A&A 441, 597-604 (2005)
DOI: 10.1051/0004-6361:20053125
A. D. Schwope1 - V. Hambaryan1 - F. Haberl2 - C. Motch3
1 - Astrophysikalisches Institut Potsdam, An der Sternwarte 16,
14482 Potsdam, Germany
2 -
Max-Planck-Institute für Extraterrestrische Physik,
Giessenbachstr., 85748 Garching, Germany
3 -
Observatoire Astronomique, CNRS UMR 7550, 11 rue de l'Université, 67000
Strasbourg, France
Received 24 March 2005 / Accepted 20 May 2005
Abstract
We present a multi-epoch spectral and timing analysis of the
isolated neutron star RBS1223.
New XMM-Newton data obtained in January 2004 confirm the spin
period to be twice as long as previously thought,
s. The combined ROSAT, Chandra, and XMM-Newton data (6 epochs)
give, contrary to earlier findings, no clear indication
of a spin evolution of the neutron star.
The X-ray light curves are double-humped with pronounced hardness ratio
variations suggesting an inhomogeneous surface temperature with two spots
separated by about
160
.
The sharpness of the two humps suggests a mildly relativistic star with a
ratio between
,
the neutron star
radius at source, and
,
the Schwarzschild-radius,
of
.
Assuming Planckian energy distributions as local radiation sources,
light curves were synthesized which were found to be in overall
qualitative agreement with observed light curves in two different
energy bands. The temperature distribution used was based on the
crustal field models by Geppert et al. (#!Geppert04!#)
for a central temperature of
K and an
external dipolar field of
G. This gives a mean atmospheric
temperature of 55 eV. A much simpler model with two homogeneous spots with
eV and 84 eV, and a cold rest star,
eV, invisible at X-ray wavelengths, was found to be similarly
successful.
The new temperature determination and the new
suggest that the star is older than previously
thought,
yr. The model-dependent distance
to RBS1223 is estimated between 76 pc and 380 pc
(for
km).
Key words: stars: neutron - stars: individual: RBS1223 - stars: magnetic fields - X-rays: stars
RBS1223 was discovered in the course of an optical identification program of
the more than 2000 RASS X-ray sources at high galactic latitude with count
rate
s-1 (Schwope et al. 2000). The RASS error circle of
source #1223 in this catalogue was
without obvious optical counterpart at the DSS limit. Subsequent ROSAT HRI
observations gave an improved X-ray position. Deep Keck imaging,
,
remained without optical counterpart, suggesting a high ratio
,
which excluded anything else than a neutron star
as the X-ray source (Schwope et al. 1999, Paper I).
Follow-up Chandra observations revealed a periodically modulated X-ray signal
with an oscillation period of about 5.16 s (Hambaryan et al. 2002, Paper II).
Retrospectively, we also found those oscillations in the HRI data. The derived
spin-down rate of the neutron star of
s s-1 implied an ultra-high magnetic field,
G, and a characteristic age of around 104 years, if interpreted
as due to magnetic dipole braking. The decay of the field would also
provide an explanation for the X-ray luminosity of the source.
However, at the implied young age the source would
not have been able to travel far away from its birth place. The
non-detection of a close-by supernova remnant at radio wavelength or in X-rays
is then at least puzzling.
We obtained XMM-Newton observations at a first epoch in 2001
(satellite revolution rev 377)
with all three EPIC cameras and both RGS'. A timing and spectral analysis of these
data, together with a second epoch observation,
was published by Haberl et al. (2003, Paper III). The much better photon statistics showed a double-humped
X-ray light curve and a spin period of twice the previously reported
value,
s. Simple blackbody models did not fit the mean X-ray spectra
at both epochs. A much better representation of the observational data was achieved
after including a Gaussian-shaped absorption line superposed on a black-body spectrum.
Under the assumption that the observed line is caused by a proton cyclotron
line, the observed position indicates a field strength
of
G. This value would still imply a detectable spin-down
rate although much smaller than derived in Paper II. If the line were an
electron cyclotron line, the implied field strength would be about
1011 G. One would not expect to be able to detect any spin variation
caused by the field decay over the comparatively short period of time
that this object is observed now (1996-2004).
In this paper we present new data from calibration observations of RBS1223 obtained with XMM-Newton in January 2004 and GO time data obtained with the Chandra LETG on March 30, 2004. We combine all available data sets in order to update the spin history of the object. We describe a model which mimics the pulsed X-ray light curves of RBS1223. It allows the determination of the sizes and temperatures of the spots on the neutron star's surface and also gives a distance estimate.
We re-analysed the ROSAT and the Chandra AO1 observations and also found evidence
of periodic variability in the interval between 10.3110 s and 10.3116 s
in those datasets, although with lower significance than at the 5.16 s
periodicity. The probability that we have a periodic signal in the mentioned
time interval based on the ROSAT data is 23%. For all other cases, the
probability is almost 100%. The measured spin
periods as derived for the six epochs between June 1997 and March 2004 are
listed in Table 1 and displayed in Fig. 1. The values
given there are the most likely periods (highest peaks in the odds' ratio
periodogram),
while the errors given are
uncertainties (68% confidence region,
"posterior bubble''; see Gregory & Loredo 1992 and Paper II).
Weighted linear fits to all data in Table 1 were
performed using different weighting schemes. We either used pure statistical
errors, taking the larger of the two values listed in the table into account,
or gave weights to individual data points according to the time resolution
element of the observation, the number of detected photons, or the probability
of detection of the periodic signal via the odds ratio.
Using statistical weights only,
s s-1 is derived. If weights according to the observed
number of photons and the significance of detection are given in addition, the
resulting
s s-1 (fit shown
in Fig. 1).
Both fits indicate a small spin-down of the star.
The evidence of a true spin-down, however, is rather weak,
and rests mainly on the ROSAT HRI
observation in June 1997 where only 521 photons were collected.
When the first data point from
Table 1 is omitted, a weighted fit to the remaining data even
indicates a spin-up of the neutron star,
s s-1.
We conclude that the presently available data are insufficient to determine
the spin history unequivocally, due to the fact that the period
determination for each observation has insufficient accuracy to connect
subsequent observations without cycle count alias. However, our original
value for the spin-down of RBS1223 is clearly ruled out.
Table 1: Spin period of RBS1223 at given epoch with given instrument. All periods listed are newly derived values.
![]() |
Figure 1: The period history of RBS1223 based on X-ray observations at six epochs. The straight line represents a weighted linear fit to all data. |
Open with DEXTER |
![]() |
Figure 2: Phase-averaged light curve and phase-resolved spectroscopy of the XMM-Newton observations of RBS1223 performed in revs. 561 and 743, respectively. The light curve in the top panel encompasses the spectral band 0.12-1.2 keV. Vertical lines indicate the boundaries of data groups for subsequent spectral analysis. In the panels below are shown the best-fit blackbody temperature, the centre, the width, the integrated flux, and the equivalent width of the Gaussian absorption line, respectively. All these parameters were determined by unconstrained fits (indicated by filled circles), by fits with fixed Gaussian line position (open triangles), and by fits with fixed Gaussian line width (open circles). |
Open with DEXTER |
The background-corrected mean countrate detected in the pn was
s-1 and
s-1
for the CALs. We could not detect any
long-term photometric variation among the observations with XMM-Newton.
In Paper III we have already shown that the Chandra data from AO1
are compatible with the XMM-Newton AO1 data; i.e. long-term changes of the
X-ray brightness between 2000 and 2004 are insignificant.
Using the 72 151 and 69 435 photons detected in the two CALs we created phase-folded X-ray light curves for the two epochs. We then determined a relative phase-shift by correlating the two light curves and finally created one common phase-averaged light curve for both CALs with 100 phase bins in the energy range 0.12-1.2 keV. The result is shown in Fig. 2, top panel.
For our spectral analysis we defined eight phase intervals indicated by
vertical lines in Fig. 2: the centres of the
main and secondary pulses; the centres of the main and secondary
minima; the rise and
the fall to the main and secondary pulses.
We adopted the same spectral model as in Paper III,
i.e. we used a blackbody plus Gaussian absorption line absorbed by the same
amount of interstellar matter
(represented by the column density ). The spectral analysis was
performed within XSPEC (version 11.3).
Three different flavours of our adopted model were tried, the first
with unconstrained parameters, a second with the centre of the
Gaussian absorption line kept fixed and a third with the width of the
line kept fixed.
Visual inspection of the
residuals, as well as an F-test, revealed
that the observed spectrum at any given phase cannot be fitted with just one
spectral (black-body) component. Including the Gaussian absorption line
resulted in a satisfactory fit to all 8 spectra.
The temperature drops from about 90 eV in
the first hump to about 82 eV in the second hump, somewhat dependent on the
chosen model. The spectral parameters are depicted in the bottom panels of
Fig. 2. We found indication for a small change of the line
width, broadest in the light curve humps.
Similarly, the absolute flux and the equivalent width of the Gaussian
absorption line is likely to be highest in the humps. Our search for
a variation of the line centre remained inconclusive.
We regard temperature inhomogeneities as the main source of the observed variations. Here we make an attempt to model the light curves assuming an inhomogenous temperature distribution over the surface of the neutron star. This is a first-order approximation, as a full model would have taken anisotropic radiation into account according to the local magnetic field in a surface element. Those models are not yet available for arbitrary angles between the surface normal and the orientation of the field. As a further complication, the numbers of degrees of freedom of the models would be raised by an order of magnitude. Without a proper normalization scheme and perhaps even much better data in terms of spectral and time resolution, it would be very difficult to discern different models. For the time being we think that the simple approach we are following in this paper is justified by the data.
The surface of the star is tiled into small areas using constant steps in latitude and longitude. A temperature is assigned to each of the surface elements. In model "A'' two spots of circular shape were located at certain positions, while in model "B'' a temperature distribution according to the recent crustal field models by Geppert et al. (2004) is assumed. We assume black-body radiation from each of the elements for the assumed temperature. The modeled photon spectrum is folded through the response of the XMM-Newton EPIC-pn camera. The predicted count-rate at any given time (phase) is the summed rate of all visible tiles at that phase, with an appropriate fore-shortening factor for each of the tiles. A similar approach was used by Page (1995), who gives a full description of the details.
The number of visible tiles at any given time depends on the
inclination i of the observer (angle between the rotation axis and
the line of sight at infinity), the angle between the magnetic axis
and the rotation axis ,
and the compactness of the neutron
star, parameterized by the ratio
(
:
radius of the neutron star (rest frame),
:
Schwarzschild-radius).
The radius and temperature in the observer's frame at infinity are related
to the quantities in the star's rest frame by
,
,
(z: gravitational redshift).
The relevant parameters for modelling the light curves are then the
radius and temperature distribution of the star, the compactness, the
distance, the geometry, and the column density
of the interstellar
absorption. Due to the unknown physical nature of the absorption
feature at 0.3 keV, we restricted our light-curve analysis to the spectral
regime above 0.6 keV, which is mainly unaffected by this
feature. Throughout the analysis we fixed
at the value
derived in the previous section,
cm-2. We perform our analysis in two energy bands,
B1
= 0.6-0.8 keV,
B2 = 0.8-1.0 keV; the corresponding hardness ratio
is defined in the usual manner as
HR = (CR2 - CR1)/(CR1+CR2),
with CR1,2 the count rates in the two defined energy bands.
For any given
,
the hardness ratio is a direct measure of
the blackbody temperature.
Beloborodov (2002) has studied the bending of light in the vicinity of
compact objects. He derived a simple approximate relation between the
local angle of photon emission
and the escape direction,
:
.
It has a
maximum deviation of 3% for
with respect to the
exact theory (Pechenick et al. 1983). The visible fraction of the star
is derived from this relation by setting
(see his
Fig. 1). We use this approximation in
order to determine the visible fraction of the neutron star's surface,
as well as the foreshortening angle of individual surface
elements.
Beloborodov also studied the possible types of light curves as a
function of i and
if two spots are located diametrically
opposite on the magnetic axis. His class III describes the current
situation of RBS1223 with the two spots subsequently rotating into view.
This type of light curve is observed when
,
i.e. when both angles i and
are sufficiently large. Although a large
range of combinations between i and
should
produce light curves of type III, the pulsed fraction is maximized
for
,
which we assume in the following.
We ran a number of simulations exploring the parameter space to search for an optimum solution by a fit-per-eye. We finally accepted a light-curve solution giving best overall qualitative agreement with the observed data which is shown in Fig. 3. The parameters in our model "A'' are constrained by the following observed features.
The pulsed fraction of the light curve is rather
high,
in the
soft band B1, which constrains the compactness of the star. For
too large a fraction of the star becomes visible at any given
time. This damps the amplitude of the variation below the observed value.
In the following we assume
.
This retrospectively justifies
the use of Beloborodov's approximative formula for the description of light
bending, which is applicable only to objects that are not too compact.
The observed large pulsed fraction also requires a
high inclination i (and a high
for our assumed geometry with
),
,
although to some extent one can trade a higher
compactness
parameter against a lower inclination and vice versa. The model light
curve shown in Fig. 3 was computed for
and
.
The temperature of the main spot is fixed by the observed hardness
ratio in the main hump of the light curve to
eV.
The width of the
main hump constrains the maximum extent of the spot to about
(full opening angle).
![]() |
Figure 3: Soft X-ray light curves of RBS1223 in energy bands 0.6-0.8 keV and 0.8-1.0 keV ( upper panel) and corresponding hardness ratio ( lower panel). The solid lines are predicted light curves based on the two spot model as described in the text. |
Open with DEXTER |
The location of the second spot is constrained by the observed separation
between the maxima and minima of the light curve. We locate the second
spot at an offset angle of
with respect to the magnetic
axis and at an azimuth of about
.
The offset
is not very
well constrained by the observations, and could become larger
if the spot is located closer to the meridian through the rotation and
the magnetic axis. Under extreme circumstances,
can become as large
as
,
although the fit then clearly deteriorates;
hence, we estimate the minimum separation between the two spots to be
130
.
The temperature of the second spot is determined by the observed spectral
hardness in the pulse maximum,
eV. The width of the second
pulse and the relative brightness of the second with respect to the
first spot then suggest a full opening angle somewhat larger than for
the main pulse,
.
![]() |
Figure 4: Observed data as in Fig. 3. The model light and hardness ratio curves are based on crustal field temperature profiles according to Geppert et al. (2004). Model parameters are described in the text. |
Open with DEXTER |
The contribution of the remaining surface in this model is negligible,
since the light and hardness ratio curves are almost completely
determined by emission from the two spots. This constrains the
atmospheric temperature to
eV. The maximum
possible
is 50 eV. This limit is derived under
the extreme assumption
.
In this non-relativistic
limit the two spots rotate alternating into view, and the emission
between the two humps is entirely due to photospheric
emission.
Finally, the modeled light curve is normalized to the observed
curve by a
factor. For our model "A'' we derive a distance Dof only 76 pc for an assumed
km star (
km
for
). Our best-fit model curve
for model "A'' is shown in
Fig. 3. The corresponding temperature variation along stellar
latitude is shown in Fig. 5 for an assumed atmospheric temperature
of 42 eV.
It is again emphasized that the fit of the light curve is based
on an adaption per eye and not due to a formal least-squares fit.
Having set the main scenery with our ad hoc two-spot model, we proceed
by comparing the observed light curves with the predictions for the
temperature distribution in a magnetized neutron star crust. Such
distributions have recently become recently (Geppert et al. 2004), and these
authors kindly provided tabulated data. Geppert et al. calculated the
temperature distributions in the atmosphere for the "core'' and
"crustal'' field scenario. They solve the equation of heat transport through
the star's atmosphere in the presence of a given dipolar field, the main
parameters which determine the temperature distribution being the (isothermal)
core temperature
and the dipolar field strength. The T-distribution of
their crustal models for sufficiently high magnetic field strength resembles
the situation studied here with two warm spots. Moreover, the maximum
temperature of their
K model is about the same as the spot
temperature in RBS1223. The "core'' field scenario à la Greenstein & Hartke
(1983) always produces too a flat T-distribution and is not
considered here any further.
In order to compute synthetic lightcurves, we used the published
T-distributions from Geppert et al. (plus one unpublished distribution for
K and
B = 1014 G).
We parameterised the maximum temperature as a function of given
and Busing a bi-polynomial interpolation. This allows easy
scaling of a chosen T-profile to the observed maximum temperature.
For lightcurve synthesis we used the same viewing and magnetic geometry and
the same compactness parameter
as above for our initial toy model "A''.
The star was divided into two
half-spheres so that different T-distributions could
be applied to the two halves. This somewhat artificial approach was
necessary in order to simulate the two observed unequal spots.
The two half spheres were slightly inclined with respect to each
other, in order to match the observed phases of the light curve maxima and
minima.
The normalised T-profiles chosen for the main and secondary humps
correspond to
K,
,
and
K,
,
respectively.
A maximum temperature of 98 eV in the main spot was achieved
by assuming a core temperature
K
and a dipolar field of
.
In order to model the observed lower
temperature in the second spot, we had to assume a slightly lower temperature
K and a slightly lower field strength
G as
well. This parameter combination then yields the maximum temperature in the
spot of 91 eV.
The fit based on the described combination of parameters
is shown in Fig. 4. As above, we used a radius of the neutron star
km. The implied distance derived for the crustal field
model is then 380 pc. This value is clearly much larger than that of the
simple two-spot model, since a much larger fraction of the star's surface
contributes to the observed light.
As a result, the crustal field temperature profiles seem to be equally suited
reflecting the observed light curves of RBS1223. We note, however, that our approach
is not self-consistent, since the maximum spot temperature and the T-profiles were
adapted in separate steps. In addition, the simulated T-profiles
were computed for a more compact star,
km, M =
1.4
,
than assumed here.
This might explain the inconsistency
between the values of the parameters
used to determine the maximum
temperature in the spots and the corresponding T-profile used by us.
The temperature
distribution which matches the observed narrow pulses is steeper
than implied by the maximum temperature in the spot.
![]() |
Figure 5: Temperature distribution as a function of magnetic colatitude for models "A'' (dashed line) and "B'' (solid line), respectively. The temperatures are given in eV. The stellar temperature of model "A'' is an upper limit as derived from the light curve modeling. This limit violates the constraint from the broad-band spectral energy distribution as sown in Fig. 6, which requires a remaining star of almost zero temperature. |
Open with DEXTER |
![]() |
Figure 6:
Broad-band spectral energy distribution of RBS1223.
Shown with solid lines are the absorbed blackbody spectra which agree with
the XMM-Newton EPIC pn data at phase of the main hump. Shown
with dotted lines are the unabsorbed extrapolations to low
energies for models A (lower red lines) and B (upper black lines),
respectively. The data point at 2.4 eV represents the
![]() |
Open with DEXTER |
The long-term observations and the analysis presented here shed some new light on the likely nature of RBS1223. We mention firstly, that there is no long-term change of the brightness of RBS1223. This makes Bondi-Hoyle accretion as powering mechanism of the X-ray source unlikely.
The spin-phase averaged light curve is double-humped with two humps of different count rate and spectral hardness. The humps are separated by about 0.47 phase units, the minima by about 0.43 phase units, which indicates a slight asymmetry of the shape of the individual humps. The double-humped light curves are indicative for the presence of a moderately strong field which is not axisymmetric. A displaced dipole or contribution(s) from higher multipoles are likely causes for the asymmetry.
We performed a phase-resolved X-ray spectral analysis of the two calibration observations in 2003 and 2004. A successful fit to the spectra at all phases was achieved by the combined black-body plus Gaussian absorption line model. There is a clear temperature variation over the spin cycle, while the line centre, the line flux, and the line equivalent width also show cyclic changes. The combined new data do not allow us to constrain the nature of the Gaussian further, but we still regard the cyclotron absorption line scenario as possible.
The observed variations of the spectral parameters are, however,
different from those observed in RX J0720.4-3125 (henceforth RX 0720) by Haberl
et al. (2004). In RX 0720 the temperature variations are much smaller,
only
eV. The equivalent widths of the putative
cyclotron absorption lines change in both systems by about 50%, and the
equivalent width in RBS1223 is a factor 2-3 larger than in RX 0720.
While RX 0720 shows a trend toward larger equivalent width corresponding
to lower blackbody temperature, there is no such trend in RBS1223.
The double-humped light curve is suggestive of a spotted atmosphere of the
star with two spots separated by about 160 degrees. We presented two
similarly successful fits to the X-ray light curves in two energy
bands. The chosen energy bands for our experiment,
0.6-0.8, and 0.8-1.0 keV,
are thought to be unaffected by the absorption feature,
so that simple blackbody spectra could be used for light curve
synthesis. Taking the view
of the simple model "A'', the star is described by two visible spots of size
8
and 10
and a, for X-ray eyes, invisible remaining surface.
The atmospheric temperature of less than 42 eV (500 000 K) leads to a revised
age estimate of about
years, derived from cooling curves of
Lattimer & Prakash (2004). This revision solves the problem between the
estimated young age and the non-detection of an associated supernova remnant
(Paper II).
In Fig. 6 the broad-band spectral energy distribution of
RBS1223 is shown. The best-fit X-ray spectra for the main hump based
on our models "A'' and "B'' (mean spectra of the visible hemispheres)
are extrapolated to the optical.
The data point in the optical wavelength range at 2.4 eV
stretching from 3000 to 9000 Å represents the brightness of the
mag tentative optical counterpart dicovered by Kaplan et al. (2002a). The extrapolation indicates that the brightness of the likely
optical counterpart is compatible with our models. Any further low
temperature component would raise the predicted flux at optical
wavelength over the observed value. However, the blackbody models which were
adjusted in the energy bands 0.6-0.8, and 0.8-1.0 keV, already
overpredict the X-ray flux below 0.5 keV where the Gaussian absorption line is
observed. Extrapolation into the optical may, therefore, be done only with
caution.
The predicted distance to RBS1223 based on model "A'' is
uncomfortably short, 76 pc, i.e. nearer than the much brighter prototypical
system RX J1856.5-3754 (
pc, Kaplan
et al. 2002b). The short distance is due to the simplicity of the
temperature structure and geometry. Any temperature profile with a
more gradual variation in the temperature and corresponding larger
parts of the stellar surface contributing to the observed light will
require larger distances to the star.
Model "A'' was used as a toy model, in order to constrain
the geometry and the compactness, but it lacks a physical interpretation.
This physical interpretation is possible with our model "B''
based on the crustal field model by Geppert
et al. (2004). This reflects the data equally well.
In this model the spots are more extended,
i.e., larger parts of the atmosphere contribute to the observed
radiation. The estimated distance is consequently much larger,
pc; all distance estimates were made for a 1
star with radius
km and
.
The average temperature of the
atmosphere is higher than for model "A'', 55 eV, but still clearly lower than
previously assumed, so that the age problem is also solved with model "B''.
Given the very restricting assumptions of our spectral model, pure
black-body emission with a simple foreshortening law,
the results of the study presented here can be regarded as only a first
step towards a full model of the star. As worked out by Pavlov et al. (1996)
and re-addressed by Zavlin & Pavlov
(2002), atmosphere model fits give temperatures
significantly
lower than the blackbody temperature
,
dependent on the chemical
composition. Light-element models, however, which would show the most drastic
effect are likely to be ruled out, since they predict much higher fluxes in the
RJ-part of the spectrum. They also would predict much shorter distances
compared to bb-models, which doesn't seem very likely.
Fe-models, on the other hand, do not show dramatic
differences as far as
and the predicted distance are
concerned. Since they produce less optical flux, they seem to be much better
suited to fit the spectral energy distribution than bb- or light element
models. Another point of uncertainty is the type of the assumed
foreshortening. Zavlin & Pavlov (2002) computed the angular characteristic of
specific intensities with an assumed field perpendicular to the surface, which
shows strongly peaked emission normal to the surface. If applicable to the
case of RBS1223, this would clearly affect the spin-phased light curves.
Future light curve models would
strongly benefit from the availability of a dense grid of models for
different temperature, magnetic field strength, chemical composition, and
angle between the local magnetic field and the surface normal. Based on such
a grid, light curve modeling would be possible via a regularisation scheme in
the multidimensional parameter space.
In sum, the revised value for ,
together with the revised value for
the mean atmospheric temperature, suggests a nature of RBS1223 as a medium-aged
neutron star (
yr) on its cooling track. The nature of the
Gaussian is not understood, but it could be a cyclotron absorption line in a field
of a few 1013 G, still compatible with the (uncertain) spin-down. The
present distant estimates make the star similarly close to the prototype
RX J1856, which has a large observed proper motion and a measured parallax.
Although very much fainter,
(Kaplan et al. 2002a),
a similar measurement
for RBS1223 seems to be feasible and rewarding in order to further constrain
the likely distance, radius, and thus the nature of this star.
Acknowledgements
We thank U.R.M.E. Geppert for providing tabulated temperatures of his crustal field models and for fruitful discussions. V.V.H. is supported by the Deutsches Zentrum für Luft- und Raumfahrt (DLR) under contract No. FKZ 50 OX 0201.