See discussions, stats, and author profiles for this publication at: [Link]

net/publication/51945563

Magnetic Field and Atmospheric Chemical Abundances of the Magnetic Ap

Star HD 318107

Article in Astronomy and Astrophysics · October 2011

DOI: 10.1051/0004-6361/201117856 · Source: arXiv

CITATIONS READS

19 65

8 authors, including:

John Darlington Landstreet Luca Fossati

The University of Western Ontario Austrian Academy of Sciences (OeAW)
382 PUBLICATIONS 13,279 CITATIONS 527 PUBLICATIONS 10,752 CITATIONS

SEE PROFILE SEE PROFILE

Claudia Paladini James Silvester

European Southern Observatory Uppsala University
138 PUBLICATIONS 1,493 CITATIONS 50 PUBLICATIONS 1,293 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Luca Fossati on 21 May 2014.

The user has requested enhancement of the downloaded file.

 Astronomy & Astrophysics manuscript no. 17856 c ESO 2011
October 17, 2011

Magnetic Field and Atmospheric Chemical Abundances of the

Magnetic Ap Star HD 318107⋆
J. D. Bailey1⋆⋆ , J. D. Landstreet1,2 , S. Bagnulo2 , L. Fossati3 , O. Kochukhov4 , C. Paladini5 , J. Silvester6 , and G. Wade6

1
Department of Physics & Astronomy, The University of Western Ontario, London, Ontario, N6A 3K7, Canada
2
Armagh Observatory, College Hill, Armagh, Northern Ireland
arXiv:1110.3053v1 [[Link]] 13 Oct 2011

3
Department of Physics & Astronomy, Open University, Milton Keynes, UK
4
Department of Astronomy & Space Physics, University of Uppsala, Uppsala, Sweden
5
Institut für Astronomie, Universität Wien, Türkenschanzstrasse 17, 1180 Wien, Austria
6
Department of Physics, Royal Military College of Canada, Kingston, Ontario, Canada K7K 7B4
Received 8 August 2011 / Accepted 27 September 2011

ABSTRACT

Context. A new generation of powerful and efficient spectropolarimeters has recently been used to provide the first sample of magnetic
Ap stars of accurately known ages. Modelling of these data offer the possibility of significant new insights into the physics and main
sequence evolution of these remarkable stars.
Aims. New spectra have been obtained with the ESPaDOnS spectropolarimeter, and are supplemented with unpolarised spectra from
the ESO UVES, UVES-FLAMES, and HARPS spectrographs, of the very peculiar large-field magnetic Ap star HD 318107, a member
of the open cluster NGC 6405 and thus a star with a well-determined age. The available data provide sufficient material with which to
re-analyse the first-order model of the magnetic field geometry and to derive chemical abundances of Si, Ti, Fe, Nd, Pr, Mg, Cr, Mn,
O, and Ca.
Methods. The models were obtained using ZEEMAN, a program which synthesises spectral line profiles for stars that have magnetic
fields. The magnetic field structure was modelled with a low-order colinear multipole expansion, using coefficients derived from the
observed variations of the field strength with rotation phase. The abundances of several elements were determined using spectral
synthesis. After experiments with a very simple model of uniform abundance on each of three rings of equal width in co-latitude and
symmetric about the assumed magnetic axis, we decided to model the spectra assuming uniform abundances of each element over the
stellar surface.
Results. The new magnetic field measurements allow us to refine the rotation period of HD 318107 to P = 9.7088 ± 0.0007 days.
Appropriate magnetic field model parameters were found that very coarsely describe the (apparently rather complex) field moment
variations. Spectrum synthesis leads to the derivation of mean abundances for the elements Mg, Si, Ca, Ti, Cr, Fe, Nd, and Pr. All
of these elements except for Mg and Ca are strongly overabundant compared to the solar abundance ratios. There is considerable
evidence of non-uniformity, for example in the different values of hBz i found using lines of different elements.
Conclusions. The present data set, while limited, is nevertheless sufficient to provide a useful first-order assessment of both the
magnetic and surface abundance properties of HD 318107, making it one of the very few magnetic Ap stars of well-known age for
which both of these properties have been studied.
Key words. stars:magnetic fields - stars: chemically peculiar

1. Introduction cal values for normal stars of similar mass. The magnetic field
strength, spectral line strengths and shapes, and brightness in
The existence of main sequence A- and B-type stars with strong various photometric bands all vary with the rotation period of
magnetic fields has been known for decades. These peculiar stars the star. This variability is explained using the oblique rotator
(referred to as Ap stars) show anomalous abundances of partic- model: the magnetic field axis and rotation axis of the Ap star
ular elements, for example of Cr, which may be as much as 102 are not aligned with the line of sight, nor with one another, and
times (2 dex) overabundant compared to the Sun. It is also not several elements are distributed non-uniformly over the surface
uncommon to find atmospheric abundances of rare earths well in in a pattern that is not axisymmetric about the rotation axis. This
excess of solar (Ryabchikova, 1991). Almost all Ap stars have means that observations through the rotation cycle lead to vary-
angular momenta that are of the order of 10% or less of typi- ing field measurements as a result of observing the magnetic
⋆
field from different aspects. Similarly the spectrum, and (as a
Based on observations made with the European Southern result of line blocking and backwarming) photometric magni-
Observatory telescopes under ESO programmes 065.I-0644(A), 075.C-
0234(B), 079.C-0170(A), 081.C-0034(A), and 082.C-0308(A),obtained
tudes and colours, vary as different parts of the star are observed
from the ESO/ST-ECF Science Archive Facility. (Ryabchikova, 1991).
⋆⋆
Based on observations obtained at the Canada-France-Hawaii
Telescope (CFHT) which is operated by the National Research Council HD 318107 (=NGC 6405 77), a very peculiar magnetic Ap
of Canada, the Institut National des Science de l’Univers of the Centre star, has an effective temperature T eff = 11 800 ± 500 K, lumi-
National de la Recherche Scientifique of France, and the University of nosity log L/L⊙ = 1.92 ± 0.1, and mass M/M⊙ = 2.95 ± 0.15
Hawaii. (Landstreet et al., 2007). It has a large global magnetic field;
2 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107

the mean line-of-sight magnetic field component hBz i is some- Sect. 5 discusses the abundance models obtained; and Sect. 6
times as large as 5 or 6 kG, and the typical size of the mean summarizes the work presented.
field modulus hBi is about 15 kG. The star is a member of the
open cluster NGC 6405, so that it is known to have an age of
log t = 7.80 ± 0.15 (yr). This means that the star has completed 2. Observations
about 17 ± 7% of its main sequence lifetime (Landstreet et al.,
2007, 2008). It is still a rather young star. 2.1. New data
Initially, a rotation period of 52.4 days was proposed by
North (1987); however, more extensive photometric data ob-
tained by Manfroid & Mathys (2000) combined with measure-
Previous observations of HD 318107 include photo-
ments of the mean surface magnetic field hBi led to the conclu-
electric photometry reported by North (1987) and by
sion that the rotation period is P = 9.7085 ± 0.0021 days.
Manfroid & Mathys (2000) and used for determination of the
The magnetic field of this star was modelled by rotation period; a series of (unpolarised) spectra obtained by
Landstreet & Mathys (2000) based on data from Mathys et al. Mathys et al. (1997) from which the mean field modulus hBi was
(1997), Mathys & Hubrig (1997) and further unpublished mag- measured, and used by Manfroid & Mathys (2000) to redeter-
netic field measurements by Mathys. Landstreet & Mathys mine the rotation period; and a small number of spectropolari-
(2000) adopted a model of colinear magnetic dipole, quadrupole, metric observations reported by Mathys & Hubrig (1997) from
and octopole components of strength +23700, -23600, and which the mean longitudinal field hBz i was measured. There is
+8300 G respectively. Observations of the hemispherically aver- also one new measurement of hBz i using ESO’s FORS1, reported
aged line of sight component of the magnetic field hBz i indicate by Bagnulo et al. (2006).
that hBz i is always positive, which leads to the model constraint
The spectra used in this project were obtained more re-
that the sum of the inclination of the rotation axis to the line of
cently than the observations noted above, and come from a va-
sight, i, and the angle between the magnetic field axis and the
riety of sources. The data available for spectral line synthesis
rotation axis, β, must be less than about 90 degrees: i + β . 90◦ .
consist of three newly acquired ESPaDOnS spectra and a new
This implies that the negative magnetic pole is never directly
FLAMES-UVES spectrum (Pasquini et al., 2002), together with
observed. For this model, the values for i and β were chosen
seven HARPS spectra and one UVES spectrum from the ESO
to be 11◦ and 78◦ , with uncertainties discussed in the text of
archive.
Landstreet & Mathys (2000). Note that this model does not re-
ESPaDOnS is a cross-dispersed echelle spectrolarimeter lo-
produce the observations very well. It appears that the field struc-
ture of HD 318107 is more complex than the axisymmetric low cated at the Canada-France-Hawaii telescope, which can mea-
sure all the Stokes parameters (I, Q, U, V). We have three flux
order multipole field geometry used, and is probably not even
(I) and circular polarisation (V) spectra which cover almost the
axisymmetric. Nevertheless, the model provides a first order ap-
proximation to the magnetic field geometry of HD 318107. entire window from 3700 Å to 1.04 µm with a resolving power
There are rather few hot Ap stars with strong magnetic fields of R ≃ 65 000. These spectra typically have signal-to-noise ra-
for which detailed chemical abundance analyses have been car- tios (S NR) of the order of 200 over much of this window. Two of
ried out. HD 318107, with T eff = 11800 K, measured mean the spectra were obtained one after the other, and thus are at es-
field modulus of hBi ∼ 15 kG (Mathys et al., 1997) and v sin i ≈ sentially the same rotational phase. The I spectra were used for
7 km s−1 , is an excellent object for which to carry out such an measuring the mean field modulus, and for modelling the chem-
analysis. The low v sin i of the star together with the large field ical element abundance distribution over the stellar surface. hBzi
strength make it possible to observe resolved Zeeman splitting was measured from the I and V spectra and used to constrain the
in many spectral lines. New measurements of hBi and hBz i are magnetic model.
reported here, with which it is possible to improve the precision HARPS is located at the European Southern Observatory
of the rotation period and re-examine the magnetic model. (ESO) La Silla 3.6 m telescope. It is a cross-dispersed echelle
HD 318107 is also interesting in that it has an age known spectrograph which covers a spectral range of 3780 - 6910 Å
with relatively high precision. The large sample of magnetic with R = 115 000. The available spectra have S NR between
Ap stars in open clusters for which magnetic fields have about 40 and 100. Since the available HARPS spectra do not
been measured and characterised by Bagnulo et al. (2006) and contain any circular polarisation information, these data were
Landstreet et al. (2007, 2008) have already provided the first used for measurement of hBi for improvement of the rotation
clear evidence of evolution of magnetic field strength of the fos- period, and for abundance modelling of HD 318107.
sil fields of middle main sequence stars with time (or stellar age) UVES is a high resolution, cross-dispersed optical spectro-
during the main sequence phase of evolution. This data set, much graph located at ESO’s Paranal Observatory. One spectrum was
of which is based on relatively high-dispersion (polarised) spec- obtained using the FLAMES instrument, with a resolution of
tra, also offers the possibility of studying empirically the evo- R = 47 000 in a spectral window covering the ranges 4200 to
lution with time and stellar age of the chemical peculiarities of 5160 Å and 5230 to 6200 Å. This spectrum was used to measure
magnetic Ap stars. This is now being undertaken; HD 318107 is the mean field modulus hBi. The other spectrum was obtained
the first star from this new sample to be analysed for its chemical with the standard UVES red arm setting at 580 nm, using image
abundance and distribution properties. slicer #3, resulting in a resolving power of 110 000.
This paper will discuss our efforts to model the magnetic The new data are summarised in Table 1. The first column
field and the chemical abundance distributions of several ele- reports the instrument used, and the second gives the Julian Date
ments for HD 318107. The following section discusses the spec- at the middle of the observation. The meaning of the remaining
troscopic and spectropolarimetric observations used and the field columns will be discussed in more detail below.
strength values deduced from them; Sect. 3 discusses improve- The radial velocity of HD 318107 during each observation
ment of the precision of the period; Sect. 4 describes the mod- was measured either by fitting an appropriate model spectrum to
elling technique and the spectral line synthesis program used; that single observed spectrum during the process of modelling
 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107 3

Table 1. The new spectroscopic and polarimetric data used in our analysis. Recorded are the HJD and derived phases for each
spectra. The values of hBz i for the ESPaDOnS spectra are shown followed by hBi values for all spectra measured from three spectral
lines: Fe ii λ6149, Nd iii λ5050 and λ6145.

Instrument HJD Phase SNR hBz i hBi (G) Window RV

(G) Fe ii λ6149 Nd iii λ5050 Nd iii λ6145 (Å) (km s−1 )
ESPaDOnS 2454553.119 0.991 250 4900 ± 45 13900 ± 380 12600 ± 340 12400 ± 280 3690-10481 −9
ESPaDOnS 2454553.150 0.995 250 4856 ± 42 3690-10481 −9
ESPaDOnS 2454652.793 0.258 320 1985 ± 31 15400 ± 380 16300 ± 340 16800 ± 280 3690-10481 −9
HARPS 2453463.898 0.803 80 13600 ± 420 14900 ± 410 15100 ± 420 3783-6914 −9
HARPS 2453582.645 0.034 90 13600 ± 500 11900 ± 680 12000 ± 670 3783-6914 −9
HARPS 2453583.693 0.142 40 13900 ± 420 14100 ± 550 14200 ± 570 3783-6914 −9
HARPS 2454223.642 0.056 70 13200 ± 310 12300 ± 510 13000 ± 410 3783-6914 −9
HARPS 2454633.613 0.283 55 14700 ± 250 16400 ± 410 16400 ± 340 3783-6914 −8
HARPS 2454869.850 0.615 70 13800 ± 420 15700 ± 680 15300 ± 570 3783-6914 −8
HARPS 2454870.879 0.721 60 13600 ± 420 14700 ± 680 15900 ± 570 3783-6914 −8
UVES 2451692.894 0.390 220 15900 ± 210 17000 ± 280 16400 ± 380 4795-6745
FLAMES–UVES 2454256.625 0.453 150 16500 ± 340 16400 ± 510 16900 ± 420 4200-6200 −8

the chemical abundance distributions (see below), or by mea- nomenon that is found in several of the measurements reported
suring the wavelengths of a few spectral lines of low Zeeman by Bagnulo et al. (2006) (for NGC 2169-12, NGC 2244-334,
sensitivity. The measured radial velocity varies somewhat be- HD 149277, CD -48 11051, and HD 318107). In each case the
tween different methods of measuring it, as the line profiles of star has a hBzi field of several kG, and all have T eff values above
individual lines are (often strongly) distorted by Zeeman split- about 11 500 K. Because the field values are very large, there
ting or by blends. However, the various methods usually differ is no question about the validity of the detections, but the hBzi
by no more than about 1 km s−1 , so we conservatively estimate values measured in different ways can differ by as much as a
the uncertainty of radial velocities to be about ±2 km s−1 . The factor of two. Why this kind of discrepancy occurs for some (but
measured value for each spectrum, rounded to the nearest 1 km not all) large-field stars is not yet understood. It might be con-
s−1 , is reported in the final column, labelled RV, of Table 1. The nected with the failure of the weak-field approximation which is
mean radial velocity of the star is −8.8 km s−1 , and there is used to deduce hBz i values from FORS spectra. However, it may
no hint in our data, which were taken over a time span of more simply be the consequence of magnetic field vectors which vary
than eight years, of any variability. This result is consistent with greatly over the visible hemisphere, possibly even on a fairly
the mean radial velocity of NGC 6405 (about −7 km s−1 ) as small scale, and which are sampled differently when averaged
reported by Kharchenko et al. (2005), and supports the member- using different chemical elements, which in general are not dis-
ship of HD 318107 in this cluster. However, our measurements tributed uniformly over the stellar surface.
are all quite different from that reported by Manfroid & Mathys To explore the possibility that inconsistent field measure-
(2000) who measure the radial velocity to be +9.4 ± 2.0 km s−1 . ments are due mainly to the inhomogeneous structure of the
This discrepancy may be the result of a sign error by Manfroid magnetic field and the surface distribution of various elements,
& Mathys, as our new radial velocities show no indication of we have remeasured the longitudinal field hBzi in our three
variability. ESPaDOnS spectra using spectral lines of only a single element
at a time. There are enough lines of Si, Ti, Cr and Fe to provide
LSD field measurements of high enough accuracy to distinguish
2.2. Longitudinal field strength measurements differences at the few hundred G level. The results of this exper-
The new ESPaDOnS V and I spectra provide measurements of iment are listed in Table 2.
both hBz i and hBi. The hBzi values were obtained by using the
first-order moment of the high signal-to-noise V line profile cre- Table 2. Measurements of hBzi using LSD masks composed
ated by the technique of Least Squares Deconvolution (LSD). of lines of single elements (Si, Ti, Cr, and Fe) from the three
The method used is discussed in detail in Landstreet et al. ESPaDOnS spectra (cf Table 1).
(2008). The line list initially used for the LSD processing was
a generic list of about 2900 lines suitable for an Ap star of JD hBz i(Si) hBz i(Ti) hBz i(Cr) hBz i(Fe)
T eff = 12 000 K, containing primarily lines of the iron peak ele- −2450000 (G) (G) (G) (G)
ments Ti, Cr and Fe, together with some lines of lighter elements 4553.119 7031 ± 379 4193 ± 231 4215 ± 139 5430 ± 50
such as Mg and Si. The integration limits were chosen by eye for 4553.150 7107 ± 370 4222 ± 221 4124 ± 132 5237 ± 47
each spectrum. The resulting measurements are listed in the fifth 4652.793 2007 ± 269 708 ± 148 878 ± 92 2417 ± 32
column of Table 1. The stated uncertainties were obtained by
propagating errors from the underlying V and I LSD spectra.
In addition to the three measurements reported here, one
other new measurement of hBz i is available from Bagnulo et al. From Table 2 a number of conclusions may be drawn. First,
(2006). This measurement was obtained using FORS1, and sep- the very satisfactory agreement of the individual entries in the
arate values are reported there for the hBz i value obtained us- top two lines, derived from two independent spectra obtained
ing only the Balmer lines (+6519 ± 55 G), and the value from within an hour of one another indicates that the claimed pre-
the metallic spectrum (+3784 ± 59 G). It is obvious that these cision (measured by repeatability) of the field measurements is
two measurements agree poorly with one another. This is a phe- believable. On the other hand, in each of the ESPaDOnS spec-
4 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107

tra, the values of hBz i measured using different elements reveal mean field modulus. The sigma components of both Nd iii lines
remarkable differences. Cr and Ti seem to sample the field in are nearly symmetric, which allows for the measurement of hBi
similar ways; both near phase 0 and phase 0.25 the measured to be nearly approximation free. This is consistent with the pro-
hBz i values are not significantly different. Fe, however, seems to cedures outlined by Manfroid & Mathys (2000). The results of
sample more strongly regions of field lines parallel to the line of these measurements are summarized in Table 1 in columns 6
sight than the other two iron peak elements, and this tendency through 8.
is still stronger in the sampling of the field associated with Si, To estimate uncertainties, multiple measurements for each
which, however, does not sample the field in the same way as spectrum were made by repeatedly fitting Gaussians to indi-
Fe. vidual Zeeman components of measured lines, using the fit-
The measurements of the field using different elements thus ting capability of the IRAF splot function. The location of the
reveal exactly the same general phenomenon already found in Zeeman components that determine ∆λ can be determined to
the comparison of the hBz i value obtained using lines of H to within ±0.01 Å or less. The corresponding ∆λ was then cal-
the values obtained using metal (primarily Fe) lines in the sin- culated for each measured line. The measurements made from
gle field measurement with FORS1. These results strongly sug- Equation (1) are extremely sensitive, with deviations in ∆λ of the
gest inhomogeneous distributions of various elements, perhaps order of 0.01 Å changing the measured hBi values by as much
combined with significant field structure variation on a small or as 800 G (depending on the spectral line being measured).
medium scale, have large enough amplitude to greatly affect field It is reassuring that the measurements of hBi for both Nd iii
measurements. lines agree within the estimated uncertainties; again, this sug-
gests that the uncertainties are realistic. We notice in the hBi
2.3. Field modulus measurements measurements the same general phenomenon observed in the
hBz i measurements, namely that the field strengths as measured
The ESPaDOnS, HARPS and FLAMES-UVES spectra all show from a single spectrum using lines of different elements are not
spectral lines with resolved Zeeman splitting. We have made the same. The difference is not as striking for hBi as for hBz i, but
measurements of the mean field modulus hBi, which we have presumably points to the same phenomenon: the abundance dis-
compared to those of Manfroid & Mathys (2000) in order to im- tributions of the elements used, and perhaps the field structure,
prove the precision of the rotation period of HD 318107. The are sufficiently inhomogeneous on a scale smaller than a stel-
measurements were made using the splitting observed in the Fe ii lar radius that different elements sample the field geometry in
λ6149 line, and in Nd iii λ5050 and λ6145. The Fe ii λ6149 and significantly different ways. However, it is also remarkable that
Nd iii λ6145 lines were used by Manfroid & Mathys (2000) to the value of hBi varies through the rotation cycle by only about
measure hBi and so to permit direct comparison are used in this ±10 − 15% from the median value observed.
study as well. In addition to these two lines, we also measured
the Nd iii λ5050 line because its Zeeman pattern is very simi-
lar that of Nd iii λ6145. In the stellar spectra this line has three 3. Refinement of the rotation period
clearly distinct groups of Zeeman components, and because the
Landé splitting factors for the lower and upper levels are almost Manfroid & Mathys (2000) used photometric data, and hBi mea-
equal, the actual detailed splitting pattern does not deviate much surements taken between August 1992 and September 1998,
from a simple triplet. Thus Nd iii λ5050 is an excellent line for to test the rotation period of HD 318107 proposed by North
measuring hBi in order to have a check on our measurements (1987). They found that the the 52.4 d period did not fit all
of the λ6145 line. We were able to take advantage of this line their data. They showed that instead a period of P = 9.7085 ±
because of the the large spectral windows covered by our new 0.0021 d fit both the photometry and hBi measurements. When
spectra. we use this period, we choose the zero point for phase to be
The atomic data required to convert splitting measure- JD 2449397.828 which is equivalent to the zero point used by
ments into hBi values were taken from the Vienna Atomic Manfroid & Mathys (2000), but close to the mid-point of their
Line Database (VALD) (Kupka et al., 1999; Ryabchikova et al., hBi dataset. We find that the phases of our recent data have a
1997; Piskunov et al., 1995), except for some lines of Si, which phase which is uncertain by approximately 0.1 cycle relative to
were given Landé factors assuming L-S coupling. To obtain hBi the older measurements (this is the phase uncertainty computed
values, the equation from the one-sigma period uncertainty). Our new hBzi and hBi
data, and the I spectra we wish to model for abundances, can-
∆λ not be accurately phased with the older measurements without
hBi = , (1) improving the precision of the period.
4.67x10−13λ2o z
In principle, the period can be improved both by fitting
was used, where ∆λ is the shift of the σ component from the zero our new hBzi measurements to the five measurements used by
field wavelength, λ0 is the location of the spectral line with zero Landstreet & Mathys (2000), and by optimising the fit of our
field, and z is the effective Landé factor of the line (respectively new hBi data to the well-determined variations of this field mo-
1.17, 1.00, and 1.35 for λ5050, 6145, and 6149). For a more ment as measured by both the Fe ii and the Nd iii line. In practice,
complete discussion, refer to Mathys et al. (1997) or Landstreet however, the variation of hBzi with phase is too sparsely sampled
(2009). The goal was to reproduce the methods used by Mathys by both the older field measurements and our own for compari-
and collaborators as closely as possible so as to have hBi values son to allow us to refine the period. This problem is made more
that could be directly compared to theirs, in order to use the new difficult for HD 318107 by the very different values of hBzi that
measurements to improve the precision of the rotation period. are found at any particular phase by only moderately different
For the Fe ii line, the two sigma components of the Zeeman pat- measurement methods.
tern are separated by as much as 1 Å and are easily measurable. Hence we turn to the new hBi data. Fortunately the numerous
The Nd iii lines have three components and the shifts between the hBi values as measured by Fe ii 6149 Å and by Nd iii 6145 Å vary
central and blue-shifted components were measured to derive the in significantly different ways, and from Manfroid & Mathys
 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107 5

(2000) we have convenient polynomials describing the varia- Landstreet et al., 1989). This programme is designed to com-
tions of hBi with phase for these two elements. Note that in pute the emergent spectrum of a star of specified T eff and log g,
Table 2 of Manfroid & Mathys (2000) the coefficients A of with a specified magnetic field strength and geometry (currently
the polynomials for the two spectral lines have been inadver- characterized as a sum of colinear dipole, quadrupole, and oc-
tently exchanged (private communication with Mathys). Varying tupole, at a specified angle β to the rotation axis). Either a uni-
the period through the range allowed by the uncertainty of form surface abundance distribution or a distribution which is a
±0.0021 d, we find that our new hBi measurements fall very simple function only of latitude in the frame of reference of the
nicely on the appropriate polynomials for a limited range of magnetic axis can be specified. The parameters of the magnetic
periods. The best fit as judged by the χ2 of the fit for Fe ii field model are chosen to match the observed field strength mea-
is P = 9.7087 ± 0.0012 d, while both Nd iii lines agree on surements of hBz i and hBi as a function of phase, generally as
P = 9.7089 ± 0.0007 d, for a conservative global average of described by Landstreet & Mathys (2000). The magnetic model
P = 9.7088 ± 0.0007 d, and an adopted zero point of JD can be further tested by comparison of computed line profiles
2449397.828. This reduces the phase uncertainty of our new data to observed ones in cases such as HD 318107 where line split-
with respect to the data from the 1990s to about ±0.03 cycle. All ting is an important line broadening mechanism. In principle,
phases reported here are on this system. since ZEEMAN computes all four Stokes parameters, compari-
son could be made with observed polarised spectra, but this pos-
sibility is not implemented at present.
The abundance variation with magnetic latitude is currently
A comparison of our new hBi data to the polynomials of specified in the form of uniform abundance values on each of
Manfroid & Mathys (2000) is shown in the upper two panels of one to six rings of equal extent in latitude. The abundance of one
Figure 1. It is clear from this figure that the agreement of our data element at a time, and its variation with latitude, can be varied by
with the previous hBi measurements is very satisfactory, and it is the programme to optimise the fit to a set of spectral line profiles
also obvious why significant phase shifts of especially the Nd iii in one or several spectra taken at various rotational phases. The
data rapidly lead to decreased quality of the fit between the two output is then a coarse model of the abundance distribution of the
data sets, so that the new period is accurately determined. element in magnetic latitude, effectively a very low resolution
The improved period is then used to phase the older hBz i data 1D map. This modelling is repeated for various elements.
shown by Landstreet & Mathys (2000) with the new measure- It is important to understand that the abundance model used
ments. The results are shown in the bottom panel of Figure 1. by ZEEMAN was adopted to describe the kind of abundance
Several points emerge from this figure. First, the new data are variations observed in 53 Cam (HD 65339), a magnetic Ap star
generally consistent with the older measurements provided that in which there are huge differences in large-scale mean abun-
we use the metal line hBzi value from FORS1 at phase 0.91. dances between the two hemispheres surrounding the two mag-
At this phase the Balmer line field strength is about 2500 G netic poles (Landstreet, 1988). In this case, a first order descrip-
higher than the older measurement at almost the same phase. tion that accounts only for abundance variation with magnetic
Secondly, the variations of both hBi and hBz i (from φ = 0.5 co-latitude is an appropriate approximation. As will be discussed
to φ = 1.0) appear to be rather non-sinusoidal. These facts, as below, such large-scale variations does not seem to be a promi-
discussed above, very strongly suggest that either the detailed nent feature of HD 318107, so this particular model is not a
metal abundance distributions or the field geometry, or probably good approximation to the true abundance map. However, us-
both, are complex. This star appears to belong to the small sam- ing it does appear to provide some useful information about the
ple of stars, including HD 32633 (Glagolevskij & Gerth, 2008), nature of the abundance distributions.
HD 37776 (Kochukhov et al., 2010), HD 133880 (Landstreet, ZEEMAN includes a reasonably complete model of LTE line
1990), HD 137509 (Kochukhov, 2006), and τ Sco (Donati et al., formation and radiative transfer in a magnetic field. The stellar
2006) in which the magnetic field may depart in an important atmosphere model is interpolated from a grid of precomputed
way from the generally dipolar topology of the magnetic fields solar abundance ATLAS 9 models. The four equations of radia-
of most Ap stars. (The field and surface He distribution of the tive transfer for polarised light are solved on a grid with 0.01 Å
outstandingly peculiar magnetic He-strong star HD 37776 have spacing. Local line profiles are computed as Voigt profiles based
recently been mapped by Kochukhov et al. (2010), and the re- on atomic parameters taken from the VALD database or rea-
sulting map reveals a quite astonishing degree of complexity.) sonable approximations. Line blending is correctly computed,
Thirdly, the fact that the FORS1 metal line hBzi value agrees by adding (polarised) line opacities due to various lines before
with the older metal line measurement, while the Balmer line solving the equations of transfer. The correctness of many of the
hBz i measurement does not, is certainly consistent with the idea oscillator strengths used has been confirmed by synthesis of a
that the difference between the two FORS1 hBzi values is due to large number of non-magnetic stars (for which the abundance
the combination of a rather complex field distribution and sub- of each element is essentially constant over the surface). Note
stantial differences in the way that the field is sampled by the pre- that the modelling uses atmosphere models computed for solar
sumably uniformly distributed H and the strongly patchy metal abundance throughout; the atmosphere model is not updated as
distributions. It is clear that it would be extremely interesting to the abundance analysis proceeds. With rather large overabun-
obtain V (and probably Q and U spectra) of this star with rather dances of some Fe peak elements found in this analysis, this fact
fine phase resolution, at least of order 0.05 cycle, to make it pos- certainly compromises the accuracy of the resulting abundance
sible to explore more fully the field and abundance distribution values (Khan & Shulyak, 2007).
structure. The models of the field geometry (a low order axisymmet-
ric multipole expansion) and of the distribution of elements over
4. Modelling technique the surface (a few simple discrete rings axisymmetric about the
magnetic axis) are quite schematic. Compared to the sort of map-
To model the spectrum of HD 318107 we use the magnetic ping done by Kochukhov et al. (2004, 2010), who map field vec-
spectrum synthesis programme ZEEMAN (Landstreet, 1988; tor and element distributions over the surface with ∼ 20◦ spatial
6 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107

17000 FeII-6149
MM 2000
16000
Bs (G)
15000

14000

13000

18000
NdIII-6145
17000 NdIII-5050
16000
Bs (G)

15000
14000
13000
12000
11000

6000

5000

4000
Bz (G)

3000

2000

1000

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

Phase

Fig. 1. The top two plots show our new hBi data phased with our best fit period, superimposed on the polynomial fits of
Manfroid & Mathys (2000) to their larger hBi dataset (green dashed curve). In the top panel the second, slightly different, hBi curve
(red dash-dotted line) is the variation of hBi predicted by our adopted field geometry. The top panel data shows our individual mea-
surements of hBi using Fe ii λ6149 (data with error bars are ours, dots are from Manfroid & Mathys (2000)); the second panel shows
our measurements of hBi using Nd iii λ5050 (filled symbols) and λ6145 (open symbols). The bottom figure depicts the hBz i field
variations observed in HD 318107. The data points in green (open squares) are the measurements shown by Landstreet & Mathys
(2000), shifted in phase to the present system; the red points (filled squares) are the new data from Table 1 plus (at phase 0.91) the
metal line hBz i value reported by Bagnulo et al. (2006). The dash-dotted line denotes the hBz i rotational variation predicted by our
adopted field geometry and phase system.

resolution in both surface coordinates, our models, in general, elling we carry out here is quite appropriate for the limited data
represent only a very rough approximation. However, the data available, and provides a first exploration of physical conditions
set required for detailed mapping is far more extensive than the and chemical abundances in the atmosphere of this star, which
few, irregularly distributed, mostly I spectra, obtained with a va- can be used for statistical studies in which each star is charac-
riety of resolving powers, and covering a large range in SNR, terised by only a few numbers, and for identifying stars of such
that are available at present for HD 318107. The type of mod-
 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107 7

great interest that they warrant the extensive investment in ob- structure, thus specifying uncertainties would give an unrealistic
servations needed for detailed mapping. impression of how accurate this model is. However, experimen-
tation with fits to the variations of hBz i and hBi suggest that we
can best reproduce the observed data when the model parameters
4.1. Magnetic field model
are in this region of parameter space.
Before starting abundance analysis, we need to establish an ap- The resulting geometry is shown in Figure 2. The figure is
proximate magnetic field model to use. An important quality of a graphical representation of the oblique rotator model that dis-
this model should be to reproduce reasonably well the phase plays the relationship of the line of sight to the two angles i and
variations of Zeeman splitting, so that the computed spectral line β and to the field axis for the phases 0.4 (left figure) and 0.9
profiles are similar in form to the observed lines, and model fits (right), when the line of sight is respectively farthest from and
are primarily sensitive to chemical abundance. nearest to the visible magnetic pole.
To estimate the inclination i of the rotation axis and the pa-
rameters of a suitable model magnetic field structure, we follow
Preston (1967, 1970) and Landstreet & Mathys (2000).
An estimate of the radius of the cluster member HD 318107 4.2. Chemical abundance model
can be made from the values of log(T eff ) = 4.072 ± 0.02 and
log(L/L⊙ ) = 1.92 ± 0.1 from Landstreet et al. (2007). The stellar With the stellar parameters and a model of the magnetic
radius is then found to be R = (2.22 ± 0.34)R⊙. Our adopted field established, the elemental abundance distributions are ob-
rotation period of P = 9.7088±0.0007 days (Sect. 4) and v sin i = tained using the synthesis program ZEEMAN described above.
7 ± 2 km s−1 (see below) may then be used in the equation ZEEMAN compares the calculated line profiles to the observed
profiles and varies the abundance to minimize the reduced χ2
(v sin i)P
sin i = , (2) of the difference between the model and the observed I spectral
50.6R lines. The programme automatically adjusts v sin i and the radial
where v sin i is in km s−1 and R is in solar units, to obtain i = velocity vR to optimal values at each iteration. For the analy-
37◦ ± 15◦ . This is not a very strong constraint, but provides a sis here, T eff was set to 11 800 K, and log g to 4.2, consistent
useful check to the value of i = 22◦ derived below. with the value 4.22 ± 0.13 determined from M = 2.95 ± 0.15M⊙
From the form of the hBzi and hBi variations with rotational (Landstreet et al., 2007) and the radius found above. Note that
phase, we can get a rough idea of the parameters the multipole because only one ionisation stage is detected for almost all ele-
model of the field will require. The fact that both magnetic mo- ments, and because the shapes of spectral lines are fit only ap-
ments (especially hBz i) vary significantly with phase indicates proximately at best, we do not obtain very strong constraints
that i must be substantially different from zero, as we found on the value of T eff from modelling the observed spectrum.
above. The fact that all the hBz i data have the same sign indi- However, we note that our adopted T eff is consistent with the
cates that, as the line of sight executes a cone around the rotation model fits of Kurucz (1979) when comparing the computed and
axis on the star, it always remains within one magnetic hemi- observed Hβ profiles.
sphere, so i + β < 90◦ . We expect that the sum of the polar fields Atomic data were taken from the VALD. The best-fitting
will be of order 15 kG. From the phase behaviour of hBz i, it ap- value of v sin i was found to be 7 ± 2 km s−1 . The best fit re-
pears that the closest approach of the line of sight to the magnetic quired some experimentation and visual evaluation, as the v sin i
axis occurs at about phase 0.9, even though the value of hBi is broadening is smaller than the Zeeman splitting, so the optimal
near minimum there; apparently the local field strength increases value is best found by looking at how various v sin i values blend
fairly rapidly into the magnetic hemisphere that is observed only the Zeeman components together. The microturbulence parame-
briefly around phase 0.4. ter was set to 0 km s−1 , as the strong magnetic field is expected
To obtain a detailed model the FORTRAN programme to completely suppress the convective motions underlying this
FLDSRCH (Landstreet & Mathys, 2000) was used. The pro- line broadening agent.
gramme takes as input values of hBzi and hBi at four equidistant With the twelve spectra listed in Table 1, we have fairly good
phases. The relative importance of these two kinds of data can phase coverage of the rotation cycle of HD 318107. The largest
be varied by adjustable weights. Since we are particularly con- gap is between phases φ = 0.80 and 0.99. However, some of the
cerned with having a field model that predicts the correct amount HARPS spectra have quite low SNR. Considering that none of
of Zeeman splitting in lines at various phases, we have given a the spectral lines seem to vary very much in equivalent width,
relatively large weight to the hBi data. FLDSRCH calculates hBzi it appears that an adequate sampling of the rotation cycle for
and hBi as functions of phase for a field composed of colinear this first exploratory modelling would be to start with a consid-
dipole (Bd ), quadrupole (Bq), and octopole (Boct ) components. erably smaller number of spectra, focussing on the best avail-
The programme iteratively searches for the values of i, β, Bd , able ones. We have chosen to model only five spectra: the two
Bq , and Boct that best fit the values hBz i and hBi provided as in- ESPaDOnS spectra at φ = 0.995 and 0.258, the UVES spectrum
put. Along with the best fit parameters, the programme provides at φ = 0.390, and two HARPS spectra at φ = 0.615 and 0.803.
calculated values of hBz i and hBi as functions of phase. For lines below 4795 Å, we replace the UVES spectrum with the
The best fit model found in this way (with of course some lower resolution FLAMES-UVES spectrum (φ = 0.453). With
substantial arbitrariness because the data are clearly more com- this choice of spectra to model, we are using the spectra with the
plex than can be reproduced by the simple magnetic field model best resolving power and SNR, we have roughly uniform spac-
adopted) are i = 22◦ , β = 65◦ , and polar field strengths of ing of about ∆φ ∼ 0.2 in phase (and therefore sample coarsely
Bd = +25600 G, Bq = −12800 G, and Boct = +900 G. The the full range of available directions of the line of sight relative
hBz i(φ) and hBi(φ) predicted by this model are plotted in Figure to the magnetic axis), and we reduce the computation time con-
1, and are in reasonably good agreement with observations of siderably.
hBi but reproduce only the general amplitude of the hBzi vari- Spectral lines respond non-linearly to changes in abun-
ations. This model is a poor representation to the actual field dance. Thus fitting several lines of different strengths provides
8 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107

Fig. 2. The geometry adopted for HD 318107. The vertical axis is the rotation axis of the star. The angle between the line of sight and
the rotation axis is i = 22◦ . The angle between the rotation axis and the magnetic field axis is β = 65◦ . For HD 318107, i + β < 90◦
and therefore, as the star rotates, the negative magnetic pole is never directly observed. The left panel depicts an orientation where
the magnetic axis is farthest from the line of sight (φ = 0.4) and the right panel is when the magnetic axis is nearer alignment with
the line of sight (φ = 0.9). The bands used for dividing the abundance distribution are shown as solid lines and are normal to the
magnetic field axis.

a stronger test of the correctness of the model than fitting one but no consistent pattern emerged from modelling different spec-
single line. Where possible, the best-fit abundance distributions tral regions. We eventually concluded that the best fit models
were found by simultaneously fitting multiple unblended lines with abundances which seem to vary strongly with magnetic co-
of a given element within a window of a few 10s of Angstroms. latitude were probably actually fitting some other kind of abun-
The spectral lines modelled are discussed in the sections on in- dance inhomogeneity, probably on a smaller scale, as already
dividual elements. strongly suggested by the very different values of hBz i found at
HD 318107 does not fulfill the conditions for which Zeeman a given phase using lines of different elements (cf. Table 2).
was designed (see Sect. 4). The line of sight remains firmly in a
single hemisphere, and in fact, because the i value of our model Our conclusion was thus that the models found by ZEEMAN
is small, the line of sight only varies in magnetic co-latitude over with strong variation in abundance with magnetic co-latitude
a total range of about 45◦ . Furthermore, the strength of most are unphysical. In fact there is no convincing evidence for
spectral lines vary with rotational phase only rather modestly, strong abundance variations on a global scale. We have there-
suggesting that the pole-to-pole variations of abundance may not fore adopted a much simpler approach to abundance modelling.
be very large. The abundance of each element studied is assumed to be uni-
Nevertheless, we experimented extensively with multi-ring form over the stellar surface. The best fits we obtain to groups
models (i.e. we allowed the abundance to vary with magnetic of lines in various spectral windows of our chosen five spectra,
latitude when we searched for the best fit abundance model for sampling the rotation cycle of the star, of course do not yield
each element). We found that a maximum of three rings should the same mean abundance in all windows. We use the standard
be used (abundances constant on rings 60◦ wide measured from deviation of the abundances found in different windows as a sim-
the visible magnetic pole) in order to achieve convergence to ple measure of the uncertainty of the value of the best fit mean
the same model regardless of the initial abundances assumed. abundance. It will be seen that the uniform abundance values de-
When we ran 3-ring models using different spectral regions (and termined for different spectral windows are roughly concordant,
thus different sets of spectral lines), we found for a few elements and that with this simple assumption we can fit all spectral lines
(particularly Si and Ti) that the best fit was for a model with modelled fairly well at all phases, giving us some confidence that
substantial differences (∼ 1 dex) between one ring and another, the derived average abundances are meaningful.
 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107 9

0.8

0.6

0.4

0.2

-0.2

-0.4
Ti 2

??

Cr 2
??
Nd 3

-0.6 Ti 2
-0.8
4796 4798 4800 4802 4804 4806 4808 4810 4812 4814
Wavelength (A)
Fig. 3. Spectrum synthesis of the region 4795 – 4815 Åusing mean abundances. Black curves are observed spectra, red curves are
models. From top to bottom phases 0.991, 0.258 (both ESPaDOnS), 0.390 (UVES), and 0.615 and 0.803 (both HARPS). The line
of sight is closest to the magnetic pole at phase 0.9, and to the magnetic equator at phase 0.4. Note that some strong lines are still
unidentified.

5. Chemical abundances of individual elements Table 3. Mean abundances of elements studied

Element log(nX /nH ) ± σ log(nX /nH )⊙

He ≤ −2.50 −1.07
O −3.65 ± 0.2 −3.31
Mg −5.70 ± 0.2 −4.40
Because of the large wavelength coverage of the available Si −3.65 ± 0.15 −4.49
spectra of HD 318107, our data are sufficient to obtain first ap- Ca −5.55 ± 0.2 −5.66
proximations to the chemical abundance distribution for O, Mg, Ti −5.05 ± 0.1 −7.05
Si, Ca, Ti, Cr, Fe, Nd, and Pr, and useful upper limits for the Cr −4.50 ± 0.15 −6.36
abundance of He and Mn. The large wavelength coverage also Mn ≤ −6.00 −6.57
means that multiple lines of many elements can be found that Fe −3.00 ± 0.2 −4.50
have a variety of strengths and splitting patterns. Pr −6.35 ± 0.2 −11.28
Nd −6.20 ± 0.15 −10.58
The mean abundance values found for the elements stud-
ied are listed in Table 3, together with the uncertainties esti-
mated from the scatter of fits to (usually two) different spectral
windows. For elements for which only a single line or region
was available, we have estimated the uncertainty as ±0.2 dex.
For comparison, the solar abundances reported by Asplund et al. upper limit to the abundance of helium of at least a factor of 20
(2009) are tabulated as well. The quality of the fits is illustrated below the solar abundance. HD 318107 is definitely a He-weak
with short spectral windows, showing the comparison of uniform star.
abundance models with the five observed I spectra used for the
fitting, in Figures 3 and 4. In each figure the model spectra have 5.2. Oxygen
been computed with the global mean from Table 3 rather than
the local best fits. The lines used and the consistency of the fits The nine weak O i lines at 6155-56-58 Å are blended with lines
are discussed element by element below. of Pr, and are too weak to model usefully. In contrast, the intrin-
sically stronger lines at 7771-74-75 Å are clearly visible and ap-
5.1. Helium pear to be unblended. However, this spectral region is only avail-
able in the ESPaDOnS spectra, and so we have determined the
The strongest lines of He i in the available spectra should be mean abundance using only two spectra. The blend of Zeeman-
the triplet lines at 4471 and 5876 Å. Neither of these lines is split features is rather well fit in the two spectra about 0.25 cycle
unambiguously detected, and fitting these regions provides an apart, and apart from possible non-LTE effects in these lines, the
10 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107

0.8

0.6

0.4

0.2

-0.2

-0.4

Fe 2
Fe 2

Fe 2

Nd 3
Fe 2

-0.6
Si 2

Si 2
-0.8
5038 5040 5042 5044 5046 5048 5050 5052 5054 5056 5058
Wavelength (A)
Fig. 4. Spectrum synthesis of the region 5038 – 5058 Åusing mean abundances. Black curves are observed spectra, red curves are
models. From top to bottom phases 0.991, 0.258 (both ESPaDOnS), 0.390 (UVES), and 0.615 and 0.803 (both HARPS). The line
of sight is closest to the magnetic pole at phase 0.9, and to the magnetic equator at phase 0.4.

O abundance is unambiguously determined, and appears to be 5.6. Titanium

about 0.3 dex lower than the solar value.
Lines of Ti occur throughout the blue spectral region available
in almost all the spectra. We have modelled lines at 4468, 4488,
5.3. Magnesium 4518 Å, and also lines at 4798 and 4805 Å. The two lines in the
The only clean line found of sufficient strength to model is Mg ii longer wavelength window are both well fit at all phases (see
λ4481. This line is reasonably well fit at all phases with a uni- Figure 3), but in the shorter wavelength window the computed
form abundance, and indicates that Mg is about a factor of 20 line is somewhat too strong (and not quite the correct shape)
lower in abundance than in the Sun. for the strong line at 4468 Å and somewhat too weak for the
weak line at 4518 Å, which might be blended with another line
not included in the line list. Nevertheless, synthesis of the two
5.4. Silicon spectral regions yield very similar mean abundances, showing
In the available spectra, the cleanest reasonably strong lines are that Ti is overabundant with respect to the solar abundance by
approximately 2 dex.
the single line at 5041 Å, the doublet at 5055-56 Å, and the sin-
gle lines at 5955 and 5978 Å. The fit to the two longer wave-
length lines is quite good at all phases with a single mean abun- 5.7. Chromium
dance. The fit to the shorter wavelength lines is a little less sat-
isfactory, as will be seen in Figure 4. The fit to these lines is The abundance of Cr has been derived from modelling lines of
roughly equally good at all phases, suggesting that no large-scale multiplets 39 and 44 at 4539, 4555, 4558 and 4565 Å, and lines
abundance variation from hemisphere to hemisphere is detected. of multiplet 30 at 4812 and 4824 Å. All lines are fit reasonably
well with a uniform abundance. The various Zeeman split line
profiles are approximately reproduced (see Figure 3). An over-
5.5. Calcium abundance of almost 2 dex with respect to the solar abundance
Useful lines of Ca ii are found in the ESPaDOnS and HARPS is found. This result is roughly consistent with the variations
spectra, in the near ultraviolet at 3933 Å and in the ESPaDOnS of abundance with T eff found in Ap stars by Ryabchikova et al.
infrared spectra at 8498, 8542, and 8662 Å. The infrared lines (2004).
are all blended with Paschen lines, and so the abundance of Ca
was determined using only the λ3933 line in the two ESPaDOnS 5.8. Manganese
spectra. The abundance of Ca appears to be approximately solar.
A weak line of Ca ii nearly coincides with a strong unidentified The strongest visible lines of Mn are in the region below the
feature at 4799.9 Å, but the computed strength of this line using Balmer jump, between 3430 and 3500 Å. Since these lines are
the abundance derived from λ3933 is negligible (see Figure 3). not included in any of the available spectra, we modelled several
 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107 11

lines of Mn ii between 6122 and 6130 Å. These high-excitation variations of hBz i and hBi with rotational phase. Fitting the pre-
lines are not detected, so they only provide an upper limit of dicted hBi and hBz i variations approximately to the observed
about 0.6 dex above solar abundance. It is interesting that al- variations allows us to determine the parameters of the model
though the even-Z iron peak elements Ti, Cr and Fe are over- magnetic field as well as the inclination i of the rotation axis
abundant by 1–2 dex, this does not seem to be the case for Mn. and the obliquity β of the magnetic field to the rotation axis.
Because the variations of the field moments are rather complex,
and in fact poorly determined for hBz i, the adopted model is only
5.9. Iron a rather coarse first approximation to the actual magnetic field
Iron has many usefully strong lines scattered through the spec- geometry of HD 318107. However, this model is appropriately
trum. We have fit lines in three windows, 4488 – 4542, 4820 – simple considering the limited information available about the
4830, and 5026 – 5037 Å. All the numerous Fe ii lines in these star, in particular the small number of hBz i measurements and
regions are reasonably well fit with an abundance about 1.5 dex high-dispersion polarimetric spectra available.
above the solar abundance. The weaker lines seen in Figure 4 The qualitative form of the field structure is clear from the
suggest that the abundance of Fe might be even a little larger nature of the field data shown in Figure 1. The fact that hBzi does
than this value; in any case, they show the tendency, also found not reverse sign indicates that the line of sight to the star is con-
for Ti, that computed weak lines to be weaker than the observed fined to one magnetic hemisphere, which in turn means that the
lines. This result is roughly with the variation of Fe abundance sum i+β . 90◦ . The fact that the minimum value of |hBz i| is con-
with T eff found in Ap stars by Ryabchikova et al. (2004). siderably smaller than the maximum value shows that the line of
sight does approach the magnetic equator, while the large ratio
of the maximum value of |hBzi| relative to the value of hBi indi-
5.10. Praseodymium cates that the line of sight does approach the magnetic pole. The
Four useful lines of Pr iii are found in the region 6098 – 6196 Å. actual parameter values adopted are discussed in Sec 4.1. The re-
All are fit reasonably well with a mean abundance almost 5 dex sulting magnetic model does not reproduce particularly well the
larger than the solar value. detailed shape of the variations of hBz i, but it works surprisingly
well as a description of the Zeeman broadening (the dominant
source of broadening) of the spectral lines we analysed.
5.11. Neodymium With only a dozen I spectra scattered through the rotation
cycle (but including spectra obtained near times of maximum
Nd iii has lines scattered through the visible window. We have
and minimum hBz i; i.e. when the line of sight is near the mag-
modelled lines at 4796, 4822 (Figure 3), and 5050 Å(Figure 4), netic pole and near the magnetic equator), it is not possible to do
and find a mean abundance about 4.3 dex larger than the solar detailed mapping. However, a preliminary reconnaissance of the
abundance. abundances of several elements, and a first estimation of their
variation over the surface, is possible. We have considered a
6. Discussion & conclusions model that uses three co-axial rings, one around each magnetic
pole and one around the equator, but there is no strong evidence
This paper is the first in what is expected to be a series of reports for strong abundance variations on the global scale. The model
describing modelling of stars from the survey of magnetic Ap finally adopted assumes uniform abundance over the stellar sur-
stars in open clusters of known age. The aim of this project is to face. This model provides the simplest possible quantitative de-
establish a preliminary model of the magnetic field structure, to scription of abundance variations on the star.
estimate the chemical abundances of a number of common el- Assuming uniform abundances over the stellar surface, we
ements, and to extract whatever information is available in our find, as expected, that most elements have distinctly non-solar
data about the surface homogeneity or inhomogeneity of those abundances. The abundance of O is slightly below solar, while
elements by means of simple, rather schematic models whose Mg appears to be significantly lower than the solar abundances.
parameters are derived by fitting synthesised spectra to observa- Ca has nearly solar abundance. Upper limits to the abundances
tions. of He and Mn are possible, the former unambiguously classify-
HD 318107 is a relatively hot magnetic Ap star, with T eff = ing HD 318107 has a He-weak star and the latter suggesting an
11 800 K. It has log L/L⊙ = 1.92, and mass M/M⊙ = 2.95. The abundance no more than 0.6 dex above the solar value. The other
rotation period of the star is 9.7088 d, and it has v sin i = 7 km studied elements (Si, Ti, Cr, Fe, Pr, and Nd) are all present with
s−1 . From its membership in NGC 6405 we know that its age higher abundance than in the Sun. For all of the elements stud-
is log t = 7.80 and that it has completed about 17% of its main ied, the abundances measured over the part of the surface visi-
sequence lifetime. The star has a strong magnetic field, of order ble in our spectra do not appear to vary strongly with magnetic
15 kG globally. co-latitude. However, for some elements a fit to the five spectra
We have used previously published (and modelled) magnetic modelled is significantly better using the three-ring model than
field measurements of Mathys et al. (1997), Manfroid & Mathys with a uniform abundance; we interpret this result as reflecting
(2000), Landstreet & Mathys (2000), and Bagnulo et al. (2006), substantial small-scale abundance variations, as already hinted
together with measurements of our own, to improve the rotation by the very different hBz i values measured in the ESPaDOnS
period of the star to P = 9.7088 ± 0.0007 d, which makes it spectra using lines of different elements.
possible to phase recent magnetic data and spectroscopy accu- HD 318107 appears to be a star which may merit further
rately with previous measurements, and thus to use the full set study. With its extremely large and apparently rather complex
of magnetic and spectral data available to us to model the star. field, and large (and apparently fairly non-uniform) abundance
Our model of the magnetic field is a simple, low-order ax- anomalies, this may be a good candidate for more detailed map-
isymmetric multipole expansion which retains the global topol- ping, which would require a major observing campaign to obtain
ogy of a dipolar field structure. This model is used in the frame- phase coverage in all the Stokes components. The main diffi-
work of the oblique rigid rotator model to describe the observed culties of such a study are the lack of much Doppler broaden-
12 J. D. Bailey et al.: Magnetic Field and Abundance of HD 318107

ing of the lines to provide spatial resolution in azimuth, and the

small value of i which limits the line of sight to one magnetic
hemisphere. But even before this is done, obtaining a few more
spectropolarimetric V observations should help to clarify the na-
ture of the variation of hBz i with phase, enable more detailed
modelling of the field geometry, and help us to understand the
discrepancies between the values of hBz i measured with Balmer
lines and with various metals.
Acknowledgements. JDB, JDL, JS, and GAW are grateful for support by the
Natural Sciences and Engineering Research Council of Canada. GAW acknowl-
edges support from the DND (Canada) Academic Research Program. Astronomy
research at the Open University is supported by an STFC rolling grant (LF).
OK is a Royal Swedish Academy of Sciences Research Fellow supported
by grants from the Knut and Alice Wallenberg Foundation and the Swedish
Research Council. CP acknowledges the support of the project P19503-N16 of
the Austrian Science Fund (FWF).

References
Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481
Bagnulo, S., Landstreet, J. D., Mason, E., et al. 2006, A&A, 450, 777
Donati, J.-F., Howarth, I. D., Jardine, M. M., et al. 2006, MNRAS, 370, 629
Glagolevskij, Y. V. & Gerth, E. 2008, Astrophysical Bulletin, 63, 259
Khan, S. A. & Shulyak, D. V. 2007, A&A, 469, 1083
Kharchenko, N. V., Piskunov, A. E., Röser, S., Schilbach, E., & Scholz, R.-D.
2005, A&A, 438, 1163
Kochukhov, O. 2006, A&A, 454, 321
Kochukhov, O., Bagnulo, S., Wade, G. A., et al. 2004, A&A, 414, 613
Kochukhov, O., Lundin, A., Romanyuk, I., & Kudryavtsev, D. 2010, ArXiv e-
prints
Kupka, F., Piskunov, N. E., Ryabchikova, T. A., Stempels, N. C., & Weiss, W. W.
1999, A&A, 138, 119
Kurucz, R. L. 1979, ApJS, 40, 1
Landstreet, J. D. 1988, ApJ, 326, 967
Landstreet, J. D. 1990, ApJ, 352, L5
Landstreet, J. D. 2009, in EAS Publications Series, Vol. 39, EAS Publications
Series, 1–20
Landstreet, J. D., Bagnulo, S., Andretta, V., et al. 2007, A&A, 470, 685
Landstreet, J. D., Barker, P. K., Bohlender, D. A., & Jewison, M. S. 1989, ApJ,
344, 876
Landstreet, J. D. & Mathys, G. 2000, A&A, 359, 213
Landstreet, J. D., Silaj, J., Andretta, V., et al. 2008, A&A, 481, 465
Manfroid, J. & Mathys, G. 2000, A&A, 364, 689
Mathys, G. & Hubrig, S. 1997, A&AS, 124, 475
Mathys, G., Hubrig, S., Landstreet, J. D., Lanz, T., & Manfroid, J. 1997, A&AS,
123, 353
North, P. 1987, A&AS, 69, 371
Pasquini, L., Avila, G., Blecha, A., et al. 2002, The Messenger, 110, 1
Piskunov, N. E., Kupka, F., Ryabchikova, T. A., Weiss, W. W., & Jeffery, C. S.
1995, A&A, 112, 525
Preston, G. W. 1967, ApJ, 150, 547
Preston, G. W. 1970, in IAU Colloq. 4: Stellar Rotation, 254–+
Ryabchikova, T., Nesvacil, N., Weiss, W. W., Kochukhov, O., & Stütz, C. 2004,
A&A, 423, 705
Ryabchikova, T. A. 1991, in IAU Symposium, Vol. 145, Evolution of Stars: the
Photospheric Abundance Connection, ed. G. Michaud & A. V. Tutukov, 149–
+
Ryabchikova, T. A., Piskunov, N. E., Kupka, F., & Weiss, W. W. 1997, Baltic
Astronomy, 6, 244

View publication stats