Index du Forum

 FAQFAQ   RechercherRechercher   MembresMembres   GroupesGroupes   S’enregistrerS’enregistrer 
 ProfilProfil   Se connecter pour vérifier ses messages privésSe connecter pour vérifier ses messages privés   ConnexionConnexion 


Poster un nouveau sujet   Répondre au sujet     Index du Forum -> Actualités -> NewPolitique&Social
Sujet précédent :: Sujet suivant  
Auteur Message

Hors ligne

MessagePosté le: Lun 4 Juil - 07:35 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Lawrence Berkeley Laboratory
Physics Division
S. Perlmutter (PhD. Thesis)
November 1986
TWO-WEEK LOAN COpy This is a Library Circulating Copy
which may be borrowed for two weeks.
Prepared for the U.S. Department of Energy under Contract DE-ACG3·76SF00098
An Astrometric Search for a Stellar Companion to the Sun
Saul Perlmutter
Lawrence Berkeley Laboratory and The University of California, Berkeley
Berkeley, CA 94720
Thesis submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy in Physics
25 November 1986
An Astrometric Search for a Stellar Companion to the Sun
Saul Perlmutter
A companion star within 0.8 pc of the Sun has been postulated to explain a possible 26 Myr periodicity in mass extinctions of species on the Earth. Such a star would already be catalogued in the Yale Bright Star catalogue unless it is fainter than mv = 6.5; this limits the possible stellar types for an unseen compan- ion to red dwarfs, brown dwarfs, or compact objects. Red dwarfs account for about 75% of these possible stars. We describe here the design and development
of an astrometric search for a nearby red dwarf companion with a six-month peak-to-peak parallax of >2.5 arcseconds. We are measuring the parallax of 2,770 candidate faint red stars selected from the Dearborn Observatory catalogue.
An automated 3D-inch telescope and CCD camera system collect digitized images of the candidate stars, along with a 13'X16' surrounding field of back- ground stars. Second-epoch images, taken a few months later, are registered to the first epoch images using the background stars as fiducials. An apparent motion, rna' of the candidate stars is found to a precision of CTma :::::: 0.08 pixel :::::: 0.2 arcseconds for fields with N fiducial > 10 fiducial stars visible above the back- ground noise. This precision is sufficient to detect the parallactic motion of a star at 0.8 pc with a two month interval between the observation epochs. Images with fewer fiducial stars above background noise are observed with a longer interval between epochs. If a star is found with high parallactic motion, we will confirm
its distance with further parallax measurements, photometry, and spectral stu- dies, and will measure radial velocity and proper motion to establish its orbit.
We have demonstrated the search procedure with observations of 41 stars, and have shown that none of these is a nearby star.
I would like to thank the people who have contributed to this research work. My research advisor, Rich Muller, has shown me that it is possible to do top- quality physics without narrowing one's range of interests. The Nemesis hypothesis, on which this work is based, is but one example of the fruitfulness of his approach. Luis Alvarez, his thesis advisor, has taught me the ingredients necessary to get results in science: perceptive questions and a determination to find the answers.
My thesis committee members have played roles in my graduate career well before I began my thesis. Buford Price gave me my first stimulating introduction to forefront physics research. Walter Alvarez has offered me glimpses of a different science frontier, the geological past.
My work builds on the efforts of the entire Nemesis/Supernova Search team within the Lawrence Berkeley Laboratory astrophysics group. Carl Pennypacker has led these searches with optimistic enthusiasm. Jordin Kare and Shane Burns lay the groundwork for the automated telescope, CCD, and image analysis system in their doctoral work, and Roger Williams wrote much of the original software. Frank Crawford analyzed the optics and the CCD performance. When the day-
to-day work threatened to degenerate into pure computer programmmg, James Graham helped restore the focus on physics. Peter Friedman has shared with me many hours of work and discussion throughout our graduate careers.
I thank Terry Mast for his insightful advice on the data analysis, and partic- ularly for his constructive criticism of the drafts of this thesis.
Throughout my work, my parents, Daniel and Felice Perlmutter, have pro- vided unceasing support and encouragement.
This work was supported in part by the Department of Energy under contract DE-AC03-76SFOOOgS.
Table of Contents
1. Introduction 1
2. Choice of Search List
2.1. Astrophysical Constraints on a Solar Companion 6
2.2. Companion Star Search 2.3. The Search List
3. Data Collection
3.1. Overview 3.2. Generating an Observation List 3.3. Observations 3.4. Future Improvements
4. Analysis and Software
4.1. Overview 4.2. Preprocessing 4.3. Image Analysis 4.4. Field Comparisons 4.5. Postprocessing
5. Results and Discussion
5.1. Calibration Results 5.2. Error Analysis 5.3. Search Strategy
6. Conclusion 7. References
9 00........................... 12
15 16 20 28
31 32 36 40 43
45 49 56
61 62
Binary Stars
1. Introduction
Studies of binary stars suggest that more than 84% of solar-type stars are in binaries (Abt, 1983). The sun has generally been considered one of the excep- tions, since the past century's cataloging of nearby stars has failed to discover a companion star bound to the sun. Occasional articles have proposed that a nearby star may have escaped detection or identification as a solar companion. Van de Kamp (1982), for example, in discussions on wide-binary pairs suggests that a wide-binary companion could have been missed in the current catalogues. In response to the new observational capabilities of IRAS and the Hubble Space Telescope, Reynolds, Tarter and Walker (1980) also discuss the possibility of finding an unseen faint companion. These proposals have not :fet been followed through; there has been no concerted all-sky search for a solar companion.
Recently, evidence from the more earthbound fields of paleontology, geology, and geophysics has excited renewed interest in a solar companion. The back- ground was set by the proposal of L. Alvarez et ai. (1980) that the Cretaceous- Tertiary extinction was caused by the impact on the earth of a large comet or asteroid. Since comets and asteroids are known to have a higher concentration of iridium than the Earth's crust, this theory explained the anomalously high con-
centration of iridium in the clay layer that marks the Cretaceous-Tertiary boun- dary.
The impacts could cause the mass extinction through a number of possible mechanisms. L. Alvarez et ai. (1980) suggested that the impact would raise enough particulate matter into the atmosphere to block the sunlight reaching the earth. This would interrupt photosynthesis in the biosphere, the food chain on earth would be disrupted, and entire families of species would become extinct. Toon et at. (1982) pointed out that the opaque atmosphere would also produce sub-freezing temperatures on the continents, causing extinctions of many land animals. This scenario has been discussed in recent "nuclear winter" analyses of the outcome oi a large-scale nuclear war (Turco et at., 1983).
Since Alvarez et at. first proposed the impact explanation fOf the Cretaceous-Tertiary extinction, substantial evidence has been collected supporting the correlation between impacts and mass extinction events. The high concentra- tion of iridium at the extinction boundary has been found to be distributed
world-wide. Iridium has also been found in concentrations above the background at other extinction boundaries besides the Cretaceous-Tertiary. At many of these boundary sites there is shocked quartz, which is found at known impact sites, and mineral spherules (probably altered microtektites), which could have originated as droplets of impact melt. Alvarez (1986) reviews this and other evidence.
Periodic Extinctions and a Solar Companion
The theoretical situation became more complicated when Raup and Sep- kokski (1984) reported evidence for a 26 million year periodicity in the fossil record of mass extinctions on the earth; the Cretaceous-Tertiary extinction fit the phase of the periodicity. This periodicity was difficult to reconcile with the come- tary impact theory, since impacts were considered rare and random occurrences. In response to this theoretical difficulty, Davis, Hut and Muller (1984) and Whit- mire and Jackson (1984) independently suggested that a solar companion could cause periodic comet showers on the earth. Other explanations for the periodic extinctions were also proposed (Whitmire and Matese, 1985; Rampino and Stoth- ers, 1984), but they fail to be supported by further analysis (Muller, 1985).
The solar-companion model of Davis, Hut and Muller proposes a star in an orbit about the sun with an eccentricity e >0.7, and a semi-major axis a chosen to give a period T = 26 Myr:
a [AU] = (T [yr]f/3 = 8.8X104 AU = 0.43 pc (1)
where the units have been chosen such that Kepler's constant k = T 2/a3 = l. Every 26 Myr, when this star passes through perihelion, its gravitational field ran- domizes the motions of the 1011_1013 comets in the Oort cloud, currently thought to orbit the sun at distances up to _105 AU. Hut (1985) has shown using numeri- cal simulations that the Companion star mass must be >O.06M0 to perturb the motions of the Oort cloud comets.
With each close passage of the Companion, some of the perturbed comets will enter the inner solar system, which is usually kept clear of comets by the sweeping action of Jupiter's gravitational field. Davis, Hut and Muller estimate that during each of these periods of cometary presence in the inner solar system an average of 2-25 comets will hit the earth, causing periodic mass extinctions. (The name they suggested for the star, Nemesis, seems appropriate.)
Other Evidence
The discovery of periodicity in the dates of craters on the earth has provided the strongest supporting evidence for periodicity. Alvarez and Muller (1984) found a 28.4 ± LO Myr periodicity that corresponds to Raup and Sepkokski's extinction periodicity within the errors of the geologic time scale. Rampino and Stothers (1984) and Shoemaker (1985) used slightly different sets of craters and found respectively a 31 Myr periodicity and a 30 Myr periodicity. These results differ mostly in the phase of the most recent events, since the earlier time scale is uncertain enough to allow cycles of differing frequency to match. It is also important to note that if a Companion Star's orbit is the ultimate cause of the periodicity, then the comet showers are expected to be only approximately periodic. Hut (1984) has shown that the Companion's orbit is expected to be per- turbed by passing stars and the Galactic tidal field, and thus the period between
impacts should vary by 100/0-20%.
This Search
The evidence supports the claim that periodic impacts on the earth have caused mass extinctions. It does not show that the periodic comet showers are caused by the close passage of a solar companion. The most incontrovertible evi- dence would be the discovery of a companion star with the appropriate orbital period. We are now searching for such a companion star, and hope to finally dis- cover if the Sun is a member of a binary system. We describe here the design and development of an observational program to measure the parallax, and hence the distance, of candidate stars in the northern hemisphere.
2. Choice of Search List
2.1. Astrophysical Constraints on a Solar Companion For a star orbiting the sun with semi-major axis a = 0.43 pc and eccentricity
e = 0.7, the maximum distance from the sun will be r max = (l+e)a = 0.73 pc (2)
Using numerical simulations Hut (lQ85) found that the Galactic tidal field slightly increases the a and rmax necessary for a T = 26 Myr orbit. He obtained r max = 0.8 pc. The six-month peak-to-peak parallactic motion at this distance is
Bp _ p [arcsec] = 2 = 2.5 arcsec (3) rmax[Pc]
(The conventional definition of parallax is half of this peak-to-peak motion: 1r =Bp_ p/2 =1.25 arcseconds.) This is the minimum motion we could expect for a Companion, obtained if we are presently halfway between two perihelion pas- sages, as evidenced by mass extinctions. Since this phase is not certain, the Com- panion could be closer and have a higher parallax. Using this limiting value for the Companion's distance, we can ask how current astrophysical knowledge con- strains the attributes of a solar companion.
Proper Motion Surveys
Surprisingly, little information can be gleaned from the catalogues of nearby stars, because these stars are mostly discovered in surveys for large proper
motion. For a star bound to the sun with a 26 Myr periodicity, the proper f "t·lon II at .. u7;11 hO
A.fi_v ,,.,..,, u, max n 11 UV
v(rmaJ Jl( rmax) = ---:.--:::;~
r max
~ 0.01 arcsec/yr
This is well below the Jl = 0.18 arcsec/yr limit set by the Luyten Catalogue com- plete sky survey (Luyten, lQ7QjlQ80). The proper motion is larger at closer dis- tances, but only exceeds the Jl - 0.18 arcsec/yr limit when the companion is near perihelion. The parallactic motion, Bp_ p = 2.5 arcseconds, for a companion star at rmax = 0.8 pc will be above the Luyten Catalogue limit. However, the proper motion surveys use observations collected at the same time of year for both epochs, and therefore are not sensitive to parallactic motion.
The Yale Bright Star Catalogue
Even at rmax' the Solar Companion would be the closest star in the sky, and the brightest star of its spectral type and luminosity class. The Yale Bright Star Catalogue (Hoffieit, 1982) is complete to apparent magnitude m v = 6.5, and no star in the catalogue has the parallax, proper motion, or spectroscopic parallax consistent with the orbit of a Solar Companion. This constrains a Solar Compan- ion to be among a small set of stellar types. Figure 1 shows the Yale Catalogue's
brightness limit on the H-R diagram. Any star within 1 pc would be contained in the Yale Catalogue if its stellar type lies above the line at apparent magnitude
-,- -.0. -.,.
-3I-- 0~
3I-- , 6I--
0 -0

-:. -. •- - 
'Y 'Y
,- - . .-
• •
M-( _6 I--
'V 'V
00 Ao S
'V \] 'V
r I
12I-- 15I--
Red dwarfs ..-
Cl. ... In
White dwarfs
I wd -
'V T

05 80 AO FO GO KG MO
Figure 1
The Hertzsprung-Russell diagram for the 100 nearest stars and the 100 closest stars (Schaifers and Voigt, 1982). The Yale Bright Star Catalogue's completeness limit is indicated by the horizontal line.
t. I
mv = 6.5, or absolute magnitude Mv = 11.5. Thus a companion star, if it exists, must be a red dwarf, white dwarf, or one of the stellar types not shown on Figure 1, the brown dwarfs, neutron stars, and black holes.
2.2. Companion Star Search
Red Dwarf Stars
Of these possible stellar types, we have chosen to search for a red dwarf com- panion. Much work has already been done by the Dearborn Observatory's sur- vey for faint red stars (Lee et al, 1Q43, 1Q44, 1Q47; Nagy, 1979), which has cataloged all of the red stars in the northern hemisphere to a limiting magnitude for late M dwarfs of mv ::::::: 13 (Stevenson, 1Q84). (No comparable survey has been done for the southern hemisphere.) As can be seen in Figure 1, this Dearborn Observatory Catalogue would include any red dwarf star within 1 pc. This catalogue does not distinguish, however, between red dwarf and red giant stars,
and thus the problem is reduced to finding the nearby dwarfs among the many distant giants. We have developed an astrometric parallax search to find a nearby star if it is in the Dearborn Observatory Catalogue. This search is clearly only a first attempt to find a Companion with the highest chance of success for the least effort, since the Companion could be in the southern hemisphere and/or may not be a red dwarf.
Other Stellar Types
There is no a priori expectation that a solar companion should be of a cer- tain stellar type. A search of just the red dwarfs does, however, include most known stars. Table 1 shows the Luminosity Function of Bahcall and Soneira (1980), which gives the number of stars per cubic parsec in each magnitude inter- val. We find that -54% of all known stars are red dwarfs below the Yale Bright Star Catalogue Mv = 11.5 limit. Only -4% are stars massive enough to become white dwarfs, neutron stars, or black holes in the T (3) ~ 4.5 X10g year lifetime of the Sun. Extrapolating back in time 4.5 billion years increases the proportion of massive stars only slightly to -5%. Thus there are about 0.1 times as many com- pact object (white dwarf, neutron star, or black hole) candidates as red dwarf candidates..
The proportion is less clear for brown dwarfs, which have not yet been found, but are postulated to exist with masses <0.08 M(3). These stars are not massive enough for hydrogen burning in the core, and thus are expected to have low luminosities. Bahcall (1986) has suggested that the missing mass in the galaxy may consist of unseen brown dwarfs. If all of the missing mass, Pmissing = 0.10 ± 0.05 M(3) pc-3, is in the form of brown dwarfs with mass less than 0.08 M(3) , but greater than the 0.06 M(3) necessary for a Companion to per- turb the Oort cloud, then the brown dwarfs would account for >95% of the stars that could be a Companion. This is not likely, since there is no reason to believe that the bulk of brown dwarfs would not have ffi::lSSeS closer to 0.001 M(3) , partic-
Luminosity Function Mass
Lifetime on Main Sequence
log T (yr) MS
M ¢(Mv) M/M
Table 1
The luminosity function for all stellar types in the Galaxy (after Shapiro and Teukolsky, 1983)
-6 -5 -4 -3 -2 -1
0 1 2 3 4
5 6 7 8 9
12 13 14 15 16
(stars pc-3 mag-1)
1.49 X 10-8 7.67 X 10-8 3.82 X 10-7 1.80 X 10-6 7.86 X 10-6 3.07 X 10-5 1.04 X 10-4 2.95 X 10-4 6.94 X 10-4 1.36 X 10-3 2.26 X 10-3
117.5 6.42 63.1 6.50 33.9 6.58 18.2 6.84
9.8 7.19 5.2 7.68 2.8 8.36 2.3 8.62 1.8 8.93 1.5 9.24 1.2 9.60
3.31 X 10-3 4.41 X 10-3 5.48 X 10-3 0.63 6.52 X 10-3 0.51 7.53 X 10-3 0.41 8.52 X 10-3 0.33 9.54 X 10-3 0.27 1.06 X 10-2 1.17 X 10-2 1.29 X 10-2 1.41 X 10-2 1.41 X 10-2 0.09
1MS <10
NJv below Yale Catalog
0.95 9.83 0.78 10.28
0.21 0.17 0.14 0.11
ularly if the Luminosity Function continues, as in Table 1, to rise with decreasing mass. If we estimate that there is approximately the same density of stars in the 0.06-0.08 M0 mass bin as in the 0.08-0.10 M0 mass bin of Table 1, we find that there would be --0.23 as many brown dwarf Companion candidates as red dwarf candidates. Considering red dwarf, compact object, and brown dwarf can- didates, our search of red dwarfs thus includes (1 +0.1 +0.23)-1 = 0.75 of the possible candidates in the northern hemisphere.
2.3. The Search List
We have selected stars from the Dearborn Observatory Catalogue of Faint Red Stars to be the candidates for the search. Only stars that He below the mv = 6.5 limit shown on Figure 1 are placed on the Search List. This limits us to a subset of the red dwarfs, those with spectral types cooler than M4. The Dear- born Observatory Catalogue contains 5,518 stars of types M4-M9. This is expected to be a complete list of the hydrogen-burning cool stars that are candi- dates within 1 pc in the northern hemisphere.
Magnitude-Color Relation
We use the "main sequence" correlation between spectral type and absolute magnitude (shown in the first two columns of Table 2) to further trim our Search List. Many stars within the M4-M9 subset are too faint to be stars of their spec- tral type within 1 pc of the Sun. Taking into account the -0.5 magnitude uncer- tainty of the Dearborn Catalogue, we eliminate 2,748 stars using the brightness
criterion listed in the fourth column of Table 2. Figure 2 shows which stars these cuts remove from the Dearborn Observatory Catalogue. The Companion Search thus begins with a Search List of 2,770 candidate stars.
Table 2
Cuts used on the Dearborn Observatory Catalogue
Spectral Type
M(a) mv v at 1 pc
mv cut number used cut
number remaining
8.0 2005 440 10.0 743 936 0 933 0 395 0 63 M9+ 03
Totals: 2748 2770 (a) Mv from Schaifers and Voigt (1982)
M4 M5 M6 M7 M8
11.3 6.3 12.3 7.3 13.5 8.5 14.3 9.3 16.0 11.0
E :>8 m
10 ~ '---
'4 + - - - - . . . . . L . . - r - . . . . . L . . - - . . . . . L . . . - . . . . . L . . - - - - , - - - - - , - - - - - - r - - - - - - , M4 M5 M6 M7 M8
Spectral Type
Figure 2
Scatterplot of Dearborn Observatory Catalogue stars of spectral types M4-M8. Boxes show faint stars deleted from the Search List.
3.1. Overview
3. Data Collection
The goal of the search is to discover any star on the Search List with a paral- lactic motion greater or equal to 2.5 arcseconds peak-to-peak. The search pro- cedure is as follows: Every night with available telescope time, we observe a sub- set of stars from the Search List. Images are collected with a CCD camera, digi- tized, and recorded on magnetic tape. These tapes are later read into the lab computer and analyzed. If any image is overexposed, underexposed, out of focus, or otherwise unusable the Search List is updated with corrected observation parameters (e.g. exposure time) for a new observation. For usable images, the analysis program finds the stars in the image, both the Search List star and the background stars, and writes the coordinates and description of each star into a data base.
When a sufficient number of months has passed for us to be able to detect parallactic' motion, we re-observe the stars, and compare the first epoch coordi- nates to the second epoch coordinates for each star field. Using the background
stars to "register" the two images with each other, we find the overall offset, rota- tion and magnification changes between the two images. Taking these changes into account we calculate the apparent motion, ma, of the Search List star. Any star with a small ma is removed from the Search List and no longer considered a Companion Star candidate. Any star with significant ma is subjected to close
scrutiny, reobservation, and reanalysis as a possible Companion candidate. Fig- ure 3 charts this process for the second-epoch observation.
The Companion Star Search shares much of its procedure and equipment with the Berkeley Supernova Search. Detailed descriptions of the telescope, auto- mation, camera, and data path can be found in Burns (1985), Kare (1984), and Kare et al. (to be published). The remainder of this chapter will describe the details of the Companion Star Search procedure. The data analysis will be dis- cussed in the following chapter.
3.2. Generating an Observation List
For each night's observations, we select stars from the Search List to create an Observation List. We take a number of factors into account in making this
Atmospheric Dispersion
The stars on our Search List are red, while the typical background stars in our exposures are hotter, with a black-body peak in the yellow part of the spec- trum. If the two epochs' images are taken at different zenith angles, we will measure a non-zero ma for a Search List star with respect to the background stars, because of the atmospheric dispersion of the different colors. We attempt to observe both epochs at the same zenith angle, but it is difficult to schedule the second-epoch observations at just the times of night when the zenith angles agree.
Search List
Generate Observation List
Observation List
Observe stars at telescope
Data tape
Analyze images
No > -------... observation
Reobservation and reanalysis; follow-up studies of candidate.
Figure 3
Remove star --1.-1 from Search
Flowchart for second-epoch observation and analysis.
We therefore try to take the exposures at small zenith angles to minimize the dispersion. For the Observation List, we select stars that will be near the zenith at the time of both the first and second epoch observations.
Parallactic Molion
Even if all of the stars on the Search List were at 0.8 pc, they would not all appear to move the same distance between two arbitrary dates of observati-on. We can maximize the parallactic motion by choosing observation dates for a given star according to its location on the "celestial sphere." In particular, stars which are at low ecliptic latitudes-coplanar with the earth's orbit-appear to trace out a sinusoid in Right Ascension versus time, making it necessary to choose observation dates with care. They reach their maximal excursions when they are
just visible at dusk and at dawn, but move fastest at the midpoint between these two dates, when they are directly overhead at midnight.
Stars at high ecliptic latitudes on the other hand can be observed most of the year, since they trace out circles in the sky annually; this motion can be measured at any two points sufficiently far around the circle. In selecting candidates for the Observation List, we therefore first choose stars that are at low latitudes and are about to exhibit their greatest parallactic speed in the following few months. The remainder of a night's Observation List can then be filled with the high-latitude stars, for which the observation dates are less important.
Dewar Configuration
The telescope cannot be pointed above 58 0 in declination unless the liquid- nitrogen dewar (used to cool the CCD) is mounted at an angle allowing it to clear the telescope's supporting fork. At this angle the dewar cannot be completely filled with liquid nitrogen, so it is usually mounted in a more perpendicular posi- tion. Thus depending on how the dewar will be mounted on a given night, stars
over 58 0 are either preferentially included or excluded from the night's observa- tion list.
Seeing Limitations
Some selections are based on the seeing at the time of observation. If the seeing is poor, stars are spread over more pixels and therefore brighter stars, whicn would otherwise saturate the image, can be safely added to the list. We prepare alternate observation lists for different seeing conditions.
Observation List Contents
An Observation List contains the following data for each star:
a) Dearborn Observatory Catalogue number b) Coordinates: RA and Dec. c) Spectral type: M4, M5, ..., M8 d) Apparent magnitude, m v
e) Exposure time (seconds) f) Sequence number on the Observation List
3.3. Observations
The search employs the 3D-inch Ritchey-Chretien reflector telescope at the
University of California's Leuschner Observatory (longitude 122.1570 37.918 0 ) . Dr. R. Treffers, of the Berkeley Astronomy Department, has extensively automated the Leuschner facility, and the Lawrence Berkeley Laboratory's astro- physics group has improved the telescope speed and pointing accuracy, and the on-line computing resources. While minor refinements and major upgrades are continually changing the operation of the telescope, the current system is sufficiently evolved to carry out the Companion Star Search. Figure 4 shows a block diagram of this system.
For the Companion Star Search, the goal of the data collection is to image each star field on the CCD camera and record the image. (The plate scale is 2.4 arcseconds per CCD pixel. One pixel thus roughly corresponds to the typical see- ing at the telescope, 3 arcseconds, and to the parallactic motion of a companion at 0.8 pc, ()p_p = 2.5 arcseconds.) An IBM XT controls most of the operations, making it possible to collect hundreds of such images each night. Using the observation parameters for a star from the observation list, the XT points the telescope, and opens and closes the shutter. It then reads out the CCD camera's digitized data onto videotape, adding bookkeeping information describing the observation onto the tape's audio track. This process is repeated until the obser- vation list is exhausted.
With ideal camera, optics, and software, this simple repetitive procedure
Dec Drive Moters
Flasher & Shutter Control
Amplifier & NO converter
Control Computer
Leuchner 30" T elescope
audio track
Display IBM
VCR tape 00
to lab
Video Display
Figure 4 Block diagram of current observatory system.
would suffice to achieve the precision we desire for our final results. In practice, some additional steps are required both before an observation list is begun, and during the observation of each star.
Telescope Pointing
Since the optics that focus the light from the telescope onto the CCD are not ideal, some vignetting and pincushion distort the final image. The vignetting reduces the amount of light at the top of the CCD frame to less than half of the light in the middle of the frame. The pincushion distortion moves images radially by ~r= por3, where the constant of proportionality for our optics is Po:::::: 6.5XIO-8 pixeis-z. These distortions shift the apparent positions of stars, a larger shift occurring for stars that lie further from the center of the image. We cancel this effect by pointing the telescope so that the position of the second- epoch image on the CCD's pixels matches the position of the first-epoch image.
The telescope pointing algorithm is initialized while the telescope is imaging a fiducial star of known coordinates on the center of the CCD's pixel array. To locate the fiducial star image within a few pixels of the array center, we send the digitized image to an IBM AT computer, which has a Tecmar graphics display card. Here we can display the image on a composite-video monitor, and manually move a superposed cursor to determine a star's position to within one pixel. For the first-epoch observations, a catalogued bright star with well-determined coordi- nates is used as a fiducial. It is placed within three pixels of the center pixel for
the initial telescope pointing, and in the course of the night the pointing will stay within about 6 pixels RMS. For the second-epoch exposures, a star seen in a matching first-epoch image is used as the fiducial.
"Flash£ng" the CCD
The CCD detector is itself not an ideal instrument. It operates by collecting photoelectrons that photons knock out of its substrate into an array of 320X512 potential wells. The amount of charge collected in each potential well thus represents the amount of light that has fallen on that pixel. A bucket brigade of charge transfers brings each packet of charge down the columns and across the bottom row to a single amplifier at one corner of the array. The amplifier meas- ures the charge in each pixel, and the result is then digitized by an analogue/digital (A/D) converter, with an overall gain of ~28 photoelectrons per A/D count.
The charge transfers in the bucket brigade are not perfectly efficient, and some fraction of the charge gets left behind at each transfer. For a star image, this causes a tail of inefficiently transferred charge to grow behind the peak as it moves across the CCD. A good CCD will have an inefficiency of less than c=lO-4, so that after the 832 transfers from the farthest corner of the CCD, the amount of charge left behind in the tails is only
1 - (1-£)832 ::::::: 0.08 Measurements of the CCD, RCA model SID53612XO, by Burns (1985) yielded
transfer inefficiencies of
1.6XlO-4 < f. < 5XlO-4
-88°C> TCCD > -130 ·C where TCCD is the operating temperature of the CCD.
More recently, we have found that the transfer efficiency of the CCD also depends strongly on the amount of charge being transferred. It is apparently harder to transfer the first electrons out of the potential wells, an effect which has been referred to as "fat zero" and "sticky bucket." The effect can clearly be seen in Figure 5, which shows the shapes of a bright star image and a dim star image. The dim star is in the high f. regime, and thus has grown a tail of inefficiently transferred charge as it was transferred from (Row 130, Column 254) to the amplifier at (0,0). The bright star is in the low f. regime, and thus has a round profile characteristic of efficient charge transfer.
A dim star with a tail will appear to have a centroid shifted toward higher coordinates. This effect can be seen in Figure 6(a), which shows results of meas- urements made in the lab, with a simulated star image held at one spot on the CCD for each curve. The apparent centroid for a star image is plotted against the brightness of the star, and the different curves show the effect of placing the star image at different distances form the amplifier corner. The apparent centroid of the star clearly shifts more as the star becomes dimmer, and the effect becomes more pronounced as the star image is moved away from the amplifier.
6!J40 [f? § t]A-
~ zfJ6 13ee
(a) Bright Star
(b) Dim Star ~l I
§6ft ,¢9
~306 63 0
130 132 134
Figure 5
::J v oLa
0(0 0 .....
173 175 177 row
TTwo-dimensional histogram and contour plot, showing the number of photoelec- trons in a box of lOXIO pixels containing (a) a bright star and (b) a dim star. The asymmetric tail of the dim star is due to CCD charge-transfer inefficiency.
+-' Q)
CJ) '+- '+- o
Q) +-' co c:
""0 (; 0.5
ou ""0
o~ +-' c: a.>
Q) 10
"E o U C
Column 270
Column 160 Column 55
•\ • \
. Integrated brightness (AID units)
a 200 1000
--.;;:: :~~~.
172.8 I 172.6 172.4
• o electrons
E =s o 172.2 U
Zo 172
............" .~

c: Q)
• -;;:::::-U--oO
6,850 electrons - - . 171. 8 -!------;,----r--.--r--.-rr-.-------r--.---r--.,-....-,,--,--,---,----,
200 1000 10000 50000
Integrated brightness (AID counts) Figure 6
Flash =

110000 50000
The effect of star brightness on charge-transfer inefficiency, shown by the change of centroid for a simulated star held at a fixed location on the CCD. (a) The curves show different locations on the CCD. (b) The curves show different levels of background light.
Since E is a function of the amount of charge in a pixel, we can improve the shape of a dim star, and the accuracy of its centroid, by adding a flat field of background light to the image. Figure 6(b) again plots apparent star centroid against star brightness, but the three curves show that the effect is reduced by adding 2,500 electrons or 6,850 electrons of background light.
We ran computer simulations of a CCD transferring charge with an efficiency which depends on the amount of charge, in an attempt to model the shape of the images found on our CCD. Results indicate that the inefficiency varies from €=10-2 to €=1O-4, as the number of electrons in the potential well rises from zero to 20,000. Above this brightness the inefficiency approaches the values of Burns (1985), 1.6XlO-4 < € < 5XlO-4•
For the search observations, we want the improved transfer efficiency obtained with more electrons in the wells. We therefore designed a flash unit con- sisting of four LED's mounted in the corners of a frame, which can be inserted in the slot intended for color filters. This slot is out of focus in the parallel beam between the Bronica lens, which receives the focussed light from the telescope, and the Nikkor lens, which focuses the light onto the CCD. [For further details of the optical configuration, see Kare (1984).] A 1.3-2.0 second flash during each observation adds a diffuse background of 20,000--30,000 electrons, with a peak- to-valley spread in intensity of -10%. This changes the faint star images read from the CCD from asymmetric "comets" to round stars. A measure of "round-
ness," R is defined to be
R = 2 O"Row - O"Column O"Row +0"Column
where O"Row and O"Column are the widths of the best-fit gausslans III the row and
column direction. We find that after flashing the typical faint star has a round- ness I R I < 0.2.
We incorporate this flash into the observation cycle at two points. First, the IBM XT powers the flash unit, typically for 1.5 seconds each time the shutter is opened. Second, we intersperse a series of three calibration exposures approxi- mately every twenty minutes of observation. These calibrations are a series of 1.5 second exposures of the flashed background plus three different dim star fields. By taking the median (not the average) of three exposures, we remove the star fields and obtain an image of the background light that each observation exposure has in common. Repeating the calibration exposures throughout the night pro- vides a check for drift in the flash background pattern. We have found this pat- tern to be stable to within 1-2 pixels over the hour required for a typical observa- tion list.
3.4. Future Improvements
. A num ber of changes to the search procedure are being introduced as they are perfected. In the near term, we will begin using a DEC microVA,,{ II com- puter, which has recently been installed at the telescope and networked to the existing system via Ethernet (see Figure 7). The microVA,,{ will control the tere- scope pointing and the readout of the CCD camera through the existing IBM
RA& Dec Drive Moters
L Q ] ~--- TK50
T ape
Amplifier & AID converter
Lab Computer Cluster
Leuchner 30" T elescope
VAX 8600's
. . . - -
Figure 7 Block diagram of future observatory system.
Ethernet r~ I
micro VAX II
computers. More importantly, it will analyze the images immediately after they are observed. Thus we will be able to adjust the pointing, focus, exposure times, flash length, and observation list selection, in response to real time assessments of the night's sky conditions.
We are also preparmg for the delivery of a new CCD detector, a Texas Instruments model VPCCD. This detector is reported to have better transfer efficiency and quantum efficiency than the current RCA CCD (Janesick et at., in press). The TI CeD's 800 X 800 pixel format will make it easier to accurately measure the position of a star by oversampling the star image (Le. spreading the images over more pixels), or alternatively to increase the number of background fiduciai stars on short exposures by enlarging the field of view.
4.1. Overview
4. Analysis and Software
We analyze the data in four stages: 1) The preprocessing procedures move the raw data to the analysis computer in a usable format, and "clean" the images to remove non-astronomical features. 2) The image analysis programs find and analyze the stars in the image and write computer files recording coordinates, brightness, and shape. 3) The field comparison programs compare the star data for fields of stars observed on two different dates, and calculate the apparent motion of the Search List star. 4) Lastly, the postprocessing tasks collect calibra- tion statistics, flag stars with significant apparent motion, and remove stars from the Search List that are no longer candidates.
Most of these outlined steps are carried out using a cluster of five VA,"'\: 8600 computers, operated by Lawrence Berkeley Laboratory's Computing Services divi- sion. For the image analysis we are using VISTA, a software system developed at the University of California's Lick Observatory (Lauer et al., 1984). Our version of VISTA has been extensively modified to include analysis algorithms specific to our application, and to use the peripherals available at our site.
4.2. Preprocessing
Data Decod£ng and Trans£t
Typically a block of data from a single night is reduced at one time. This consists of 200-400 digitized images and bookkeeping records on Beta-format videotape. These -50 Mbytes of data must be decoded, error checked, reformat- ted into a more useful computer standard, and moved to the VAX cluster for analysis. We use a PDP 11/44 computer to read the videotapes and reformat the results. We then transfer the data to the VAX cluster via traditional magnetic computer tapes.
This decoding and transporting process is time-consuming: 200 images take five hours to read into the PDP, two hours to write onto magnetic tapes, and another half hour to read into the VAX cluster. We are preparing an alternate data path, using the microVAX at the telescope to write the images on TK-50 magnetic tape cartridges. These will read directly into the VAX cluster VIa a microVAX on the Lawrence Berkeley Laboratory computer network.
Eventually we will perform the first two stages of data analysis at the tele- scope in real time, thus reducing the amount of data to be transported each observation night from -50 Mbytes to -0.05 Mbytes. This will be especially necessary with the Texas Instruments CCD, which collects almost four times as much data per observation.
Clean£ng Images
For our purposes, a perfect image would be one in which any variation of electron counts from pixel to pixel would correspond to a variation in light inten- sity across the sky. The images we obtain are imperfect, primarily because of the following factors:
a) Dark Noise. Even in the dark, the potential wells of the CCD begin to collect electrons. Certain pixels and regions of the CCD are particularly "hot," and always register some electron counts. The CCD readout amplifier adds addi- tional noise, usually seen as 60 Hz or 120 Hz stripes in the image with a peak-to-peak variation of about 200 electrons.
b) Flashed Ba~kground. The added background light from the flash unit is not perfectly uniform, as described in the previous chaptero
c) Vignetting. Due to the optical design, the intensity of light transmitted falls off sharply toward the edges of the field of view.
d) Quantum Efficiency Variation. The CCD pixels vary in their sensitivity to light by < 2% pixel-to-pixel.
The effects of the dark noise and the flash background are similar: both introduce extra electron counts varying from pixel to pixel, but constant (within poisson noise) from image to image. We remove these extra counts by subtract- ing from each image the pixel-by-pixel median of the three 1.5 second calibration
exposures. This median has only the counts in each pixel that are constant from image to image.
The vignetting and quantum efficiency can be corrected by a similar stra- tegem, since they both affect the pixel-to-pixel variation in sensitivity to light. The twilight sky provides a relatively flat light source, and thus the median of several twilight exposures gives a map of the responsiveness of each pixel to the same light level. From this map, we subtract the median of dark noise images obtained by reading out the unexposed dark CCD. The remaining image is then divided pixel by pixel into an observation exposure to correct the vignetting and quantum efficiency.
We can see the effect of these corrections graphically by plotting a histogram of the light intensities before and after each correction (see Figure 8). The histo- gram of Figure 8(b) suggests that the subtraction is correctly removing the extraneous electron counts from the image, leaving behind the star light plus the approximately gaussian noise of the counting statistics. The width of this almost-gaussian peak, (J'::::::: 250 electron counts, is consistent with the poisson noise expected after adding -30,000 counts of flashed background to the image and then subtracting the flashed-background median frame. As an additional check of our ability to remove the background light, we display the subtraction of one medianed flash image from another and look for any systematic gradients of brightness. This subtraction has always appeared as a flat pattern of random noise, as we hoped.
Figure 8
Histograms of pixel intensities (in AID counts):
(a) for typical star field with 1,000 AID counts of flashed background;
(b) after subtracting median of three flashed calibration images from (a);
(c) after dividing (b) by twilight-sky quantum-efficiency map. Dotted line shows (b) for comparison.
>< '0. 2-
.8 2E
E .fi
8. I o
E 1.5 -6
o 50 100 160 200 Intensity
Intensity (AID counts)
)( 2.15
o ...J
-t 2.5 ><
'a. '0
~a 2
o 1.5 E
-6 .<:
8' I ...J
The vignetting and quantum efficiency correction do not greatly improve the image quality, as can be seen by comparing histograms (b) and (c) of Figure 8. We have found empirically that our star center-finding results are sensitive to the background correction, but not to the vignetting and quantum efficiency correc- tion. Most of the image processing is done on images that have only been corrected for background light.
4.3. Image Analysis
Finding 8tars
Once we have a "cleaned" image to work with, the next step of the analysis is to find all of the stars in the field. We only want stars bright enough to use as fiducial background stars for measuring the position of the Search List star; the brightness cutoff thus depends on the position accuracy for a given stellar bright- ness. A typical stellar image covers 4-5 pixels, with our 2.4 arcsec/pixel plates- cale. We can find the position of this image using the number of photoelectrons,
8 1 and 8 2 , in the brightest two pixels in a given dimension. The x-coordinate position of the star is given by a function x = f(r), where r = 8 1/82 is the ratio of the brightest pixel counts, and the function f(r) can be exactly calculated for any image point-spread-function (PSF). The accuracy of the location measure- ment is
where we have used the signal-to-noise ratios Sr/N1 and S21N2 to calculate the error (jr of the ratio r. King (1983) has shown numerically for a Gaussian PSF with an RMS width of -1 pixel that af lar :::: 1 pixel. To first approximation, we use the signal-to-noise ratio SiN of a typical pixel near the peak of a star (SIN =SllNl ::::::: s21N2 and r::::::: 1) and thus estimate the uncertainty of the star location to be (jx ::::::: NIS pixels.
If we only want fiducial stars that are better than 0.2 pixels, we must then look for stars with an intensity of S = 5N counts above background. From the width of the gaussian in Figure 8(b), we know that the background noise is Nb ::::::: g AID counts::::::: 250 electrons, while the signal poisson noise is VS. We have then
(9) Solving for S, we find we are looking for stars with a brightness of S ::::::: 1250 elec-
trons ::::::: 45 AID counts above background.
In finding and measuring the stars, we cannot assume that the background sky level is flat and uniform. The cleaning corrections are not perfect, clouds and haze may intrude, and an occasional cosmic ray may have gone through the CCD during the observation, filling a pixel or two with charge. We therefore
specifically search for star-shaped peaks 45 counts above the local background. We use an algorithm due to Stetson (1986). First the image is convolved with a two-dimensional function based on a gaussian of width matching the seeing for the night of observation. A constant is subtracted from the gaussian such that the integral under the convolution function is zero. Thus the convolution output is insensitive to changes in the local background. Surprisingly, it is also possible to normalize the convolution function so that the value of any point in the con- volved image is equal to the amplitude of the gaussian that best fits the nearby pixels. We will use this gaussian amplitude, Agau88' as the measure of a star's brightness throughout the following.
Vv'e scan the convolved image for values of Agau88 > 45 counts. When a pos- sible star is found, two more tests screen its roundness and width before it is added to the list of confirmed stars. Stars with a roundness of I R I > 1.0 or a width differing from the seeing by more than a factor of two are rejected.
We have found that an iterative first-moment calculation gives a robust measure of the location of a star. First moments are, however, sensitive to the choice of background level, so we must first calculate a value for the local back- ground around the star. We look at the pixels around the perimeter of a 7X7 box centered on the star's peak. For a gaussian of full width 2.4 pixels, this is in the region where the tail of the star is lost in the background. The analysis program
takes the median of these pixel values to obtain a robust measure of the back- ground level.
Mter subtracting off this background from each pixel within the box cen- tered on the peak of the star, the program proceeds to calculate the first-moment centroid. A new box of pixels is now centered on this centroid, and a new first moment centroid is calculated. The program iterates this last step until the cen- troid value converges. (If the program takes >6 iterations without converging, the star has an un-starlike shape, and we flag it as a bad fit.)
Star Description and Storage
The image analysis program also calculates other useful characteristics of each star in the field. A Star Data File is written for each image; it contains the following data for each star:
a) Row coordinate from first moment b) Column " c) Best fit gaussian amplitude d) Peak pixel value
e) Summed brightness above background of pixels within radius=3 f) Roundness g) Width h) Amplitude of gaussian fit to three points in row
i) Width " j) Amplitude of gaussian fit to three points in column k) Width "
Some of these values are of frequent use, others are of occasional use, and some are helpful for diagnostics when a night's results prove unsatisfactory. One particularly useful value in the "occasional" category is the peak pixel value. A
value of 16,383 AID counts indicates a saturated pixel (at the limit of the AID converter) and thus flags the calculated centroid as unreliable.
4.4. Field Comparisons
Following the collection of two epochs of observations of the same star fields,
and reduction of the images to create the Star Data Files, we are ready to com-
pare the star fields. We begin by creating a list of matching stars for each star
field pair. This list is corrected and pruned to contain only those stars useful in
registering the two fields. We then fit the rotation, magnification, and offsets
necessary to make the two fields match, and use the results of the fit to calculate
th o ",",,",,,~o~+ ~~+:~~ ~l' t h _ ~;:>arch List stat The iollowing sections describe the "'.........'" u.y,Pal. v .u.u. .lllV\JIVU VI "'liC .......'-' •
details of this procedure.
Preparing the Matched Lists
We have found that stars that lie above row 400 on the CCD gIve poor results in the fitting procedures, because the vignetting in this region becomes so sharp that even the twilight-image correction leaves a poor image. Similarly, saturated stars, with more than 16,383 AID counts in their peak pixel, result in poor fits. Before matching the fields, we delete these stars. The field comparison program then finds matches between the two epochs, for the remaining stars. The matching is accomplished by an exhaustive search through all possible matchings of stars. This is a reasonably efficient algorithm for our lists, which are both ordered by increasing row coordinate. Typically, one of the first few
match attempts is correct, with epoch-to-epoch partners found for >75% of the stars.
At this stage, we also make a correction to the star coordinates to account for the pincushion optical distortion ~r = por3, where r is the distance from the optical center at (Column 140, Row 234). If a star is located at rl in the first- epoch image and at r2 in the second-epoch image, then the differential pincushion distortion, which the program corrects, is
6.dijferential = poCr? - rg)
where Or =- rl-r2 is the measure of the mismatch in telescope pointing between the two epochs. For our optical system we have measured the pincushion con- stant to be Pc = 6.5XlO-8 + 405XlO-8 pixels-z. We estimate the systematic error, (jpincushion' introduced by the uncertainty in the pincushion constant, OPo = 4.5XlO-8 pixels-2 to be
(jpincushion = a6. dijferential oPo oPo
With the RMS pointing error 0 6 pixels, and a typical distance r = 160 pixels of a fiducial star from the optical center, we find the systematic error introduced by the corrected pincushion to be approximately (jpincushion ::::= 0.01 pixels.
Field Fitting
The field comparIson program next goes twice through a night's worth of matched lists. The first time through, a "downhill simplex" fitting algorithm [see Press [i]al. [/i](1986)] finds the values of four parameters, rotation, magnification, row offset and column offset, that best align each pair of star fields. (The fitting algorithm weights each star in the field by its brightness.) We use the global aver- age of all of the magnification and rotation values to describe the difference in conditions at the telescope between the two nights. The second time through the matched lists the fitting program finds individual values for the row and column offsets between the star fields, but uses the global average of magnification and rotation in its calculations. This two parameter fit is less affected by outliers and
sparse fields with few fiducial stars.
We would like to do this fitting procedure leaving out the Search List star. The Dearborn Catalog, however, gives coordinates that can be inaccurate by as much as 3 arcminutes ~ 75 pixels, and there may be more than one likely bright
star within this radius of error. Our solution is to perform the field fitting once for each likely bright star near the center of the image. The program leaves the likely star out of the fit, and then uses the resulting fit parameters to measure the apparent motion, ma, of the likely star. The extra inadvertent measurements of stars which are not on our Search List does not reduce the probability of finding a companion star; they simply provide extra calibration points as distant stars which do not move.
4.5. Postprocessing
Candidate Selection and Rejection
Thus we reach the primary goal of the analysis: we have measured the apparent motion of each Search List star (pius a few inadvertent extras). The. last step is to dispose of each Search List star according to its measured rna. We delete candidates with small rna from the Search List (using criteria described in Section 5.3), and re-examine the stars which are in the outskirts of the motion distribution. Generally none of these (except one) should corne close to the amount of motion expected for a star closer than 0.8 pc, but they will allow us to
eliminate systematic errors in our experiment. Improvements of the experiment should seek to improve the precision of our measurement of apparent motion. This will reduce the time required between first- and second-epoch observations.
Candidate Follow-up Studies
When and if we do find a star with rna of the magnitude and direction appropriate for a nearby star, we will check our result in a few ways. First, we will follow its motion for a longer time, and if it follows the expected path in the
sky this would be conclusive proof of its distance. For immediate confirmation, we will take photometric exposures with different color filters. The colors can dis- tinguish between red dwarf stars and red giant stars in certain cases (Penfold, 1978). We will also obtain spectra at Lick Observatory, and look for the charac- teristic bands which differentiate dwarfs from giants.
These measurements will test whether the star is nearby, but not whether it is orbiting the Sun. We will need to measure the radial velocity and the proper motion to determine the orbital parameters. The radial velocity can be obtained from the spectra taken at Lick Observatory. The minimum annual proper motion, /1-(rmax) ::::::: 0.01 arcseconds, is well below the detection threshold of our parallax measurements, so we will probably measure /1- using an astrometric pro- gram at another observatory. However, if the star happens to be very close, or the new Texas Instruments CCD gives much better performance, we may be able to measure /1- at the Leuschner telescope.
5. Results and Discussion
5.1. Calibration Results
Table 3 describes the observations used for calibration measurements. The precision of our parallax measurement can be seen in the distribution of apparent motion, rna, that we find for stars that are not expected to move. Figure 9 shows this distribution for stars observed twice on the night of 8 August i986, approxi- mately one hour apart.
Table 3
Calibration Observations
List Name
Aug I Aug II
Oct I Oct II
8 August 1986
13 October 1986
Time Range
Universal Time
. 05:59-06:36 06:48-07:25
02:30-03:14 03:18-04:02
Airmass Range sec (()zenith)
1.01-1.22 1.04-1.28
1.05-1.29 1.13-1.44
We expect no parallactic motion in one hour, so the scatter around rna = 0 in both the row and column directions gives us an estimate of the error of our meas- urement. The RMS deviations about the mean are CTm• = CTRow = CTColumn = 0.063 pixels, with a sample size of N = 60 star fields. (For the plate scale of 2.4
Figure 9
Apparent motion, rna' for Aug I list matched with Aug II list.
47 arcseconds per pixel, this corresponds to (J'ma = 0.15 arcseconds.) The furthest
outliers from zero motion are at t::,.Row = 60Column = 0.19 pixels. Figure 10 shows similar results for a subset of the same set of stars also observed twice the same night but tvv'O months later, on October 13, 1985. For this histogram,
(J'Row = 0.054 pixels (J'Column = 0.052 pixels
6oRow = 0.16 pixels (12) 6oColumn = 0.14 pixels
for N = 42 star fields. We see that the distributions have a similar scatter under different observation conditions: the atmospheric seeing on the August night was about twice that of the October night.
.The accuracy of the measurements of rna is indicated by the mean of these distributions. We expect <rna> = 0 for stars observed twice on the same night. We find for the August night,
<rna> Row = 0.011 ± 0.008 pixels
<rna> Column = 0.002 ± 0.008 pixels (13) and for the October night,
<rna> Row = -0.021 ± 0.010 pixels
<rna> Column = 0.021 ±0.008 pixels (14)
There is some bias apparent in the October data, but it is small compared to (J'rna' which determines the smallest rna the search can detect.
+-' c Q.)
> Q)
'+- a
'- Q.)
- '-
...0 5-
'--- 0 -r I I
-0.5 -0.3
c Q.)
10 -
-0.1 0.1 0.3 0.5
rna (pixels)
o~-_--......-4------l---,-~---,.----t -0.5 -0.3 -0.1 0.1 0.3 0.5
rna (pixels)
Figure 10
Apparent motion, rna., for Oct I list matched with Oct II list.
>Q ) '+-
0 '-
The overall measurement precision of ma is best seen in the distribution of ma found for stars measured two months apart, as shown in Figure 11. This dis- tribution tests the entire data collection and analysis procedure. The mean apparent motion is near zero:
<ma > Row = -D.008 ± 0.010 pixels
<ma>Column = 0.030 ±0.010 pixels For these N = 41 -stars, we have the following deviations:
<7Row = 0.080 pixels <7Column = 0.071 pixels
l::..Row = l::..Column = 0.18 pixels
5.2. Error Analysis
Systematic Errors
(15 )
There are a number of sources of systematic error, <7systematic> at the level of -0.03 pixels. The transfer inefficiency of the CCD, for example, may not allow more precise measurements, even with the improvement due to flashing the CCD with background light. If the inefficiency changes from E = 0.0001 in the first epoch to E = 0.0002 in the second (corresponding to a temperature change of about 10 degrees), the amount of charge in the tails of a star image at (Row 512, Column 320) changes from 8% to 15%. This would shift the centroid by
0 -0.5 -0.3
+-' C <l> >
OJ ..0
E :J
+-' C <l> >
<l> ..0
E ::::l
o - + - - -0.5
-0.1 0.1 0.3 0.5
rna (pixels)
-0.3 -0.1 0.1 0.3 0.5 rna (pixels)
Figure 11 Apparent motion. rna. for Aug II list matched with Oct I list.
O"ineffic :::::::: (0.15--0.08) pixels:::::::: 0.03 pixels.
As discussed in Section 4.4, the pincushion distortion also contributes an error on the order of O"pincushion :::::::: 0.01 pixels. Evidence for this source of error is the increase in rna for fields with poorly matched pointing (> 12 pixels) between two epochs. The pincushion error can, however, be removed by fitting the pin- cushion parameter, Po, to the data set taken as a whole.
We thus estimate the overall systematic error to be
This is adequate for the search for a Companion. If more precision is needed for a different search design, these sources of error would have to be studied further. In particular, it would be important to understand if the <rna> bias and O"systematic are due to the same source.
Statistical Errors
The rna distributions shown in Figures g, 10, and 11 are clearly not simple gaussians drawn from a single normal distribution. They can be approximately described as the sum of gaussians drawn from a family of normal distributions, where the widths of the distributions depend on a number of contributing sources of error. The two dominant statistical sources of error are a) the centroid uncer- tainty, O"centrol'd' a function of the brightness, Aqauss' of a star, and b) the number of background fiducial stars, N fiducial, in the field. These two factors are correlat-
ed, since both Agauss and Nfiducial increase with increasing exposure time. For large N fiducial' both of these sources of statistical error become negligible, and the overall measurement error reaches a minimum at O"systematic.
Determining O"centroid
Figure 12 shows the dependence of O"centroid on brightness, by plotting the residual deviations of star coordinates from the overall fit against AgauSS" The plot includes fields with more than N fiducial = 20 stars, and with telescope point- ing matched to within 9 pixels for the two epochs. We expect these epoch-ta- epoch deviations, O"e-e, between two stars to be given by
2 (Je-e = V20"centroid + 2 ·
where O"centroid is the individual standard deviation of each centroid measurement, and the factor of 2 accounts for the two measurements made, in the first and second epoch.
Improving on the estimate of Section 4.3, we will take O"centroid ~ N /8 2 , where we use 82, the signal of the second brightest pixel, rather than the ampli- tude of the star's gaussian in the signal-ta-noise calculation. This choice reflects the fact that a star's location on the pixels is determined roughly by the ratio of the counts in the t\'/O neighboring brightest pixels, and this ratio's uncertainty is dominated by the uncertainty of the less bright pixel. For stellar gaussians with amplitude A gauss (in electrons) and FWI-llv1 ~ 2..5 pixels, we find 8 2 ~ .Agauss/2.
-0.5 -
.e••- , DD,:.•• '.
a.> x
ll.l I ll.l
-r J )
<D x
:. :: -;'.,,:'.:.. '.," .
.'\",."\"I. '1.# ," •••• . . r ~,w.. I I I l l . ~ ( . , . . . i ~ " " ' . ~"" • • ' ~ . . . "
··I.,~ -" :-'.:'. ~)~~;..,_."" r·;J ••.~-.-t~·.
..:.:. ,
' .
-. ... ,-...
" :
. :
..'. · ., ..

- ' '.' ."',' .
I •
Q. 0.2
q) I q)
~:i"J..,I :.f...... " .. . •
. .'
·.. .
~..... !..-.~,:....,,'T ,":.t. :.
J ::::; ••:....
' ~ ,e.
II ••
Agauss (AID counts)
'\~~ -.~
--== • 1000
<Agauss> (AID counts) Figure 12
For Aug I list matched with Aug II: (a) the residual epoch-to-epoch deviation, 0e-n in column coordinate vs. the brightness of the star, Agauss' for all stars in a field after correcting for fit offset, magnification, and rotation; (b) standard devia- tion, <7e- n for residuals in (a), averaging every 69 stars binned in order of increas- ing Agauss.

2 N22 (J"e-e = 2 - 2 - + (J"systematic
+2 (J"systematic
where Nb = 300 electrons is the background nOIse. The solid line in Figure 12 shows (J"e-e plotted against Agauw The consistency of this curve with the empiri-
cal results indicates that (J"centroid ::::::: N/S2 is a reasonable estimate. In the foIIow- ing analysis, we will use this estimate for the centroid uncertainty, (J"search' of the Search List star, and for the centroid uncertainty, (J" jiducial' of the fiducial stars.
Number of Fiducial Stars We can now write the rna-measurement error, (J"m.' for a field with N jiducial
fiducial stars, as
2 (J"~ducial N fiducial
::::::: [2
+ [i](J"systematic[/i]
(A jiducial/2)+Nb2 N jiducial (A jiduciai/2)
2 ]1/2
where A jiducial' the fiducial star brightness, is itself dependent on N jiducia/l since longer exposures give larger N jiducial and brighter fiducial stars. This dependence is shown in Figure 13. We neglect (J"search' the uncertainty of the Search List star's centroid, because it is much smaller than asystematic' The soiid curves in
".4000 E
co 0-

tl 3000 :J
v 2000 1000 0
I • •I
°1 •

0 10 20 30 40 50 60 70
<N fiducial>
Figure 13
Brightest fiducial star amplitude, A fiducia/l vs. number of fiducial stars, N fiducial' for 72 fields binned by 9 in order of increasing N fiducial' The median of the 9 fields' A fiducial is plotted.
Figures 14 and 15 show the expected erm • plotted against N fiducial' The small un- derestimate of the error can be accounted for by our use of the brightest fiducial star rather than a weighted average of all of the fiducial stars.
5.3. Search Strategy
With this understanding of the measurement preCISIon, we have chosen a search strategy that minimizes the chance of missing a nearby star on the Search List, and also minimizes the time required to complete the search.
Dim Stars: Long Exposures
Dim Search List stars require long exposures, and thus have more fiducial stars available. For those stars with more than ten background fiducial stars visi- ble in the same field, we find that any parallactic motion greater than
mrr = 2.6.Row,Column :::::: 0.4 pixels would be clearly distinguishable from the zero motion stars shown in Figure 11. This value of m rr lies at -50" for N fiducial > 10,
based on the deviation shown in Figures 14 and 15. For these long-exposure stars, two epochs separated by two months gives a little more than m rr = 0.4 pix- els for stars undergoing their maximum parallactic motion. In Figure 16, the dis- tribution of m a for Aug II matched with Oct I is replotted to show this expected two-month motion and the six-month peak-to-peak motion, (mrr)p_p-
. E.
• I. .. .
.. ...
0 20 40 60 80 100 N stsrs
o 0 0 -l-0--~----'20---r---4rO--'----6TO---'----8'0---'-~-----:1-:!O0 <Nstsrs>
Figure 14
For Aug I list matched with Aug II: (a) the apparent motion, rna' in row coordi- nate vs. the number of fiducial stars, N fiducial; (b) standard deviation, (Jrow' for apparent motion in (a), averaging every 20 stars binned in order of increasing N fiducial'
III E c:
::J 0
.. .
, .. .
oOO+---.-----,---,---r-----.----,-----.-----,---,--- o 20 40 60 80 100
Figure 15
For Aug I list matched with Aug II: (a) the apparent motion, rna' in column coordinate vs, the number of fiducial stars, Nfiducial; (b) standard deviation, (Tcolumn' for apparent motion in (a), averaging every 20 stars binned in order of increasing N fiducial.
...... C Q)
> Q)
Q) ..0
:::l Z
-0.25 rna (pixels)
a -1.25
0 0.25
0.50 0.75
1.5 2 2.5 3 !II!! ! !
...... C Q)
> Q)
Q) ..0
rna (arcseconds) -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1.5 2 2.5 3
-1 -0.75 -0.50
rna (arcseconds) -3 -2.5 -2 -1.5 -1 -0.5 0 0.5
~oFigure [/b]16
Apparent motion, ma' for Aug II list matched with Oct I list. The arrows indi- cate the average magn£tude of a) the expected parallactic motion, m7l"l for this two-month time interval, and b) the six-month peak-to-peak motion, (m".)p_p. The d£rect£on of motion depends on the ecliptic coordinates of the measured star.
Bright Stars: Short Exposures
Bright Search List stars require short exposures and thus have fewer fiducial stars available. We handle these stars as follows.
a) For the stars with at least one fiducial star, we will allow a longer time interval to pass between the epochs. If necessary, the entire six-month peak-to-peak parallax, (m1!")p-p > 1 pixel, can be used.
b) Search List stars with no fiducial stars in the field present a problem of dynam- ic range. We will take multiple exposures of a bright star without moving the telescope, reading out the CCD between each exposure. Stacking the images in software then increases the effective exposure time without saturating the image of the bright star. This brings fainter fiducial stars above the noise.
Both of these strategies will be obviated when the Texas Instruments CCD is incorporated into the system. The two- to three-fold increase in dynamic range will bring more fiducial stars above the background noise before a bright star sa- turates the CCD. The 800X800 pixel format will also allow us to share the star light of the bright star among more pixels, and/or use a larger field of view con- taining more fiducial stars.
For the Aug I and Aug II lists, the Search List stars with fewer than one fiducial star in the field account for less than 5% of the Search List. However, since these stars are the brightest in the Search List and the Companion could be a very close star, it is important that we include them in the Search.
6. Conclusion
The above results and analysis demonstrate that the Companion Star Search can accomplish its goals. The 41 stars used as a calibration in Figure 16 also represent the beginning of the Companion Search. The measured values of rna for these stars indicate that none of these stars is a nearby companion. We are currently observing the remaining 2,729 stars on the Search List, and expect to complete the observations within the next year. We hope that this time span is cut short by the discovery of a Companion.
7. References
Abt, H. A. (19S3). Normal and abnormal binary frequencies. Ann. Rev. Astron. Astrophys., 21,343-72.
Alvarez, L. W., W. Alvarez, F. Asaro, and H. V. Michel (19S0). Extraterrestrial cause for the Cretaceous-Tertiary extinction. Science, 208, 1095-110S.
Alvarez, W. and R. A. Muller (1984). Evidence from the crater ages for periodic impacts on the earth. Nature, 308, 71S-720.
Alvarez, W. (1986). Toward a theory of impact crises. EOS, 67, no. 35, 649, 653-655, 658.
Bahcall, J. N. and R. M. Soneira (1980). The universe at faint magnitudes. 1. Models for the Galaxy and predicted star counts. Astrophys. J. Suppl., 44, 73.
Bahcall, J. N. (19S6). Brown Dwarfs: Conference Summary. In Astrophysics of Brown Dwarfs, ed. M. C. Kafatos, R. S. Harrington and S. P. Maran, pp. 233-237. Cambridge: Cambridge University Press.
Burns, M. S. (19S5). Development of a CCD Camera for an Automated Super- nova Search and Observations of a Supernova in NGC 5033. LBL-2064S,
Lawrence Berkeley Laboratory, Berkeley, CA. Ph.D. Thesis.
63 Davis, M., P. Hut, and R. A. Muller (1984). Extinction of species by periodic
comet showers. iVature, 308, 715-717. Hoffieit, D. (1982). The Bright Star Catalogue, Fourth revised edition. New
Haven: Yale University Observatory. With the collaboration of C. Jaschek. Hut, P. (1984). How stable is an astronomical clock which can trigger mass
extinctions on the earth? Nature, 311, 638-641.
Hut, P. (1985). Evolution of the solar system in the presence of a solar compan- ion star. In The Galaxy and the Solar System. Proceedings of conference, Tucson, Arizona, January 1985, to be published.
Janesick, J. R., T. Elliott, S. Collins, and H. Marsh (in press). The future scientific CCD. SP/E.
Kare, J. T. (1984). An Automated Search for Supernovae. LBL-19340, Lawrence Berkeley Laboratory, Berkeley, CA. Ph.D. Thesis.
Kare, J. T., M. S. Burns, F. S. Crawford, P. G. Friedman, R. A. Muller, C. R. Pennypacker, and S. Perlmutter (to be published). An automated search for supernova explosions.
King, 1. R. (1983). Accuracy of measurement of star images on a pixel array. P.A.8.P., 95, 163-168.
Lauer, T., R. Stover, and D. Terndrup (lg84). The VISTA Programmer's Guide, Version 2. Technical Reports No. 34, University of California, Lick Observa- tory.
Lee, O. J., R. J. Baldwin, and D. W. Hamlin (lg43). Annals Dearborn Observa- tory, V. Part lAo
Lee, O. J. and T. J. Bartlett (lg44). Annals Dearborn Observatory, V. Part IE. Lee, O. J., G. D. Gore, and T. J. Bartlett (1947). Annals Dearborn Observatory,
V. Part 1C. Luyten, W. J. (1979;1980). NLTT Catalogue. Minneapolis: University of Min-
nesota. (4 vols.)
Muller, R. A. (1985). Evidence for Nemesis: a solar companion star. LBL-20438, Lawrence Berkeley Laboratory. Also in Proceedings of The Galaxy and the Solar System, Tuscon, Arizona.
Nagy, T. A. (lg7g). Report no. R-SAW-8/7g-0l, Systems and Applied Sciences Corporation.
Penfold, J. E. (1978). Observations of faint red stars at intermediate galactic lati- tude. !'vlon. Not. R. Astr. Soc., 182, 283-2gl.
Press, W. H., B. P. Flannery, S. A. Teukolsky, and W. T. Vettcrling (1986). Numerical Recipes, p. 28g. Cambridge University Press.
Rampino, M. R. and R. B. Stothers (19S4). Terrestrial mass extinctions, come- tary impacts and the Sun's motion perpendicular to the galactic plane. Nature, 308, 709-712.
Raup, D. M. and J. J. Sepkokski, Jr. (19S4). Proc. Nat. Acad. Sci. U.S.A., 81, S01-S05.
Schaifers, K. and H. H. Voigt (19S2). Astronomy and Astrophysics. In Landolt- Bornstein: New Series, ed. K.-H. Hellwege, vol. 2, p. b.25. Berlin: Springer- V erlag.
Shapiro, S. L. and S. A. Teukolsky (1983). Black Holes, White Dwarfs, and Neu- tron Slars. New Y ork: Wiley.
Shoemaker, E. M. (1985). In The Galaxy and the Solar System. Proceedings of conference, Tucson, Arizona, January 1985, to be published.
Stetson, P. B. (19S6). DAOPHOT User Manual. Victoria, British Columbia: Dominion Astrophysical Observatory. Computer program package for pho- tometry.
Stevenson, B. (1984). Private communication.
Reynolds, R. T., J. C. Tarter, and R. G. Walker (19S0). A proposed search of the solar neighborhood for substellar objects. Icarus, 44, no. 3, 772-779.
Toon, O. B., J. B. Pollack, T. P. Ackerman, R. P. Turco, C. P. McKay, and M. S. Liu (1982). Evolution of an impact-generated dust cloud and its effects on the atmosphere. Ceol. Soc. Am. Special Paper 190, 187-200.
Turco, R. P., O. B. Toon, T. P. Ackerman, J. B. Pollack, and C. Sagan (1983). Nuclear winter: Global consequences of multiple nuclear explosions. Sci- ence, 222, 1283.
Whitmire, D. P. and A. A. Jackson (1984). Are periodic mass extinctions driven by a distant solar companion? Nature, 308, 713-715.
Whitmire, D. P; and J. J. Matese (1985). Periodic comet showers and Planet X. Nature, 313, 36-38.
van de Kamp, P. (1982). Evolutionary trends in wide binaries. In Binary and Multiple Stars as Tracers of Stellar Evolution, ed. Z. Kopal and J. Rahe, pp. 81-103. Dordrecht: Reidel.

Dernière édition par sunlightwalkin le Lun 4 Juil - 07:39 (2011); édité 2 fois
Revenir en haut

MessagePosté le: Lun 4 Juil - 07:35 (2011)    Sujet du message: Publicité

PublicitéSupprimer les publicités ?
Revenir en haut
andree dole

Hors ligne

Inscrit le: 28 Déc 2010
Messages: 76
Localisation: ouest

MessagePosté le: Lun 4 Juil - 07:38 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

un résumé en trois phrases serait bienvenu
andrée dole
Revenir en haut

Hors ligne

MessagePosté le: Lun 4 Juil - 07:42 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Abstrait Une étoile de compagnon à moins de 0.8 PC du Sun a été postulée pour expliquer une périodicité possible de 26 MYR dans des extinctions de masse des espèces sur la terre. Une telle étoile serait déjà cataloguée dans le catalogue d'étoile lumineux de Yale à moins qu'elle soit plus faible que système mv = 6.5 ; ceci limite les types stellaires possibles pour un ion invisible de compan- aux nains rouges, aux nains bruns, ou aux objets compacts. Les nains rouges expliquent environ 75% de ces étoiles possibles. Nous décrivons ici la conception et le développement d'une recherche astrometric d'un compagnon voisin de nain rouge avec une parallaxe de crête à crête de six mois de > ; 2.5 arcseconds. Nous mesurons la parallaxe de 2.770 étoiles rouges faibles de candidat choisies parmi le catalogue d'observatoire de Dearborn. Un télescope 3D-inch automatisé et système d'appareil-photo de CCD rassemblent des images digitalisées des étoiles de candidat, avec un 13' ; X16' ; champ environnant des étoiles moulues back-. des images de Deuxième-époque, prises quelques mois plus tard, sont enregistrées aux premières images d'époque utilisant les étoiles de fond comme fiducials. Un mouvement apparent, rna' ; du candidat les étoiles est trouvées à une précision de CTma : : : : : : 0.08 Pixel : : : : : : 0.2 arcseconds pour des champs avec le > fiduciel de N ; 10 étoiles fiducielles évidentes au-dessus du bruit au sol back-. Cette précision est suffisante pour détecter le mouvement parallactique d'une étoile à 0.8 PC avec un intervalle de deux mois entre les époques d'observation. On observe des images avec peu d'étoiles fiducielles au-dessus de bruit de fond avec un plus long intervalle entre les époques. Si une étoile est trouvée avec le mouvement parallactique élevé, nous confirmerons sa distance avec d'autres mesures de parallaxe, photométrie, et matrices spectrales de stu-, et mesurera la vitesse radiale et le mouvement approprié pour établir son orbite.

Revenir en haut

Hors ligne

MessagePosté le: Lun 4 Juil - 07:45 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Revenir en haut

Hors ligne

MessagePosté le: Lun 4 Juil - 07:46 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Revenir en haut

Hors ligne

Inscrit le: 13 Jan 2011
Messages: 2 432

MessagePosté le: Lun 4 Juil - 08:17 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Ils trouveront toujours une excuse pour douter, tant qu'ils n'auront pas le vécu de ce qui est et de ce qui vient. 
Revenir en haut
andree dole

Hors ligne

Inscrit le: 28 Déc 2010
Messages: 76
Localisation: ouest

MessagePosté le: Lun 4 Juil - 10:20 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

merci du résumé
andrée dole
Revenir en haut

Hors ligne

Inscrit le: 27 Jan 2011
Messages: 7 953
Localisation: Is-Er Dauphiné

MessagePosté le: Lun 4 Juil - 19:49 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Bingo Finito ! 
« ...quand vous pénétrez réellement dans ce Cœur, il n’a plus ni centre ni circonférence ; il est partout. »
Revenir en haut

Hors ligne

MessagePosté le: Lun 4 Juil - 19:50 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Revenir en haut

Hors ligne

MessagePosté le: Lun 4 Juil - 19:56 (2011)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE Répondre en citant

Sympa les moeurs des terriens  Mr. Green Mr. Green Mr. Green Mr. Green Mr. Green
Revenir en haut
Contenu Sponsorisé

MessagePosté le: Aujourd’hui à 11:53 (2017)    Sujet du message: DEPUIS 1986; ILS SAVENT QU'UNE EXTINCTION MASSIVE ARRIVE

Revenir en haut
Montrer les messages depuis:   
Poster un nouveau sujet   Répondre au sujet     Index du Forum -> Actualités -> NewPolitique&Social Toutes les heures sont au format GMT + 1 Heure
Page 1 sur 1

Sauter vers:  

Index | Panneau d’administration | Creer un forum | Forum gratuit d’entraide | Annuaire des forums gratuits | Signaler une violation | Conditions générales d'utilisation
Powered by phpBB © 2001, 2005 phpBB Group
Traduction par :