone,
very likely to be associated with the X-ray source.
The object is, however, unlike any of the previously
identified X-ray sources. The composite spectrum
indicates that the object is almost certainly a binary
system, consisting of a red giant and a much hotter
source.
Astrophysical Journal, Part 1, vol. 211, Feb. 1, 1977, p. 866-871.
197 7ApJ. . .211. .8 6 6D
The Astrophysical Journal, 211:866-871, 1977 February 1
© 1977. The American Astronomical Society. All rights reserved. Printed in U.S.A.
THE OPTICAL COUNTERPART OF GX 1+4: A SYMBIOTIC STAR
Arthur Davidsen*
Department of Physics, The Johns Hopkins University
AND
Roger Malina and Stuart Bowyer
Astronomy Department and Space Sciences Laboratory, University of California, Berkeley
Received 1976 June 9
ABSTRACT
Spectrophotometry of the proposed optical counterpart of the hard X-ray source GX 1 +4 is
presented. The spectrum is that of a symbiotic star which is an M giant with a variable blue
continuum and a rich emission line spectrum including H i, He i, Fe n, [Fe vu], and probably
[Fe x]. An explanation in terms of a compact object accreting material from an M6 III companion
near the galactic center appears acceptable. The similarity of the emission line spectrum to the
spectra of Seyfert nuclei is discussed, and it is suggested that photoionization of gas in the binary
system can explain the strength of the high-excitation lines.
Subject headings: galaxies: Seyfert — stars: combination spectra — X-rays: binaries
I. INTRODUCTION
The X-ray source designated GX 1+4 by Lewin,
Ricker, and McClintock (1971) has been observed in a
number of hard X-ray balloon experiments (Thomas
et al. 1975; Haymes et al. 1975; Ricker et al. 1976).
The experimenters are in agreement that GX 1 + 4 is
the dominant hard X-ray source in the galactic center,
and that it is very likely identical with 3U 1728 — 24
(Giacconi et al. 1974), also known as GX 2 + 5 (Haw-
kins, Mason, and Sanford 1973). The spectrum ob-
tained in the balloon observations can be fitted with a
power law with an energy index in the range —1.4 to
—1.8. The data of Haymes et al. (1975) indicate that
it extends to at least several hundred keV.
Variability of GX 1 +4 has been observed on several
time scales. Lewin, Ricker, and McClintock (1971)
suggest a possible periodic variation with P æ 2.3
minutes. Copernicus observations yield a period P =
4.3 minutes with 12-20% modulation (White et al.
1976), while recent SAS-3 observations indicate
periodic variability with P = 122.607 seconds (Doty
1976). The Copernicus observation might represent an
alias of a true modulation at 130 seconds, similar to,
but significantly different from, the SAS-3 period
(White et al. 1976). A factor 3 increase in intensity
from 1970 to 1972 has been reported by Ricker et al.
(1976).
An accurate position for the source was given by
Hawkins, Mason, and Sanford (1973) and further
improved by Mason (cf. Davidsen, Malina, and Bowyer
1976). An optical candidate, first proposed by Glass
and Feast (1973), has been discussed further by
Davidsen, Malina, and Bowyer (1975, 1976), where
coordinates and a finding chart may be found. In this
* Alfred P. Sloan Foundation Research Fellow.
paper we present the results of spectrophotometric
observations of the proposed optical counterpart.
II. OBSERVATIONS
Observations of the optical candidate were made
with the Robinson-Wampler image-tube scanner (ITS)
attached to the 3 m telescope of Lick Observatory on
1974 July 14 and 1975 June 13 and August 12. In view
of the faintness of the object (K æ 19), the individual
spectra, obtained at a resolution ~10Â, have been
averaged together to improve the detectability of weak
features. The resulting spectrum is shown in Figure 1.
The most striking feature of the spectrum is the
overwhelming dominance of the Ha emission line.
However, there are also a strong continuum which
rises sharply in the infrared and a large number of
other emission lines with fluxes on the order of 1%
of the Ha flux. A list of the strongest lines and their
measured fluxes is given in Table 1. Besides Ha and
Hß, the most prominent lines are those of He i (ÀÀ5876,
6678, and 7065), O i A7774, [O m] AA4959, 5007, and
[Fe vu] AA5721, 6086. In addition, many of the weak
features throughout the spectrum may be identified
with lines of Fe h. Multiplets which are definitely
present include (48) and (49) in the yellow, (40) and
(74) in the red, and (72) and (73) in the infrared. No
clear evidence for [Fe n] lines has been found. Al-
though the blue end of the spectrum is rather poorly
detected, it is apparent that He n A4686 and the C m,
N hi blend at ~ 4640-4650 Â, which are often found
in emission in X-ray sources (McClintock, Cañizares,
and Tarter 1975), are much weaker than Hj8, if they
are present at all.
While the [Fe vu] lines, whose ionization potential
is 103 eV, indicate the presence of a high-excitation
region within the source, there is evidence for even
866
© American Astronomical Society • Provided by the NASA Astrophysics Data System
197 7ApJ. . .211. .8 6 6D
9 Z -°
Z H/Z* + WO/S
I X
/OUI) n N j
(O
9Z-0IX
C Z H / 2**U0/S/9 d 3)
(NJ
n N 3
© American Astronomical Society • Provided by the NASA Astrophysics Data System Fig. 1.—Spectrum of the optical counterpart of GX 1+4 obtained with the Lick Observatory image-tube scanner. Three observations covering the wavelength
intervals 4600-6700, 4800-7300, and 5800-8250 Â have been reduced to absolute fluxes by comparison with standard stars and averaged together. Top panel: blue
end of the averaged spectrum. Bottom panel: red end of the averaged spectrum.
197 7ApJ. . .211. .8 6 6D
868 DAVIDSEN, MALINA, AND BOWYER
TABLE 1
Emission Line Fluxes in GX 1 +4
Vol. 211
Identification Measured Flux
(10″15 ergs cm”2 s”1) Reddening Corrected Flux*
(10″13 ergs cm”2 s”1)
* Assuming/L = 3EB-V — 5.1.
higher states of ionization. All three of our individual
spectra reveal a feature at ~6373 Â, which we believe
must be partly due to the [Fe x] A6374 coronal line
(ionization potential = 235 eV). Although some of the
flux in this feature must be attributed to [O i] A6364
and to Fe n A6369 because [O i] A6300 and the other
members of Fe h multiplet (40) at 6433 and 6516 Â are
clearly present, the observed feature is both too strong
and too far to the red to be explained entirely in this
way. The flux given in Table 1 refers to the entire
feature and is therefore an overestimate of the [Fe x]
emission. The [Fe xiv] coronal line at 5303 Â has not
been detected in these observations. Additional evi-
dence of a very high ionization region is provided by
features at 5534 Â and 6919 Â, which may possibly
be identified with [Ar x] A5533 and [Ar xi] A6919,
although the first could also be attributed to Fe n
A5535 multiplet (55). There is also an unidentified
emission feature at ~6830Â, which is common in
many symbiotic stars and novae (Webster and Allen
1975).
It is not possible to obtain accurate radial velocities
from these data, but Ha and the He i lines yield V =
—100 ± 50 km s“1, and all of the suggested identifica-
tions are consistent with this result. Glass and Feast
(1973) measured V = — ISókms-1 for Ha, so there
is as yet no clear evidence for large-amplitude velocity
variations. The flux in the emission lines, however, is
observed to be variable. The flux in the He i lines
increased by a factor 2.6 while Ha increased by a
factor 1.6 over a two-month interval. The continuum
also appears to be variable, yielding V = 18.66 and
19.36 in observations separated by 11 months.
In addition to emission lines displaying a wide range
of ionization, the spectrum contains absorption
features of an M star. The large depression in the
continuum in the 7700-8000 Â region is due to a
strong blend of TiO and VO bands. The TiO band at
A7054 is also present, although it is considerably
weakened by He i A7065 emission. The A7345 band of
VO is also evident. The identification of VO bands
indicates that the spectral type is M6 or later (Albers
1974), while the strength of the 7800 Â blend requires
a spectral type M5 or later. The strength of the TiO
band at A7054 corresponds to M2, but it has already
been noted that the helium emission affects this
determination. This feature may also be filled in by
the continuum observed at shorter wavelengths, which
is much too strong for an M star. Since there is no
evidence for the Na i doublet at AA8183, 8195, we
conclude that the spectrum is not that of a dwarf (see,
e.g., Albers 1974). The presence of the VO bands also
indicates that the M star is not a dwarf (Pesch 1972).
We shall adopt the spectral type M6 III, but in view
of the complicated nature of the spectrum, this
preliminary estimate should be accepted with caution.
in. DISCUSSION
The combination of high-excitation emission fea-
tures with an M-type absorption spectrum is the defin-
ing characteristic of the symbiotic stars (Merrill 1950).
The members of this class, of which a few dozen are
known, are all variable and include the recurring novae
RS Oph and T CrB (cf. Swings 1970), whose spectra
are similar to that of the object discussed here (see,
e.g., Joy and Swings 1945). T CrB has been established
as a binary in which the M3 III component probably
fills its Roche lobe and transfers matter toward a blue
subdwarf companion (Kraft 1958). X-rays could be
produced by such a system if the companion were a
compact object. Photoionization of gas surrounding
the X-ray source might then explain the occurrence of
emission lines up to and including [Fe x].
Inspection of the Palomar Sky Survey prints for the
field of GX 1+4 indicates heavy reddening, and the
optical spectrum discussed here bears this out. If we
assume that the intrinsic Balmer decrement is that
appropriate to radiative recombination in case B
(Brocklehurst 1971), the observed ratio F(Ha)/F(Hß) =
124 implies a reddening EB_V x 3.4. The actual red-
dening is likely to be less, since the Balmer decrement
© American Astronomical Society • Provided by the NASA Astrophysics Data System
197 7ApJ. . .211. .8 6 6D
No. 3, 1977 OPTICAL COUNTERPART OF GX 1+4 869
is probably steepened by self-absorption and collisional
excitation. Another estimate can be obtained from the
infrared colors measured by Glass and Feast and our
result for the spectral type. The measured colors are
J – H = 1.54 ± 0.13 and H – K = 0.75 ± 0.11.
Using the intrinsic colors of M giants and the redden-
ing relationships given by Lee (1970), we find EB_V =
1.7 ± 0.4. A value EB-V = 3.4 would make the infra-
red colors inconsistent with an M-type spectrum. We
adopt Av = 3.07^ _ y = 5.1, which implies that the
intrinsic Balmer decrement is Ha/Hß # 20. The emis-
sion line fluxes, corrected for this amount of extinction,
are given in Table 1.
The distance to the system can be derived from the
spectral type and extinction found above. Taking the
infrared colors and reddening relations of Lee (1970)
and assuming Mv = —0.5 for an M6 III star, we find
= 10 kpc. We note that either luminosity class I or V
would lead to implausible distances. The correspond-
ing X-ray luminosity is L* # 4 x 1037 d102 ergs s“1,
where d1Q is the distance in units of 10 kpc. These
results place the system close to the galactic center and
indicate an X-ray luminosity similar to that of other
high-luminosity X-ray sources (Margon and Ostriker
1973). The expected interstellar absorption corre-
sponding to the derived reddening is A* £ 1 x 1022
cm ~2 (Ryter, Cesarsky, and Audouze 1975 ; Gorenstein
1975).
The continuum observed in the yellow region of the
spectrum is too bright to be due to the M star, whose
expected magnitude is K(M6 III) = 20.7. The observed
values V = 18.66 and V = 19.36 then indicate that a
variable blue component is responsible for most of the
light observed in the V band.
The first aspect of the emission line spectrum which
requires explanation is the enormous ratio
F(Ua)/F(Uß) x 20,
after correction for reddening. In a study of physical
conditions in the nuclei of Seyfert galaxies and QSOs,
Netzer (1975) has shown that large Ha/Hß ratios may
be obtained with Ne > 108cm-3 and with large
column densities such that self-absorption occurs in
the Balmer lines. In particular, he finds F(Hd)IF(Hß) =
19 for Ae = 109 and r0(La) = 106, r0(Ha) = 100,
where r0 is the optical depth at the line center. Similar
conditions may well exist around GX 1+4. The
presence of Fe n lines and the absence of [Fe ii]
indicates ne ^ 107cm-3 (Wampler and Oke 1967).
The weakness of the [O m] fines may also indicate
that collisional de-excitation is important and hence
ne > 106cm-3.
The self-absorption calculations of Netzer indicate
that, while the Balmer fine ratios may be severely
affected, the Ha intensity is not substantially different
from that which is obtained in the ordinary radiative
case B nebula. We may therefore derive the emission
measure from the observed Ha flux. (Hereafter “ob-
served” will refer to the reddening-corrected values.)
We assume Te = 2 x 104 K, take the case B Ha emis-
sivity from Osterbrock (1974), and find
NpNeVld1Q2 = 9.3 x 1059 cm“3 .
Assuming that the emitting gas is fully ionized, we
take Np = Ne = 109 N9. Then the radius of the H +
region implied by the Ha flux is R = (3V/4tt)113 =
6 x 1013 d102l3N9~213 cm, which is about 5 times the
radius of an M6 giant. The ionization parameter f =
LX/NR2 = 11A91/3¿io2/3 near the outer edge of this
H+ region. Although a detailed optically thick calcu-
lation will be required, this suggests that T æ 1-2 x
104 K is a reasonable approximation for the Balmer
line region (Hatchett, Buff, and McCray 1976).
Further support for the parameters suggested is
provided by the X-ray spectrum, which indicates a
column density ~4-10 x 1022cm“2 (White et al.
1976). The Balmer line emitting region calculated above
gives a column density NR = 6 x 1022 d192l3N9113
cm-2, in excellent agreement with the observed value,
and indicates that there is not a great deal of additional
neutral gas in front of the H+ region. This suggests
that R x 6 x 1013 cm is a characteristic size for the
gas distribution in the vicinity of the X-ray source.
Since the O i and Fe n lines must originate in a neutral
hydrogen region, they presumably come from the red-
giant atmosphere. The blue continuum is a factor 10
above the expected bremsstrahlung from the Balmer
fine region, but it could come from a hotter, denser
region closer to the X-ray source.
Another interesting feature of the emission fine
spectrum is the great strength of the He i lines, a
situation which also occurs in the Seyfert galaxies
(Phillips and Osterbrock 1975; Boksenberg et al.
1975). MacAlpine (1976) has suggested that the He i
A5876 fine may be enhanced by a combination of
scattering and collisional excitation processes if the
density exceeds 109 cm-3. He shows that scattering of
He i A10,830 photons may maintain a significant
population of the 2 3F level, allowing collisional
excitation of 3 3D followed by emission of A5876. From
the parameters derived for the GX 1+4 candidate it
appears that this process may also be effective there.
The GX 1 + 4 spectrum also shows a strongly enhanced
He i A7065 fine, with F(7065)/F(5876) = 0.8, compared
with an expected value 0.18 for case B radiative re-
combination at Te = 2 x 104 K (Osterbrock 1974).
This could be due to a large optical depth in 2 3*S-
3 3P A3889 if the density is high. Absorption of these
photons can lead to their conversion to 3 3*S’-3 3F
A4.3 /xm, followed by 2 3F-3 3S A7065.
The presence of strong [Fe vu] emission fines at
AA6087, 5721 is another interesting feature of the
spectrum. These fines are also common in Seyfert
galaxies and have been discussed in some detail by
Nussbaumer and Osterbrock (1970). They arise from
the first excited 1D term, about 2 eV above the ground
3Fterm, and are presumably excited by collisions with
thermal electrons. The occurrence of [Fe vu] in GX
1+4 may be due to photoionization by the X-ray
continuum, as is believed to be the case in the Seyfert
galaxies. Under conditions of photoionization and
collisional excitation, Nussbaumer and Osterbrock
(1970) have calculated line emission coefficients for
[Fe vu]. The strongest line expected is A6087. We may
write the luminosity in this line as
L(A6087) = 1 x 10-13A(Fe+6)F(Fe+6)/(«e, Te) ergs s-1,
where /(Ae, re) is an increasing function of Ne and Te
and/(109, 2 x 104) æ 1 by a small extrapolation of the
tabulated results of Nussbaumer and Osterbrock.
A(Fe+6) and F(Fe + 6) are the number density of Fe + 6
and the volume of the region in which Fe + 6 occurs.
Equating the predicted and observed fluxes, we calcu-
late F(Fe+6), and comparing this with the volume
calculated for the H + region, we find
F(Fe+6)/F(H+) = 1.4 x 10-5[A(H)/A(Fe + 6)]A9/-1.
Here A(H) is the total hydrogen density. Assuming
that all the iron is in the form of Fe+6 in the region
emitting A6087, and taking the solar abundance
Fe/H = 3 x 10″5 (Ross and Aller 1976), we find
K(Fe+6)/F(H+) = O.57V9/”1. This result appears
plausible if the ionization is indeed provided by a
hard X-ray continuum and the density is ^ 109 cm”3,
but a detailed calculation is required. This calculation
also indicates that the iron abundance is unlikely to
be much less than the solar value, as might be expected
if GX 1+4 were a Population II object, since then we
would require F(Fe + 6) > F(H+). A search for an iron
absorption edge and/or iron line emission in the X-ray
spectrum of GX 1+4 would be of interest with regard
to this point.
The suggested presence of the coronal line of [Fe x]
A6374 in the GX 1+4 spectrum is yet another point of
similarity to the Seyfert galaxies (Oke and Sargent
1968; Weedman 1971). Osterbrock (1969) found that
the occurrence of [Fe x] emission could also be under-
stood in terms of a photoionization model. For this
ion photons above 235 eV are required, so one may
infer the presence of soft X-rays at least. Osterbrock’s
calculation indicates that the [Fe xiv] A5303 coronal
line is expected to be 9 times weaker than A6374 in the
power-law photoionization model. This agrees with
the observations for NGC 4151 (Weedman 1971) and
Vol. 211
is also consistent with the absence of A5303 in the
GX 1+4 candidate.
IV. SUMMARY AND CONCLUSIONS
The observations presented here indicate that the
proposed counterpart of GX 1+4 has an extremely
unusual spectrum and is, perhaps for that reason alone,
very likely to be associated with the X-ray source.
The object is, however, unlike any of the previously
identified X-ray sources. The composite spectrum
indicates that the object is almost certainly a binary
system, consisting of a red giant and a much hotter
source. The similarity of this object to recurrent novae
such as T CrB, which is known to be a binary, further
strengthens this conclusion. The distance (10 kpc) and
X-ray luminosity (~4 x 1037 ergs s”1) deduced from
the optical data are consistent with those expected for
an X-ray source in the galactic center direction.
The emission line strengths indicate the presence of
an envelope of gas around the system with density
~109 cm”3 and radius ~6 x 1013 cm. If this gas is
interpreted as a spherically symmetric wind with an
assumed velocity ^lOkms”1 characteristic of M
giants, the implied mass-loss rate is ~ 10″6 A/© yr”1.
Such a rate is plausible and is more than adequate to
power a strong X-ray source by accretion onto a com-
pact companion. If the orbital separation of the two
components is comparable to the size of the gas cloud,
a period of the order of years may be expected.
We have suggested that the emission line spectrum
can be understood in a photoionization model, and
have referred to calculations for Seyfert galaxies which
have similar spectra. No calculations have been re-
ported for binary X-ray sources which predict the Fe
lines observed in the GX 1+4 counterpart. Since the
emission line spectrum of this object is much richer than
that of any other compact X-ray source, it should
provide a good test of theoretical models for the
transfer of X-rays through gas.
We thank H. Spinrad for his encouragement and for
donating some observing time. This work has been
supported by NSF grant AST 75-03735.
Astrological Society • Provided by the NASA Astrophysics Data System