19

THE HIPPARCOS PHOTOMETRY AND VARIABILITY ANNEXES

F. van Leeuwen

Royal Greenwich Observatory, Madingley Road, Cambridge, CB3 0EZ, UK

main mission could provide a broadband magnitude ABSTRACT

with relatively high precision. Although in terms of

passband de nition the Hipparcos main mission pho-

A summary of the photometric data obtained by the

tometry was not ideal, it was able to provide a set of

Hipparcos and Tycho missions is presented. The pho-

photometric measurements that is unique in quantity,

tometric signals and their reduction to magnitudes

homogeneity and sky coverage. This large quantity

are describ ed, as well as the passbands and their

of data could also be used for variability investiga-

evolution. The main mission photometry provided

tions of all stars observed and for veri cation of some

on average 110 observations for each of the 118 000

prop erties of ground-based photometric observations.

stars observed during the mission. The accuracies

The Tycho photometric data was primarily used to

for this epoch photometry range from a few milli-

provide much improved colour information for stars

magnitudes for the brightest stars to a few hundreds

in the Hipparcos mission. Colour information was

of a magnitude for the fainter ones. The data cov-

imp ortant in many reduction pro cesses, and ground

erage was very much a function of ecliptic latitude,

based observations were often inhomogeneous, inad-

with the most homogeneous coverage near the eclip-

equate or unavailable.

tic p oles, and the most observations close to ecliptic



latitude 47 . The Hipparcos ep o ch photometry for

The analysis of the Hipparcos photometry to ok place

all stars was investigated for variability and around

in three stages:

970 new p erio dic variables were discovered, as well

as some 7000 unresolved variables including micro

 the reduction of the pro cessed photon-counts to

variables, irregular variables, as well as variables for

magnitudes by CERGA for the FAST consor-

which a p erio d could not b e determined based on the

tium, and the RGO for the NDAC consortium,

Hipparcos data only. Median magnitudes as well as

with system and standard star de nitions done

course variability indicators, derived from the ep o ch

in Geneva;

photometry, are included in the the main astromet-

ric catalogue. All Hipparcos ep o ch photometry has

 the merging and quality control of the reduced

b een made available on the ASCI I CD-ROM, disks 2

photometric data and the preparation of the

and 3. The Tycho ep o ch photometry was of lower ac-

epoch photometry les, done at the RGO;

curacy, but provided useful colour index information.

Only ep o ch photometry for a selection of the brighter

 the variability analysis by Geneva Observa-

stars in the Tycho catalogue has b een preserved on

tory and the Royal Greenwich Observatory,

the ASCI I CD-ROM disk 4. More epoch photome-

the preparation for the atlas of light curves

try can b e obtained through the CDS in Strasb ourg.

Geneva, and the naming of the new variable

Mean magnitudes obtained from the Tycho photom-

stars Moscow.

etry are, when available, included in the main Hip-

parcos astrometric catalogue.

The analysis of the Tycho photometric data was done

by the TDAC consortium in Tubingen,  with indep en-

dent checks on a selection of the data by the RGO

Key words: space astrometry; photometry; variable

and a quality assessment at Geneva Observatory. All

stars.

analysis and merging was sup ervised by the photom-

etry working group, set up by the Hipparcos Science

Team and led by Dafydd W. Evans. Table 1 lists the

main p eople involved in the various tasks.

1. INTRODUCTION

The present pap er gives brief summaries of all the

main asp ects of the photometric analysis. More com- At an early stage in the preparations and develop-

plete descriptions can b e found in the Hipparcos and ment of the Hipparcos mission it was realized that

Tycho Catalogues, ESA 1997, Volumes 1, 3 and 4. the same signal that was used to obtain the astro-

Section 2 presents the photon count signals used as metric results would also contain imp ortant photo-

input for the photometric analysis, and a brief sum- metric information. The Star Mapp er signal as used

mary of their analysis to a measure of stellar inten- in the attitude control for the main mission as well

sity. The passbands b elonging to these signals are as for the Tycho mission could provide magnitudes in

also shown. Section 3 summarizes the main asp ects bands close to the Johnson B and V bands, while the

20

Table 1. Consortia members involved in the photometric

analysis.

Name Aliation Reduc. Merg. Variab.

D.W.Evans NDAC * *

L.Eyer INCA *

J.-L.Falin FAST *

M.Grenon INCA * * *

V.Gromann TDAC *

J.-L.Halwachs TDAC *

F.van Leeuwen NDAC * * *

F.Mignard FAST * *

M.J.Penston NDAC *

M.A.C.Perryman ESTEC

Figure1. An example of a main mission signal for a 5th

1

magnitude star. One sampling periodequals s.

1200

of the reduction of the measured intensities to cali-

brated magnitudes, and the e ect of the changes in

the passband on these reductions. It also explains

a 32/15 s interval, within which between one and

the implementation of the nal passband de nition

10 stars were observed quasi-simultaneously. The

and colour corrections. Section 4 describ es the qual-

amount of observing time sp ent per star dep ended

ity control after the main reductions, the nal cor-

on its brightness and on the comp etition for observ-

rections that were applied to the reduced data and

ing time from other stars. Accuracies on individ-

the merging into one photometric data base, the Hip-

ual observations vary accordingly. In the reductions

parcos and Tycho Ep o ch Photometry Annexes. Sec-

these intervals were referred to as frame transits. In

tion 5 presents the main statistical asp ects of the

the photometric reductions the data from these frame

Hipparcos ep o ch photometry, suchasnumb ers of ob-

transits was used. It to ok around 9 frame transits for

servations, accuracies, and data coverage. Section 6

a star to cross the 0.9 degree eld of view. In the nal

presents some of the techniques used in the variabil-

presentation of the data, only the combined, averaged

ity analysis and a summary of the overall results.

data for a eld of view crossing were preserved. They

Section 7 summarizes the di erent ways in which

provided the ep o ch photometry data.

the photometric data are presented, and nally Sec-

tion 8 gives a summary of results for comparisons

The Star Mapp er signal was obtained from the pas-

with ground-based photometric data.

sage of a star crossing four slits at unequal distances.

Two such sets of slits were present, one set was ver-

tical p erp endicular to the scan direction and one

2. THE INPUT SIGNALS

set was inclined at an angle of 45 degrees. The light

b ehind the slits was split into a B and V channel

T T

and measured with photomultipliers. Figure 2 shows

The main mission astrometric data was obtained

an example of a star mapp er signal in the B channel

T

from the signal of a stellar image moving across a

for a 6.6 mag star. This signal was mo delled using

mo dulating grid, and b eing measured by an image

calibrated single slit resp onse functions, for which the

dissector tub e photon counting device. Only a small

p osition provided information on the transit time and

area of the grid, 30 arcsec diameter, was visible for

the scaling the information on the signal intensity.

the image dissector tub e at any one time. It did,

however, see two parts of the sky simultaneously. An

example of the main mission signal is shown in Fig-

ure 1.

This signal was mo delled using 5 parameters, repre-

senting the mean intensity, the amplitude and phase

of the rst harmonic and the amplitude and phase of

the second harmonic of the mo dulation. The phase

information was used for the astrometric analysis, the

mean intensity and amplitudes of the harmonics for

the photometric analysis. Main mission photometry

derived from the mean intensity is referred to as dc-

photometry, and is for fainter stars a ected by the ac-

curacy of background estimates. Main mission pho-

tometry derived from the mo dulation amplitudes is

referred to as ac-photometry, and is in general ab out

two times less accurate than the dc-photometry, not

sensitivetobackground, but very sensitive to distur-

bances by other images.

Figure 2. An example of a star mapper signal for a

1

s. 6.6 mag star. One sampling periodequals

The analysis of the signal used data collected over 600

21

ters at ESO, La Silla. These data made it p ossible to Preliminary pro les for the Hp, V and B pass-

T T

assign reliable colours and ep o ch photometry to Hip- bands were determined b efore the start of the mis-

parcos measurements of several Mira stars and other sion. On the basis of these pro les a set of standard

long p erio d red variables. In the standard reductions star magnitudes was established. During the mis-

the extreme red stars could not b e used due to their sion, improvements were made to these determina-

variability and the lownumb er of stars available. A tions and much improved standard star selections and

nal recalibration of the colour co ecients for the magnitudes were obtained and used. All main mis-

FAST and the NDAC reductions was carried out by sion data were nally reduced to a passband which

M. Grenon. The evolution of the colour co ecients as closely resembles the actual passband on 1 January

shown in Figure 4 represents the changes in the pass- 1992. Figure 3 shows the nal passbands for Hp, B

T

band it should b e noted here that there were small and V . A full sp eci cation of the passbands can b e

T

di erences b etween the passbands for the two elds of found in Volume 1 of ESA 1997, Table 1.3.1.

view and for the dc and ac comp onents. One e ect

of these changes is that for stars that were reduced

using the wrong colour a drift o ccurs in the nal data.

Such drifts can b e corrected for afterwards, when b et-

ter colours are available. A similar problem can exist

for variable stars with large colour variations. These

colour variations could not be accounted for in the

data reductions, but can b e corrected for afterwards

if required.

Figure3. The Hipparcos and Tycho passbands.

3. THE DATA REDUCTIONS

The data reductions removed the p ositional and

colour dep endences from the measured intensities,

using a set of around 20 000 calibration stars. This

was done in di erentways by the di erent reduction

consortia. Calibrations were done in pseudo-intensity

scale or an equivalent thereof, in order to avoid bi-

ases. In addition, in order to be able to use the

mean intensities for the main mission signal, a back-

ground signal needed to b e determined, in particular

for the reduction of the fainter stars. An actual back-

ground was only measured in the Star Mapp er signal,

but was o ccasionally also available for the main mis-

sion when, due to attitude convergence problems, the

main detector was not p ointed at the intended stars.

Using this information, together with a mo delling

of the galactic and zo diacal light, the background

signal was predicted. A ma jor complication in the

background determinations was caused by the con-

tribution of the radiation from outside the satellite,

which varied strongly b oth intrinsically the satel-

lite op erated around the time of a solar maximum

and due to the orbital p osition of the satellite the

VanAllen b elts were encountered on average once ev-

Figure 4. The evolution of the colour coecients in

ery 5 hours. In order to preserve the p ossibilityto

check the in uence of the background on the nal

the main mission photometric reduction, representing the

magnitudes, all background values as implemented

changes that took place in the Hp passband. The two

by the two reduction groups have b een preserved in

curves in each graph represent the two elds of view. Top:

the photometry annex le.

the dc-photometry derivedfrom the mean intensities, bot-

tom: the ac-photometry derivedfrom the modulation am-

Although colour co ecients were used in the data re-

plitudes. Note the smal l di erencebetween the passbands

ductions, the entire colour dep endence over the mis-

for the two elds of view and between the dc and ac com-

sion was recalibrated afterwards, using a calibration

ponents.

of the passband resp onse for very red stars. This

was made p ossible through a collab oration with the

AAVSO, whose observers measured a selection of

Hipparcos stars particularly for this purp ose. Simul-

taneously some stars were measured with photome-

22

4. THE FINAL CHECKS AND DATA MERGING

Table 2. Standard errors on transit and median magni-

tudes.

Several checks were made on the nal reduction re-

sults as obtained from FAST and NDAC. Remain-

mag

ing p ositional dep endences of the photometric results

Hp Hp Hp Hp B V

ac ac

T T

dc dc

were removed, as well as some as yet unexplained dis-

3 0.003 0.005 0.0004 0.0006

crete small changes in the overall detector resp onse.

5 0.005 0.009 0.0006 0.0010 0.003 0.003

Furthermore, the error estimates on the ep o ch pho-

7 0.008 0.019 0.0009 0.0019 0.008 0.007

tometry were adjusted. These estimates were, due to

9 0.015 0.037 0.0019 0.0039 0.026 0.022

the small number of observations used in determin-

11 0.033 0.072 0.0044 0.0079 0.12 0.12

ing an average, a ected by a Student's-t distribution.

Correlations b etween the two data sets were investi-

gated to o. These correlations exist b etween the resid-

uals of individual measurements for constant stars for

the reduction results as obtained in the two consor-

tia. There are two contributors to these residuals:

the error intro duced by the original photon counts,

and the error intro duced by minor inadequacies of the

data reduction metho ds used. The rst typ e of er-

ror will b e correlated b etween the two sets of results,

while the second error will be largely uncorrelated.

When the correlation between the FAST and the

NDAC residuals is high, the in uence of data reduc-

tion intro duced errors can b e considered small. Fig-

ure 5 shows the correlation co ecients as a function

of magnitude, indicating some calibration problems

for very faint stars resulting from the background

mo delling and for the brightest stars resp onse mo d-

elling.

Figure6. The total number of observations per star as a

function of ecliptic latitude. The rst contour represents

10 stars, every next contour adds 20 stars. The highest

contours represent around 200 stars.

Figure 5. The correlation coecient between the FAST

and NDACreduction results as a function of magnitude.

The Tycho data for the fainter stars required cor-

rections for a bias in the magnitudes that resulted

from the recognition pro cess. A pro cess, called de-

censoring, was develop ed to correct this bias on the

basis of the unrecognized transits. This pro cess is

describ ed in Chapter 9 of Volume 4 of ESA 1997.

Figure7. Distribution of the lengths of gaps as a func-

tion of ecliptic latitude. Gaps longer than 1.5 days were

5. STATISTICS

counted for al l stars. The lowest contour represents 200

gaps observed, increasing by 400 until 1400, then by 600

until 2600, and then by 900 until 8900. The majority of

The statistical asp ects of the Hipparcos and Tycho

gaps is between 10 and 40 days.

data are dominated by the magnitudes of the stars

and by the way the satellite scanned the sky, using

a prede ned scanning law describ ed in ecliptic co or-

6. VARIABILITY INVESTIGATIONS

dinates. The relations b etween accuracies and mag-

nitudes is shown in Table 2. The numb ers of obser-

vations and some parameters describing the distribu-

All main mission photometric data was investigated

tion of those observations over the 3.5 years of the

2

for variabilityby rst examining the statistics of mission are shown in Figures 6, 7 and 8.

23

1982 and Scargle 1989 and the analysis of variance

as describ ed bySchwarzenb erg-Czerny 1989. In ad-

dition to this, a metho d of curve tting using cubic

splines except for Algol-typ e eclipsing binaries, for

which linear splines were used and p erio d ne tun-

ing was develop ed. This also provided minimum and

maximum luminosities, zerop oints and accuracies on

the p erio d estimates see van Leeuwen et al. 1997.

The p erio d searches were in addition supp orted bya

data base of references for photometric or other ob-

servations of stars contained in the Hipparcos Input

Catalogue.

The metho ds used in Geneva are describ ed byEyer

1997. A wider variety of tests was used for the de-

2

tection of variability, using not only a but also,

amongst others, asymmetry of the residuals, checks

for outliers and trends. This made it p ossible to de-

tect a wide range of variabilitytyp es as well as trends.

Figure8. Distribution of the number of gaps as a func-

Any trends detected were checked against the colour

tion of ecliptic latitude. Gaps longer than 1.5 days were

index used in the data reductions. As was explained

counted for al l stars. The contours represent the rela-

in Section 3, errors in the colour index would cause

tive distribution of the gaps, with the lowest contour at 1

trends as a result of the changes that to ok place in the

per cent, and every next contour 2 per cent higher.

passband resp onse. A search for outliers was made, in

order to exclude cases where the variability indicator

was raised only as a result of one or two data p oints

2

which could b e real but did not o er the p ossibility

the weighted residuals. The value was translated

for further investigations. Three metho ds were used

in a probability of o ccurrence. In case no stars were

for p erio d searches:

variable, this probability would be prop ortional to

the numb er of stars observed at a given probability.

The logarithm of the probabilities were accumulated

 Fourier analysis for small amplitude variables

in bins. Figure 9 shows the observed distribution,

Deeming 1975 and Ferraz-Mello 1981;

and, as a dashed line, the exp ected distribution for

constant stars. The excess to the right of this line is

 Stellingwerf 's metho d for large amplitude vari-

due to variability. However, on an individual basis

ables Stellingwerf 1978;

only stars very much to the right of this line with

logpr ob < 8 could b e further investigated for p e-

 Renson's metho d for eclipsing binaries La er &

rio dicity. This amounted to some 12 000 stars.

Kinman 1965, Renson 1980.

The light curves were tted with a weighted linear

regression based on a Fourier series. This ensured

continuous prop erties of the solutions in the phase

diagram. The numb er of harmonics was chosen au-

tomatically from criteria dep ending on the p erio d,

asymmetry, amplitude and noise level of the light

curve. Every light curve was checked and numb ers

of harmonics were adjusted when necessary. This

t provided the nal p erio d, minimum and maxi-

mum light and the reference ep o ch. Figure 10 shows

schematically the metho ds used in Geneva.

The variability results as obtained in Geneva and at

the RGO were regularly compared and in most cases

agreementwas reached. In few cases where this was

not the case or where p erio ds were altogether uncer-

tain, notes were added to the variability annexes. All

references are also included in the printed catalogue

and on the ASCI I CD-ROM. A summary of the de-

2

Figure9. The observed distribution of values ful l line

tections is presented in Table 3.

compared with the distribution for al l stars being constant

dotted line. The right-most bin contains al l values at or

below -15.

7. PRESENTATION OF THE DATA

The metho ds used at the RGO were describ ed byvan

Leeuwen et al. 1997. The choice of metho ds was The photometric data is made available in three

the result of a search through the literature for those forms: the epoch photometry in the Hipparcos and

metho ds b est capable of coping with the rather p o or Tycho Epoch Photometry Annexes; in the form of

data coverage. Two metho ds were used in parallel, of tables and graphs for the results of the variabil-

the discrete fourier analysis as describ ed by Scargle ityinvestigations; and in the form of median values

24

Table 3. Summary of the variable stars investigations.

Number of entries variable or p ossibly variable 11597 8237 new

Perio dic Variables 2712 970 new

Cepheids 273 2 new

 Scu & SX Phe 186 9 new

Eclipsing binaries 98 36 new

Other typ es 1238 576 new

Non-p erio dic and unresolved 5542 4145 new

Not investigated 3343 3122 new

for the Walraven system, the Geneva system and the

UBV system. Only for the Geneva system

Procedure for the analysis Johnson

there were found systematic errors as a function of

right ascension. These comparisons are describ ed in olume 3 of ESA 1997. Statistical Tests Constant Median, width Chapter 21 of V + & Amp_l Outliers

Median, width

CKNOWLEDGMENTS Variable & Amp_l A + Outlier(s)

Trend

y p eople contributed to the analysis of the vari- Sign. Slope Man

+ Median, width

stars. In addition to those mentioned in Ta- Statistical Tests & Amp_l able

Non sign. Slope +

1 I would like to thank I. Zegelaar and J. Lub

slope ble

RR Lyraes, C. Waelkens low amplitude red vari-

S. Meara literature searches, M.B. van

Median, width ables,

wen To czko statistics of variability analysis,

& Amp_l Leeu

ward librarian at RGO. The observations Median & width Micro- and I. Ho

Amplitude Test

vided by the AAVSO were of crucial imp ortance

Amplitude variable pro

for the prop er calibration of the very red stars. The

Sternb erg Institute in Moscow provided at very short

notice names to some 3000 newly discovered variable stars. Mean Period search Periodic + Min, max Fit Iteration(s) Period, epoch Curve, residuals REFERENCES Unsolved Median & width

Min, max Deeming, T.J., 1975, Astroph.&Space Sci., 36, 137

The Hipparcos and Tycho Catalogues, 1997, Median & width Period, epoch ESA,

Amplitude ESA SP{1200

Eyer, L., Hipparcos and the Variable stars, 1997 PhD

Thesis, Geneva Observatory

Figure 10. Schematic representation of the variability

Ferraz-Mello, S., 1981, AJ, 86, 619

analysis methods used in Geneva.

La er, J., Kinman, T.D., 1965, ApJS, 11, 216

van Leeuwen, F., Evans, D.W., van Leeuwen To czko,

M.B., 1997, in `Statistical Challenges for Mo dern

in the main catalogue, supplemented in some cases

Astronomy I I', ed: E. Feigelson & G.J. Babu, in

by ground-based information on star colours. The

press

epoch photometry contains information on the back-

ground level applied, a ag indicating any problems

Renson, P, 1980, A&A, 63, 125

that were considered may exist, and the p osition of

Scargle, J.D., 1982, ApJ, 263, 835

the other eld of view with a p ointer to a le giving

Scargle, J.D., 1989, ApJ, 343, 874

p ossibly disturbing ob jects noted at that p osition.

Schwarzenb erg-Czerny, A., 1989, MNRAS, 241, 153

Stellingwerf, R.F., 1978, ApJ, 224, 953

8. COMPARISONS WITH GROUND{BASED

DATA

The global nature of the Hipparcos and Tycho pho-

tometry and the accuracy of in particular the Hip-

parcos photometry as well as the large number of

measurements make this set of photometric data an

almost ideal reference data base for ground-based

photometric systems. Such comparisons were done