Searching the solar neighborhood for protoplanetary debris disks


Material Information

Searching the solar neighborhood for protoplanetary debris disks a survey of Vega-like sources and a discussion of new disk examples
Physical Description:
x, 178 leaves : ill. ; 29 cm.
Fisher, Robert Scott
Publication Date:


Subjects / Keywords:
Circumstellar matter   ( lcsh )
Astronomy thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Astronomy -- UF   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 2001.
Includes bibliographical references (leaves 172-177).
Statement of Responsibility:
by Robert Scott Fisher.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 027697059
oclc - 48198422
System ID:

This item is only available as the following downloads:

Full Text







Copyright 2001


Robert Scott Fisher

This work is dedicated to my fellow inhabitants of 317, 135, 8, 2613, 139, and 1072A.


I would first like to acknowledge my advisor Dr. Charles Telesco for his role in

this work and for his valuable friendship. In the time spanning the work for this

dissertation I have learned from him in many ways. In the office, the lab, or the control

room, I have always enjoyed our discussions. It seems like not so long ago that we met

for the first time in a small office on the second floor and talked about the idea of this

project... Not long after that came the infamous 'run of 100 ms.' Then we were off to

Chile where we learned that the polar axis of a telescope points to the South pole in the

southern hemisphere, and that cake is not always what it seems. We had nights of El Niio

there, and nights of high stakes science. Then we went to Keck and learned what D4

really means. At Keck we first used the unit 'pJy' and also learned that the PicTel can

sometimes be a good thing. Now we are at Gemini. OSCIR is still getting time on the sky

and T-ReCS is coming soon. I look forward to the planned collaboration with Charlie,

and I look forward to our continued friendship.

I would also like to thank Dr. Robert Pifia for his guidance and his friendship.

Bringing OSCIR into operation and spending time with Robert running the system are

some of my best memories of this work. I admire his ability to diagnose a problem, and I

will always relish coming into his office to 'check this out.' I hope that I somehow

contributed to the many talks about the OSCIR interface and the acquisition system. I

would like to thank Kevin Hanna for his development of the OSCIR electronics system,

and Jeff Julian for his design and fabrication of critical OSCIR components. One of the

main reasons that OSCIR has been a premier instrument for the last 6 years is the talent

these two possess. I would especially like to recognize Kevin for his unquestionable

perseverance and dedication to getting OSCIR operational. I thank him for the nights he

and Robert put in at HP and Tololo.

I also would like to thank my friends James Radomski and Jim De Buizer. I sat

here and tried to think of which of their qualities I wanted to mention, but basically I

thank them for the whole deal and for being my friends. We all know the secret of row

90, and I can think of no better way to spend a Saturday afternoon than in The Swamp

with the two of them and Brent.

My family has also played an important role in this work. I thank Brent, Jeanette,

and Wik for keeping me grounded and keeping me company. There has always been an

inner strength to our family-unit, and that has helped me in ways both obvious and subtle.

Thanks to them I never missed a flight, and I never forgot the HASP. To my extended

family in St. Augustine and in Pennsylvania I also send my thanks for their support since

day one.

The last people I want to recognize are Ann, Audrey, Debra, Glenda, and Tracey

in the main office. We all know that without them the department would not run, but no

one seems to thank them for it. So thanks.

This work was supported in part by a NASA Graduate Student Research Fellowship and

funding from the National Science Foundation



ACKNOW LED GM ENTS............................................................................................ iv

ABSTRACT ....................................................................................................................... ix


1 INTRODUCTION ..................................................................................................... 1

Star Form ation and Disk Creation.................................. ............... .......................... 4
Evolutionary Status of Vega-like Disks.................................................. .............. 8
Outline of the Project .............................................................................................. 11

2 OBSERVATIONS AND DATA REDUCTION..................................... ........... 13

Source Lists ................................................................................................................... 14
Instrum entation........................................................................................................ 17
Observational Strategy ............................................................................................ 20
Nightly Variation of the PSF...................................................... ........................... 21
Data Reduction Steps .............................................................................................. 23
Rem oval of Contam inated Data ................................................ ........................ 23
Cross-Correlation of Im ages .................................................... ......................... 25
Skyfit ................................................................................................................... 25
Calibration/Airm ass Correction ................................... .........................................26
Flat-field Correction ........................................................................................... 29
Photom etry .......................................................................................................... 29
Color Correction.................................................................................................. 30
Resolved or Unresolved: The Question ................................ .............. ....... ..... 34

3 SURVEY DISCUSSION ........................................................................................ 37

Disk Sizes...................................................................................................................... 38
Photosphere Rem oval..............................................................................................41
Grain M odels........................................................................................................... 47
Color Tem peratures and Optical Depths............................................ ............... 54
Scale Size Com parison........................................................................................ 59
A Proposed Evolutionary Sequence ............................................... ....................... 64
SED Evolution........................ ............................................................................. 64

Class IIIa: The near-IR N ine ........................................................................ 68
Class IIIb: ZAM S Sources .............................................................. ......... ...........70
Class IIIc: Post ZAM S Sources .......................................................................... ...71
Survey Sources in the HR D iagram ......................................... ................................. 73

4 THE FACE-ON DISK OF H D 169142................................................................... 79

M id-Infrared Observations at Keck...................................................... .................. 83
HD 169142 Source Size ........................................................................................ 85
M odeling of Characteristic Grain Param eters............................................ ......... 87
M odel Results.......................................................................................................90
Grain Sizes and Tem peratures................................................................................... 91
Discussion and Conclusions.......................................................... 99
Disk Orientation ..................................................................................................... 99
Evolutionary Status ................................................................................................ 101
Conclusions ............................................................... .................................................. 106

5 KECK IM AGIN G OF HD 141569 ............................................................................. 108

Observations................................................................................................................ 110
Source Size.................................................................................................................. 113
Dust Tem perature and Grain Size ............................................................................... 117
Com prison to N ICM OS Im ages ................................................................................ 119
Conclusions About the Disk of HD 141569................................................................ 122

6 SAO 26804: A QUESTION OF CLASSIFICATION ............................................... 124

The Detection and Non-detection of the SA O 26804 Disk......................................... 125
Previous Observations ............................................................................................. 125
New Keck Im aging w ith OSCIR ........................................................................... 127
First-Ascent Giant or M ain-Sequence Dw arf? ......................................................... 131
Luminosity Class V: A Main-Sequence Dwarf................................. 132
Lithium Abundance................................................................................................. 134
Position in the H-R Diagram ................................................................................... 136
Sum m ary and Conclusions.......................................................................................... 139

7 INSTRUMENTATION ACTIVITIES: THE OSCIR SYSTEM................................ 140

System Overview ........................................................................................................ 141
The Dew ar............................................................................................................... 144
Optics .................................................................................................................. 147
Entrance W indow and Filters ............................................................................ 150
Detector ............................................................................................................... 154
VM E Crate .............................................................................................................. 158
Control Com puter.................................................................................................... 162
Sum m ary of Activities ................................................................................................ 164


M ID-IR ASTRONOMY: A SHORT COURSE.............................................................. 165
Extracting the Source Signal ....................................................................................... 167
The Standard Chop-Nod Technique.............................................................................. 168

LIST OF REFERENCES ................................................................................................ 172

BIOGRAPHICAL SKETCH..................................................................................... 178

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Robert Scott Fisher

August 2001

Chairman: Charles M. Telesco
Major Department: Astronomy

Theory suggests that planets form in the dusty circumstellar disks associated with

young stars. Many classes of stars exhibit excess infrared emission that is attributed to the

dust grains in these disks. In this dissertation we present mid-IR (10 and 18 gm) imaging

observations of a sample of almost forty Vega-like stars. These are stars that have mostly

finished their pre-main-sequence evolution and are now close to, or already on the main

sequence. Our imaging shows that most of these sources are unresolved at arcsecond

resolution at 10 and 18 9Lm. We did however make the first detection of extended mid-IR

emission for two stars; HD 141569 and HD 169142. We use our observations to

investigate the properties of the mid-IR-emitting dust grains in these two resolved disks,

and the unresolved disks of 22 of the survey sources using radiative equilibrium models.

The results of our modeling show that the mid-IR emission from these disks traces the

position of grains that are 1 to 10 ltm in diameter in the disks of these stars. The mid-IR-

emitting grains orbit the stars at distances of tens of AU, similar to the size of the Solar

system, and therefore trace the region of assumed planetary formation in these disks.

Additionally, we place the survey stars into the H-R diagram, and use the characteristics

of their infrared excess in conjunction with their positions to suggest a proposed

evolutionary sequence for them. Finally, discuss the development, characterization, and

support of the OSCIR camera system that was a fundamental aspect of the work

presented in this dissertation.


This dissertation focuses on the study of circumstellar disks around main-

sequence stars. The most fundamental question one can ask is, why is the study of disks

important? There are two facets to the answer to that question. The first deals with star

formation and the evolution of stellar objects during the initial stages of their lives.

Current theories of star formation like that of Andre (1994) suggest that the formation of

a circumstellar disk is necessary for the successful formation of a low or intermediate

mass (M < -5 M.) star. In this context, circumstellar disks are just one of a myriad of

supporting players in the process of star formation, but an important one. The disks

dissipate angular momentum from the system while bi-polar outflows remove excess gas

and dust from the close stellar environment in the first few million years of its life. With

the relative explosion of observations of circumstellar disks in the last 10 years, they now

seem ubiquitous. Also important is the fact that planetary systems form from these disks.

With the recent discovery of numerous exo-solar planets and planetary systems (Marcy

and Butler 1996), understanding the formation and evolution of circumstellar disks is

more important than ever. The research presented in this dissertation relates more to the

second facet presented above. Our research is centered on the study of circumstellar disks

around stars that are either in the very last stages of their pre-main-sequence (PMS)

formation, or have already moved onto (or past) the zero-age-main-sequence (ZAMS) in

the H-R diagram. These disks are the so-called "Vega-like," or debris disks.


0.01 -

10 100 1000
Wavelength [pm]

Figure 1-1: The spectral energy distribution (SED) of Vega. Solid line is a model
atmosphere. Dotted line is a modified blackbody fit to the data with T=73 K and
Q(X) a X'1' Flux estimates from ISOPHOT and IRAS. Submillimeter fluxes from
Zuckerman and Becklin (1993). Figure from Heinrichsen et al. (1998)

The IRAS satellite discovered the "Vega phenomenon" in 1984 when it was

performing what was thought to be routine calibration observations of the photometric

standard star, Vega. The surprising result returned by IRAS was that at all X> 12 gIm

Vega exhibited levels of emission that were several times higher than what was predicted

by model stellar atmosphere calculations. In Figure 1-1 we plot the SED of Vega to show

the shape and extent of this excess emission. This excess emission was attributed to

emission from dust grains in the vicinity of the star, presumably in the form of a disk.

Three more examples of this phenomenon were quickly discovered: a PsA, P Pictoris,

and E Eri; and together with Vega they became the archetypes for this class of object

(Aumann et al. 1984). Smith and Terrile (1984) proved the disk hypothesis with their

coronagraphic image of 3 Pictoris, which showed a nearly edge-on disk (i > 800) around

the star with dimensions of 100s of AU. The publication of this image jumpstarted the

field and the study of Vega-like disks has become one of the most heavily researched

aspects in modem astronomy; since 1984 there have been almost 1700 papers published

with the name "Beta Pictoris" in the title.

Since planets are thought to form in the dusty environments around stars, there

seems to be a growing consensus that the Vega phenomenon is intimately connected with

the occurrence of planetary systems. Indeed, we live inside our own Vega-like disk, the

so-called Zodiacal "cloud" disk around the Sun. Many of the characteristics of the Vega-

like disks are similar to our own dust disk, including the fact that the lifetime of dust in

orbit around the archetypes is much shorter than the estimated age of the central stars.

This means that the dust must be replenished in some manner, which hints at the

existence of hidden planetary companions in the disks that could supply the system with

fresh dust much like comets and asteroidal collisions do in the Solar system. In the last

few years evidence has been accumulating for the existence of exo-solar planets in some

Vega-like disks. High-resolution STIS images have revealed a warp in the disk of P

Pictoris (Heap et al. 1997), and subsequent near-IR (2 um) observations confirm that

warp exists and suggest that a planet with a -30 inclination could cause it (Mouillet et al.

1997). In the mid-IR (5 to 30 gLm) the disk of HR 4796A has been imaged at -0.3"

resolution by Telesco et al. (2000), Jayawardhana et al. (1998), and Koerner et al. (1998)

showing evidence for an inner hole of approximately Solar system dimensions.

Our motivation for this dissertation is to characterize a number of these Vega-like

disks in the mid-IR. To this end we used a modern mid-infrared camera system with large

aperture telescopes to conduct a survey of nearly 40 Vega-like stars in the mid-IR (10 and

18 gpm). In the following chapters we discuss our survey results, and place the observed

Vega-like sources into an evolutionary sequence. We also discuss two new disk examples

that we resolved in the mid-IR for the first time.

In the remainder of this chapter we summarize the current paradigm of low and

intermediate mass star formation and see that the formation of a circumstellar disk is

most likely a requisite step in the process. We then contrast the disks we see around

Vega-like stars with those around stars in earlier stages of formation, like the Herbig

Ae/Be stars. We end the chapter with a statement of the dissertation project and list what

we hope to add to the current status of Vega-like disk research.

Star Formation and Disk Creation

The formation of stars is the main way that material is removed from the

interstellar medium. Star formation is also one of the main topics of study in modem

infrared astronomy. The very first stage of star formation takes place in the cores of giant

molecular clouds (GMCs). The evidence that star formation occurs in these clouds is

strong. First and foremost, direct observation easily correlates regions of star formation

with dark clouds in the sky, for example in the constellations of Taurus and Ophiuchus.

Since the newly formed stars are by definition young, we believe that they have not yet

had sufficient time to move away from their birthplaces. Additional evidence comes in

the form of the correlation between the youngest stars known and the presence of

molecular CO emission (Churchwell 1991). In fact, Churchwell (1991) shows that there

is no place in the galaxy where active star formation is occurring where there is not

significant CO emission. Since CO is a major constituent of molecular clouds, it seems

very reasonable to think that stars form out of the material in these GMCs. The earliest

stage of the formation of a star is the fragmentation and collapse of part of a GMC. The

passing of a galactic spiral density wave, or a close encounter with a nearby star may

trigger the collapse. Whatever the trigger, the process of star formation is underway when

a fragment of approximately 104 AU "breaks-off' from the main cloud and starts its

collapse. The luminosity associated with these "protostars" is at first derived from the

gravitational energy associated with this collapse. At some point in this process there is

enough luminosity associated with the protostar that it can be observed directly. We

begin our more detailed discussion of star formation at this point following the work of

Lada (1987) and Andre et al. (1993).

In Figure 1-2 we show the different stages in the evolutionary sequence of young

stellar objects. On the right side of the figure there is a model spectral-energy-distribution

(SED) of the source. The middle panels show a sketch of the system, and the left side

lists relevant masses and timescales for each Class. A Class 0 source is the earliest

observable stage of star formation. Class 0 sources are only visible in the millimeter

regime or through the detection of CO spectrally. The dust shells surrounding these

sources are still so dense they are optically thick to even mid and far-IR radiation.

Because collimated CO outflows are often associated with Class 0 sources, it is believed

that a centralized core has formed by this stage and that matter from the surrounding shell

is actively accreting onto it. At this stage the shell is likely still more massive than the

protostar at its core.

i lumd 0101 0.1 1

I-' O iiM O4.V I C I1
'Cld Bladc Bo0y : O
Age s

1 2 10 100

""I I Me


perhaps 2 tm. There are still collimated outflows of molecular gas, for example CO,
1 2 10 M00

AVS 1'yri


Figure 1-2: Schematic of low and intermediate mass star formation by Lada (1987) and
Andre et al. (1994).

The flow of matter onto the protostar continues unabated past the transition into

the Class I phase, which occurs on timescales of _105 yr. At this point the formation of

the protostar is almost complete, and it is now detectable through the infrared, down to

perhaps 2 gim. There are still collimated outflows of molecular gas, for example CO,

associated with Class I objects indicating that the inner regions of the shell are still being

accreted onto the protostar. The amount of material around the star is still large, upwards

of 0.1 Me. Class I objects are also normally associated with extended millimeter

continuum emission. Some form of disk has likely formed around the protostar at this


By the time the system becomes a Class II source the millimeter continuum has

become much more compact, and there is rarely any evidence for outflows associated

with the source. The age of the system is now ~ 106 yr and the mass of the optically thick

disks associated with these sources is close to the minimum mass Solar nebula, -0.01

Me. At this stage the SED of the source is beginning to resemble that of a Herbig Ae/Be

star with strong excess emission detected down to 2 .Lm or less. This is relevant to our

work since it is believed that the Herbig Ae/Be stars are the progenitors of the Vega-like

stars that we study in this dissertation. Indeed, some of the SEDs of our survey sources

resemble those of the Class II objects shown in Figure 1-2, although most of our sources

are clearly Class III, or later.

The Class III stars have thin disks associated with them and their SEDs are

approaching those of normal stars. Depending on the mass of the star, Class III is reached

in the few x 106 yr timescale. In this stage there is little evidence for active accretion onto

the star and absorption lines are often seen in the spectra of these sources. The mass of

the disks around Class III stars is low: around 0.003 Me; much less than that in Class II

or earlier. It is around Class III sources that you find Vega-like disks. This connection

between a Class III and a Vega-like source is beautifully illustrated by comparing the

SED of the Class III source in Figure 1-2 with the SED of Vega in Figure 1-1.

Evolutionary Status of Vega-like Disks

As seen in the last section the Vega-like sources most resemble stars in the last

stages of star formation. To more clearly define what a "Vega-like' system is we use the

definition suggested by Lagrange et al. (2000). In their review they suggest the following

criteria for defining a Vega-like system

Ldus/L* << 1
M(dust+gas) << 0.01 M
Dust dynamics not controlled by gas ; Mgas << 10 Mdust
grain destruction times less than age of star

The first criteria is set to make sure that a massive protostellar disk is not present around

the star. Criteria 2 guarantees that the circumstellar material is optically thin. Criteria 3

makes sure that the dust is separated from any remnant gas left in the system and that the

motion of the dust grains are Keplerian modified by radiative processes. This should

exclude systems where dust is condensing in winds and jets. Criteria 4 that is crucial to

the definition because it ensures that the dust around the source is not primordial. That is,

the grains have been processed within the disk. It implies that the dust grains must be

somehow replenished from larger objects. This is important since it implies that larger

grains are present, and also hints at the existence of even larger, planetesimal sized bodies

capable of sending the smaller bodies into collision-producing orbits (Lagrange et al.


There is also some issue with the name "main-sequence" being an integral part of

the definition of a Vega-like system. The definition of Lagrange et al. (2000) above

defines the class in terms of the physical state of the circumstellar matter, not the exact

evolutionary state of the central star. The star itself may or may not have reached the

Table 1-1: Resolved Vega-like Disks

Source D ) Spectral Age
Name Type (yr)

EEri 3.2 K2V (5-10)x108
a PsA 7.7 A3V 1-3x108
a Lyr 7.8 AOV 3x108
pPictoris 19.3 A5V 1-2xl07
HR 4796A 67.0 A5V (8+3)x106
HD 141569 99.0 B9V 1xl07
HD 169142 145.0 A5V (3-5)x 106
BC+31643 330.0 B5V 106-107
SEstimates from this dissertation
SOSCIR estimate, see Telesco et al. (2000)

Disk Radius at Given Wavelength (AU)
0.5 Lm 2 gm 10-20 pm 60 pm 850 .m

S 60
S 80 160
S 100 125
800 120 150 250 250
70 70 -
400 100 --
> 1000
> 1000 -----

main sequence by the time its disk meets the above criteria. This is relevant to this

dissertation since we have investigated a number of stars that seem to be transitioning

from Class II to Class III. While the stars may not yet have reached the ZAMS, they

clearly exhibit characteristics reminiscent of main-sequence stars. Lagrange et al. (2000)

coined the term 'old pre-main-sequence' stars (OPMS) to describe these sources. They

are an interesting class of object since they represent the transition between the pre-main-

sequence stars of the star forming regions, and the main-sequence stars in the solar


Because the Vega-like disks are optically thin, tenuous structures, they are very

difficult to detect. In the 17 years since Smith and Terrile (1984) published their

watershed image of P Pictoris only seven other Vega-like disks have been directly

imaged. In Table 1-1 we list the Vega-like disks that have been resolved at any

wavelength. This is however, rapidly changing. New instruments coupled with large

aperture telescopes are giving us the opportunity to detect more of these disks than ever

before. This is made evident by the fact that 12 of the 14 size estimates in Table 1-1 have

been published in the last 4 years. In particular the latest generation of near and mid-IR

instruments allows us to investigate this class of object as never before. The sub-

arcsecond angular resolution achieved by the 8 to 10 m class telescopes lets us see down

to a few AU from the stars in the closer of the sample sources, and the sensitivities of the

latest generation mid-IR cameras give us the chance to detect fainter extended emission

from these disks than ever before. This is one of the motivating factors behind this

dissertation, to use the latest generation of infrared camera systems on the largest aperture

telescopes to investigate the class of Vega-like source as a whole.

This work is unique in one important way. Investigating these disks in the mid-IR

directly probes the region where planetary systems form in them. This work is

complementary to research done in the near-IR, sub-mm, and mm regimes. At the longer

wavelengths we are detecting the coldest dust in the systems, which has temperatures in

the tens of Kelvin range. This cold dust also orbits hundreds of AU from the stars, much

farther away than any known planetary companions. At the shorter wavelengths, the near-

IR, we detect two different kinds of radiation. Some sources exhibit thermal emission

from hot dust (T = 1 to 2000 K) very close (1 to 3 AU) to the stars (e.g., HD 169142;

Chapter 4). Others like HD 141569 (Weinberger et al. 1999) and P Pictoris (Kalas &

Jewitt 1995) are associated with near-IR radiation that is scattered from the dust in their

disks. This scattered radiation is seen out to hundreds of AU in these two cases and is

likely being scattered off the dust grains that are responsible for the far-IR/sub-mm/mm

emission detected around these systems. The mid-IR forms a bridge between these two

regimes. It can be said that the long wavelength emission comes from the "Kupier belts"

and perhaps "Oort clouds" of the Vega-like sources, while the near-IR emission comes

from the very inner parts of their disks. In this scheme the mid-IR emission would come

from the Zodiacal dust of the Vega-like systems, most likely tracing the region where

planets are forming, exactly where they formed in our own Solar System. For this reason

these disks are important to study, for by answering questions about them as a class, we

may well be able to answer outstanding questions about the formation and evolution of

our own Earth. In the next section we outline the work presented here.

Outline of the Project

This dissertation is centered on mid-IR (10 and 18 pLm) imaging of a large sample

of Vega-like stars. We undertook this project with the hoping to characterize a number of

disks at these wavelengths, and we succeeded by making the first detection of extended

mid-IR emission around two stars, HD 141569 and HD 169142. We did not detect any

extension around the remainder of the survey sample.

In Chapter 2 we discuss our sample, our observations of the survey sources, and

the data reduction steps we used. Chapter 3 presents our conclusions about the survey as

a whole and we place our sample sources into what may be an evolutionary sequence. In

Chapter 4 we discuss the newly discovered disk around the star HD 169142, and

investigate the properties of the dust in that disk through radiative equilibrium models.

We present similar results for HD 141569 in Chapter 5. For HD 141569 we derive grain

size estimates for the mid-IR-emitting grains, and compare our mid-IR images of this

disk to near-IR images from HST. In Chapter 6 we present observations that "de"-resolve

the source SAO 26804, a star previously thought to be associated with a Vega-like disk.

We show that it is likely that SAO 26804 is a luminosity Class III star and not on the


main sequence. Chapter 7 discusses the OSCIR instrument and its role in this

dissertation. We close with an Appendix on mid-IR observing and discuss the specialized

techniques used in that wavelength regime.


We have used front-line infrared instrumentation and large aperture telescopes to

survey approximately 40 Vega-like sources in the northern and southern sky. Most of the

survey was conducted at the Blanco 4-meter telescope at the Cerro Tololo Inter-American

Observatory (CTIO). Parts of the survey were also conducted at the NASA Infrared

Telescope Facility (IRTF) and the W. M. Keck Observatory (Keck). The project centered

on imaging each of the survey sources at 10 and 18 p.m using the Observatory

Spectrometer and Camera for the Infrared (OSCIR). The OSCIR system is discussed in

detail in Chapter 7. For most of the sources our images are the first mid-IR observations

of them since those of the IRAS satellite. A literature search for other mid-IR

observations of our survey sources returned only two further studies. Fajardo-Acosta,

Telesco, and Knacke (1993) imaged five of the sources at -4" resolution, and Sylvester et

al. (1996) presented mid-IR spectra of a dozen of the sources on our list. A small number

of our sources have questionable classification, and have been included in studies of

Herbig Ae/Be stars also. In particular, the studies of van den Ancker et al. (1998) and van

den Ancker (2001) contain members of our sample. Although the stars in our survey have

been studied in other wavelength regimes, our mid-IR imaging here represents the

highest resolution mapping of these sources to date. Our observations typically have sub-

arcsecond resolution limits with some as low as 0.3". At the average distance of our

survey stars (77 pc) at these resolutions we are tracing emission from dust that is tens of

AUs from the star, a region where planetary systems are born and evolve. We cannot yet

directly detect any planets around these stars. However, we can detect the effects of

hidden planets on the disks they inhabit. An example of this idea is the inner holes

predicted by the SEDs of some of the sources. An inner hole with a diameter of

approximately 50 AU was predicted in the disk of HR 4796A by Jura et al. (1998) and

was then directly imaged by Jayawardhana et al. (1998), Koerner et al. (1998) and

Telesco et al. (2000). Observing our survey sources in two mid-IR passbands also gives

us the opportunity to investigate the physical properties of the mid-IR-emitting grains in

these disks. Our 10 and 18 gm observations let us estimate the temperature of the grains;

and investigate their composition, size and distance from the stars. We also comment on

the dynamical properties of the grains and discuss the roles of Poynting-Robertson drag

and radiation expulsion on the evolution of the grains. In the next sections we discuss or

sample stars and describe how we reduced the data.

Source Lists

Our primary source list is that of Walker and Wolstencroft (1988) [hereafter

WW]. One of the first searches of the IRAS database after the discovery of the four

archetypes, WW is one of the primary lists of Vega-like sources used today. WW present

a list of 34 sources that have characteristics similar to Vega and the other archetypes.

Their selection criteria are shown below.

1. Sources must be associated with bright visual sources (SAO stars), but not
with a well-known emission-line star.
2. Sources must have a 60 gm/100 gm flux density ratio similar to that of the
prototypes. The range of the accepted ratios corresponds to black body
temperatures of the dust from 60 to 150 K.
3. Sources must show evidence for extension in one of more of the IRAS bands
in the IRAS Working Survey Database.

The 34 sources that met these criteria were separated by WW into four categories: the

prototypes, section A, section B, and those associated with objects in the Gliese catalog.

The prototypes are exactly that: a Lyra, a PsA, E Eri, and P Pictoris. Section A contained

the best candidates for disks similar to the prototypes. Sources were relegated to Section

B if there was evidence for mass-loss or emission characteristics in their spectra. The

sources associated with the Gliese catalog were separated from the others since they were

not stringently required to meet Criteria #3, however, most of them were flagged as

extended in the IRAS database. All of the sources in the Gliese category are also on the

Vega-like list of Aumann (1985). WW calculated temperatures for the dust using the

IRAS 25 and 60 ltm fluxes to fit the dust emission with a blackbody curve. By using

published B and V magnitudes to model the photosphere, WW plotted the SEDs of the

sources. Their plots give us a first look at the amount of excess associated with each

source, and we used them to help prioritize the sequence in which we observed the stars.

Our secondary lists are those of Sylvester et al. (1996) and Mannings and Barlow

(1998). The Sylvester et al. (1996) list was particularly useful since their selection

process was similar to what we had in mind for this project. They included objects on

their list that maximized their chances of determining disk and grain parameters. Their

sources are therefore bright in the infrared, relatively close to the Sun, and have large IR

excesses associated with them. Their list contains a total of 24 sources, of which we

observed 20 during our survey. The Sylvester et al. (1996) list does have some overlap

with the WW list, with a total of 12 stars common to both. A useful facet of the Sylvester

et al. (1996) research is that they presented mid-IR spectra for 11 of their sources taken

with CGS3 on UKIRT. We used these spectra to help prioritize our observations, and to

compare to our results on the disks of HD 141569 (Chapter 5) and HD 169142 (Chapter

4). The better defined SEDs of the star/disk systems in Sylvester et al. (1996) were also

very useful in our source selection.

The third list we used to select our survey stars from is that of Mannings &

Barlow (1998). Their method of source identification was based on the cross-correlation

of the Michigan Catalog of Two-Dimensional Spectral Types with the IRAS Faint Source

Survey Catalog. The Mannings and Barlow list contains some 60 new Vega-like sources

not contained on either the Sylvester et al. (1996) or WW list. Our primary use of the

Mannings and Barlow sample was to search for Vega-like stars at Southern declinations.

In the decade between the publication of the Walker and Wolstencroft (1988)

paper and the Mannings and Barlow (1998) list, significant work was done on the

spectral classification of the Vega-like stars. In 1988 some of the sources on the WW list

are listed as luminosity Class I, III or IV, and a few had no luminosity classification at all.

Mannings and Barlow (1998) note that today only about half of the sources on the WW

list sources are known to be on the main sequence. At first glance this seems to have a

negative effect on the completeness of our survey; however, this is not detrimental to our

purposes. An important part of our research is to place these sources into an overall

evolutionary sequence. To start to piece together that sequence we need to see these

sources in all stages of their evolution. This includes the time immediately before, and the

time just after they are on the main sequence. It is therefore not unreasonable to make

observations of PMS objects and post-main-sequence objects that meet the criteria of the

Vega-like surveys to fill in the gaps in this evolutionary sequence we are trying to

construct. As we see in Chapter 3, it is the fact that we made observations of sources with

questionable classifications, and varying amounts of IR excess that allows us to put

together such a sequence for our survey stars.

The sources we observed for our survey are listed in Table 2-1. In total we

observed 36 sources at both 10 and 18 gm. Three other sources were only observed in

one filter because of poor observing conditions. In Table 2-1 we present which list the

objects came from and see that in total we observed 25 sources from Walker and

Wolstencroft (1988); 9 unique sources from Sylvester et al. (1996); and 4 sources from

Mannings and Barlow. One star used as part of the survey, HR 4796A (HD 109573), was

not present on any of our source lists. We also list the equatorial coordinates (J2000), and

distance to each source as listed in the Hipparcos database. Only 5 of our sample stars do

not have a Hipparcos measured parallax, and therefore distance. Values of (B-V) for each

star from the Tycho catalogs are also listed. Using these newly measured distances for the

sources allowed us to calculate absolute magnitudes for the stars, a requirement in setting

up our evolutionary sequence presented in Chapter 3.


We used the OSCIR camera system exclusively for the observations in this

dissertation. OSCIR is a mid-IR camera/spectrometer system built and operated by the

University of Florida Infrared Astrophysics Group. The camera is built around a 128 x

128 pixel Si:As blocked-impurity-band (BIB) detector from Rockwell/Boeing. Chapter 7

more fully discusses the role of OSCIR in this dissertation. As noted previously we

collected data for this research on the IRTF, CTIO 4m, and Keck II telescopes. Because

of the different apertures of the telescopes some important characteristics of OSCIR

changed as we moved from site to site. Table 2-2 lists the parameters and their values on

each telescope. The survey was conducted using the standard "chop/nod" technique with

chop frequencies of a few Hz and chop throws in the range of 10" to 30." Appendix A

Table 2-1: Survey source data.
Spectral Distance RAa DECa
Source Name Class ()a (2000) ( M (B-V) Ref.
Class (pc)a (2000) (2000)

HD 432
HD 8538
HD 9672
HD 16908
HD 17848
HD 20010
HD 22049
HD 27290
HD 34282
HD 34700
HD 35187
HD 38678
HD 39060
HD 49662
HD 74956
HD 98800
HD 101584
HD 102647
HD 104237
HD 109573
HD 123160
HD 135344
HD 139614
HD 139664
HD 141569
HD 142165
HD 142666
HD 143006
HD 144432
HD 155826
HD 158643
HD 163296
HD 169142
HD 172167
HD 188037
HD 207129
HD 216956
HD 218396
HD 233517'



01 3437.7
03 1204.5
03 32 55.8
05 16 00.4
05 1941.4
05 46 57.3
05 47 17.0
06 48 57.7
08 44 42.2
11 49 03.5
1406 12.8
15 1548.4
1541 11.3
15 53 53.9
14 50 52.7
17 31 24.9
18 24 29.7
21 45 01.1
22 54 53.5
23 07 28.7
08 22 46.7

+59 08 59.2
+60 14 07.0
-15 4034.8
+27 42 25.7
-62 38 23.5
-28 59 45.7
-09 27 29.4
-09 48 35.4
+05 38 42.7
+24 57 37.5
-1449 19.0
-51 0403.5
-24 46 39.7
-55 34 25.8
-55 34 25.8
-78 11 34.5
-3952 10.1
-37 09 16.2
-42 29 53.5
-44 39 40.3
-03 55 16.3
-2431 59.3
-22 01 40.0
-22 57 15.2
-27 43 09.8
-38 35 38.0
-23 57 45.5
-21 5721.8
-29 46 49.3
+22 27 14.2
-4731 55.8
-29 53 15.7
+21 0803.3
-34 27 06.0





SSpectral classifications from Mannings & Barlow (1999).
a Data taken directly from the Hipparcos and/or Tycho catalogs.
t Ambiguous distance and absolute magnitude, see Chapter 6
+ Distance from Walker & Wolstencroft (1988) or Sylvester et al. (1996)

Table 2-2: OSCIR/Telescope Characteristics
e Aperture Focal Field of Chop Freq.
Platescale ie Throw hop Freq.
rech (m) Ratio View (Hz)
) Ratio (arcsec) Vie (arcsec)
IRTF 3 F/35 0.223 29 30 3-5
CTIO 4 F/30 0.184 23 25 5
Keck 10 F/40 0.062 8 10 3

gives a thorough explanation of these standard mid-IR observing techniques and more

fully describes why these observational parameters were chosen.

During the planning stages of the survey we decided to observe all of the sources

in a consistent manner, with follow-up observations made when warranted. For the

survey observations, we made deep N (Xo= 10.8 Jim, AX = 5.2 gpm) and IHW18 (Xo= 18.2

gim, AX = 1.7 pm) integration with exposure times ranging from 10 to 20 minutes of

chopped-integration (5 to 10 minutes on-source) in each filter. Due to the inherent

inefficiencies of chop/nod observing the elapsed time during an observation is

approximately a factor of three longer than the on-source time. We felt that this exposure

time allowed us to efficiently use the allocated telescope time while giving us a good

chance to detect any extended emission associated with the survey sources. Given these

integration times the possibility exists of faint extended emission around some

of the survey stars that we did not detect. However, to keep our dataset consistent we

observed all of the sources in a coherent manner. Next we describe our observational

technique used for the survey observations.

Observational Strategy

Observations for this work were made over a period of three years at the above

mentioned telescopes. We followed a set observing strategy for each of the observations

made. When observing a science source we also made a series of observations in support

of the actual data on the Vega-like source itself. The sequence of observations made is as


1. Observation of photometric flux calibrator
2. Observation of a nearby point-spread-function (PSF) star
3. Observation of the science source
4. Observation of the PSF star

There was normally no need to observe a 'pointing' star for these observations as all of

our science sources are visually bright and easily detected with a 3 m class telescope. We

attempted to start each Vega-like sequence with observations of a photometric standard at

both 10.8 and 18.2 mrn, however, due to time constraints this was not always possible.

Since we were trying to determine if the science sources were resolved or not, the

observations of the PSF star were as critical as those of the science source. As illustrated

in Figure 2-3 our determination of whether or not a science source was resolved is based

in part on our comparison of scans through the science source and PSF star. To make as

accurate a comparison as possible, during the data reduction process we always used the

PSF observation that was taken closest in time to the science source observation. The

data was reduced in this way to combat the temporal changes observed in the PSF of the

various telescopes. In the next section we give an example of these changes. The PSF

stars were chosen to be bright, nearby stars with no excess associated with them. By

selecting K or M class stars with mv < 5 we could normally find a suitable star to use

within 100 of the science object. In some cases the PSF stars were up to 150 away. Our

chopped integration times for the PSF stars were 2 to 4 minutes (1 to 2 minutes on-


Nightly Variation of the PSF

During the planning stages of this work we attempted to form a viable

observational program that took into account some of the obstacles we knew we would

face. One of the problems we foresaw was the overall stability of the PSF of the

telescope-camera system. Many factors including slumping of the telescope structure,

changes in the ambient temperature in the dome, and issues with the mirror support

systems, can contribute to the instability of the PSF of the system. On modem telescopes

like Keck and Gemini, active control of the mirror surface mostly compensates for these

effects. However, even with these systems in operation some changes in the PSF are

evident. To illustrate this point we present Figure 2-1, which is a plot of scans through all

of the PSF measurements taken with OSCIR on the night of 04 May 1999. On this

particular night of observing at Keck II we obtained five measurements of the PSF at 10.8

gtm, and three at 18.2 plm. Inspection of the figure shows the magnitude of PSF changes

for the night clearly. Although it is somewhat difficult to distinguish the lines, the

sequence of 10.8 p.m measurements goes from the narrowest scan, made early in the

evening, to the broadest, made close to dawn. The 18.2 p.m PSF scans follow the same

trend. This implies that the PSF of the telescope became markedly worse as the night

went on, illustrating the reason why we used the PSF taken as closely as possible in time

to our science observations.

Note however that even though there is significant difference in the shape of the

PSFs within each filter throughout the night, the FWHM of the PSF is relatively

unaffected by these differences. From best to worst case the FWHM of the PSF changed

10.8 micron PSF measurements 18.2 micron PSF mcasurmcnts

0.4 81

Offset (arcseconds) Offset secondsd)

Figure 2-1: Nightly variation of the point-spread-function (PSF) at 10.8 gm (left) and
18.2 lm (right). Plotted are scans through the five PSF measurements at 10.8 gm and the
three measurements at 18.2 gm on the night of 04 May 1999, with OSCIR on Keck II.
The earliest measurements of the night are the narrowest, later measurements were
progressively broader.

by 0.15" at 10 lm and 0.11" at 18 gm. Similar analysis done for other nights show that

these values are slightly higher than normal. For most nights the FWHM of the PSF was

stable to within -0. 1" at Keck, and that is from the best to worst case on a night. The

variations between sequential PSF measurements were lower than this, implying that the

changes to the PSF were gradual, akin to a slow drift that occurred over the course of

hours. This is encouraging since the PSF measurements we used to compare to our

science sources were normally taken within 30 minutes of the end of the science source

observation. Along with the PSF star measurements we also observed infrared standard

stars to absolutely calibrate our data. In the next section we discuss our data reduction

steps, including that calibration.

Data Reduction Steps

Over the course of making our survey observations we acquired a large amount of

data. Including the PSF and standard star observations approximately 12 GB of data was

amassed for this project. The sheer amount of data alone made the data reduction arduous

and time consuming. The reason that the amount of raw data is so large is because the

OSCIR data is saved in specialized 6-dimensional FITS files. This unusual data structure

is required due to the chop/nod method of observing in the mid-IR (see Appendix A).

Briefly, the OSCIR detector is readout approximately every 10 ms during normal

observations. The data from the readouts is saved in buffers in a VME control crate on

the back of the telescope. Once every 2 seconds the buffers are sent down a fiber optic

cable connecting the VME crate and the control computer. Once the data is sent to the

control computer it is written to disk in one of these 6d FITS files. The advantage of this

is that the observer has access to the entire data stream of the observation after the fact.

While this produces data files in the 20 to 50 MB range, having access to the data in 2

second quanta is an invaluable asset during the data reduction. All of the data reduction

for this dissertation was done in IDL, using the "f6tools" to access the OSCIR 6d FITS

files, and the "wtools" to reduce the data in a GUI environment.

Removal of Contaminated Data

One of the advantages of having access to the entire data stream is evidenced in

the very first step of any OSCIR data reduction, the removal of contaminated data sets.

The advantage is that we can remove data in 2 second increments without changing the

overall quality of the data as a whole. The main reason that portions of the data stream

must be removed is due to fluctuations in the background. As explained in Appendix A,

44.5 -

44.0 -



425 -

0 100 200 300 400
Frame Number

Figure 2-2: Background at 10.8 pm during an observation of HD 141569 at Keck II.
Here we plot %well (a measure of how close to saturation the OSCIR detector is) vs.
frame number for an observation of HD 141569 at Keck II. During the reduction process
frame numbers 145 to 200 were removed due to the large background change.

accurate removal of the background is paramount in mid-IR observations. Using the

routine 'f6bstat' we produced plots for each survey star observation like the one shown in

Figure 2-2 above. In this figure we plot '%well' as a function of frame number for an

observation of the science source HD 141569 at Keck II. The quantity '%well' is a

measure of the background seen by the OSCIR detector during each of its 10 ms

integration. Experience has shown that an acceptable background variation is < 0.5%

change in '%well'. In this case we therefore remove the frames between number 145 to

200 because of the large change in well during those readouts. This large change in

'%well' was most likely due to a faint band of cirrus clouds moving through the field of

view of OSCIR. Once any contaminated frames are removed from the data we move onto

the next step of the reduction, which is the creation of a single image from the individual

frames in the 6d FITS file.

Cross-Correlation of Images

Once any contaminated frames are removed from the data, we have a series of so-

called 'sig' frames to work with. These are the final products of a complete nod-cycle, a

single 128 x 128 image that has the background from the sky and telescope fully

removed. For our survey mode observations we typically had 10 to 15 of these 'sig'

frames to work with for each filter.

For sources that are not bright enough to be seen in each individual 'sig' frame,

we performed what is called a "straight-stack". A straight-stack is the direct sum of all of

the sig frames in an observation, no shift-and-add processing done to the data. Most of

our sources were relatively bright in the mid-IR and were detected in the individual sig

frames. For these sources we developed and used two IDL routines that use a chi-squared

minimization kernel to cross correlate the images to find the best offsets between them

before stacking them into a single image. We find that this helps remove any tracking

errors or other systematic drifts in the data. Once the sig images are straight-stacked or

cross-correlated we move to the next step of the reduction sequence, ensuring that the sky

is zero-mean before we calibrate the image.


The skyfit routine is used to ensure that the off-source portions of the final image

are zero-mean. Due to imprecise background removal during some of our observations

the final image from a 6d FITS file may have a non-zero mean sky. To correct for this we

use the skyfit routine written by Drs. Jim De Buizer and Robert Pifia. This routine allows

you to mask any portion of the array, for example around your source, and then it fits the

remaining, unmasked, background with a 2 dimensional polynomial surface of specified

order typically 3,3 to 'flatten' the sky. This routine was used on all of our survey data

with good success. Once we have a zero-mean sky image, we are ready to calibrate.

Calibration/Airmass Correction

Our science observations were absolutely calibrated by observing infrared

standard stars as part of our observational sequence. In Table 2-2 we list the fifteen

primary standards that we observed during the survey. We also present their assumed flux

densities at 10.8 and 18.2 gm. For most of the standard stars we have model atmospheres

from Cohen et al. (1999). The flux densities for these stars were calculated by integrating

the models for the star through the passband of the OSCIR filters. We used the following

relation to calculate the bandpass averaged flux density standard star flux density Fx.

Fv (x) -(Tfilter() -Tatran (k)) dX

Fk :=--^-c---2----

2- (Tfilter(k) 'Tatran ()) dX

where Tfilter and Tatan are the transmission curves of the OSCIR filter and the atmosphere

respectively. XI and X2 are the wavelengths where the transmission of the filter drops to

zero, and c is the speed of light. On average we observed standards at both 10 and 18 p.m

3 or 4 times per night during the survey observations. Each of these data was reduced and

a calibration ('cal') value that converts the instrumental units of ADU/sec into the

physical units of mJy was calculated. The 'cal' value was then applied to the reduced

science data to create maps in mJy/pixel.

We comment here on the uncertainties associated with the photometric calibration

since they are the dominant uncertainties we dealt with for the survey observations. The

effects of atmospheric transmission always dominate the photometric uncertainties in

mid-IR. As Figure A-l shows (see appendix A), the mid-IR atmospheric windows at 10

and 20 pim are riddled with atmospheric absorption features from water vapor, C02, and

03. A variation in the column density of these species has a dramatic effect on the

transparency of the atmosphere in the mid-IR. Indeed, it is these variations of the sky

background that requires the use of chopping in the mid-IR.

Van der Bliek et al. (1996) measure short term variations in the atmospheric

transmission of -10% for the 10 tim window, and upwards of 15% in the 20 gim window.

Looking at the calibration data directly related to our survey, which consists of > 60

individual stellar observations, we find that the overall uncertainty in the absolute

calibration is 10% at both 10.8 and 18.2 gm. The discrepancy with the van der Bliek et

al. (1996) long wavelength conclusion is most likely related to the fact that our 18.2 gLm

filter has a bandpass on only AX = 1.7 gm; relatively narrow when compared to the

broadband 20 lim filters discussed in their paper. We adopt 10% uncertainties for all of

our measured OSCIR flux densities.

The airmass corrections in the mid-IR are generally small. Under good sky

conditions, a rule-of-thumb measure for the airmass correction in the N-band (10.8 glm) is

- 0.02 mag/airmass. Even so, we attempted to perform airmass corrections for all of our

Table 2-2: Standard Star Flux Densities
Flux Density (Jy)

Standard 10.8 glm 18.2 gjm
Star (N-Band) (IHW18)
aLyrt 37.8 11.9
a CMa 130.7 41.1
a CrB 5.0 1.8
a Boot 682.7 219.1
a Hyat 125.2 41.5
a Taut 600.6 200.1
pAndt 245.9 83.9
pGemt 115.5 37.0
SGru 906.0 323.1
0Pegt 352.1 122.2
yAqlt 77.2 25.6
yCru 833.9 286.7
yRet 72.6 28.7
Vel 194.6 69.4
t UMat 99.1 33.9

t Values are bandpass-averaged flux densities
calculated by integrating Kurucz models of each star
through the OSCIR filter bands also taking into account
the detector QE, and atmospheric transmission. All
other values come from the CTIO standard list.

survey data by using the observations of the photometric calibrators over a range of

airmass. At times this was impossible due to insufficient standard star observations, or

obviously contaminated standard data where there was no trend at all evident in the cal

values. When there was sufficient data of good quality available we used a linear

regression of log(cal) vs. airmass to correct the science observations for airmass. These

corrections rarely made a more than 5% change in the flux estimate for a survey source,

well within the 10% photometric errors assigned to our OSCIR photometry.

Flat-field Correction

Figure 7-6 shows an image of the OSCIR detector under uniform illumination in

the N-band (10.8 gm). There are approximately 10 hot pixels distributed across the array

that have significantly different response than the others. There are also about 10 dead

pixels. Overall the uniformity of the detector is very good, however there is a small

amount of vignetting in the lower right corer of the array. The response of the array is

generally very flat though, the flat-filed varies < 5% across the array. All of the structure

seen in Figure 7-6 is 'fixed-pattern' and is removed during the chop-nod process.

Because of the flat response of the array, no flat filed corrections were applied to the

survey data.


The last step in the reduction was to conduct aperture photometry on the science

sources to estimate their brightness at 10.8 and 18.2 gm. We used two methods to

perform the photometry on the survey stars. The first is an automated routine, which

estimates the sky background, and automatically grows an aperture around a point source

until the increase in flux obtained is less then or equal to the measured noise in the

background contained within the annulus bounded by the last two apertures. This routine

works very well for strongly detected sources and was used for the photometry for 75%

of the survey data. We also used this routine when we manually reduced standard star

data, or obtained flux estimates for PSF stars. For faint sources we used a manual

aperture for the photometry. For consistency, all manual apertures used had a radius r =

2". The measurement errors associated with the photometry (i.e. opix(Npix)112) were

always dominated by the 10% photometric error for all survey sources. Our observed

color corrected OSCIR flux density estimates for the survey sources are presented in

Table 3-1. In Chapter 3 we also discuss the removal of the stellar photosphere from those

values to calculate a values for the excess emission from the circumstellar dust.

Color Correction

Flux density estimates for objects observed through wide bandpass infrared filters

need to have a correction applied to them called a 'color correction'. This is particularly

true of the OSCIR N-band filter that has a bandpass of 5.2 pm. The color correction must

be applied since the spectral shapes of the flux calibration star and science object are

different through the bandpass of the filter. Because stellar temperatures are high (> 5000

K), the SED of a calibration star peaks well short of the mid-IR and its SED is on the

Raleigh-Jeans tail through the OSCIR filters. For our science sources though, this is not

be the case. For our science targets we are observing dust that has a temperature in the

hundreds of Kelvin. Because of this the SED of the science source may be peaking, or

has not yet peaked, in the mid-IR which means that its spectral shape through the

passband of the OSCIR filters is different than that of the calibration star. This is an issue

because we prefer to discuss monochromatic flux densities in the mid-IR, and since the

calibration star and science objects have different SEDs through the filter, any

monochromatic flux density given is dependent on the spectral shape of the source

through the passband. Since we also make the assumption that the SEDs of our science

objects follow a modified blackbody function (Fv = (1 e ). By(T)) through the

passband. The color correction term is applied to scale that function to give us the correct

in-band flux for the science source.

We outline the procedure for this correction below starting with the definition of the

effective wavelength of the N-band filter (Lef). Since the OSCIR detector is a photon

counter for these purposes ~ff is defined in terms of the photon with the average energy

in the bandpass.

The color correction factor assumes that the ratio of instrumental counts for the

calibration star (ADUS) to that of the program object (ADUP) is equal to the ratio of

number of photons detected from the calibration star (Ny) to the program source (NyP).

This equality is shown below in equation 1.

U (1)

the number of photons a source will generate through a bandpass is

N := Q sys-dv (2)

Where Fv is the flux density of the source with temperature T and optical depth tv

and is defined as

F,:= (-e /n-Bv(T) (3)

We also define the 'quantum efficiency' of the system as

Q sys:= T filte )T atran(X) QE(X)

where this relation takes into account the QE of the OSCIR detector, the atmospheric

transmission, and the transmission of the OSCIR filter. Substituting (3) into (2) gives

N :=l-e )-B V(T) (4)
:= Qsys hv

If we are given a monochromatic flux density of the source at some frequency, Fvo using

equation (3) and solving for 0 we can define the following relation

-( )vo( (5)
(Ie- VO)-B v(T)

substituting equation (5) into (4) gives

FVO (1-e -.Bv(T)
N:= Qsys e-) ) dv (6)
(1 e Bvo(T) hv

which is an expression for the number of photons produced by a source with flux density

Fv through the passband of the filter. Recalling equation (1)


we can use equation (6) to relate the number of photons received from the program

source (Np) to the counts measured from the program source (ADUP) and the number of

counts measured from the calibration star (ADUS). First solving (1) for (N/) and then

substituting in expressions for the number of photons received from the star and program

source from (6) gives

F F (I-e -B; AD FO B )
SQ s -- B dv A:=D Q Sys ) dv (7)
S- e- SYS hv BADLB ) sys hv

the absence of the (1 e~) terms on the right hand side of equation (7) is due to the fact

that the calibration star is assumed to be a simple blackbody. We can now solve (7) for

the flux density of the program object at vo, (FvoP)

Bv(T) i
y BV (Ts) hv
F vo := ADU vo J
\ADUS (8)
(1 e- T) -B VT
Q sys ----- dv
1 e o YvoS ( e B (TV) v

In equation (8) we have equated the flux density of the program source at the

monochromatic frequency vo to three known quantities, ADUP, FvoS, and ADUS. The first

term on the right hand side of (8) is the measured instrumental counts of the program

source. The second term is simply the ratio of the given flux density of the calibration

star at vo the measured instrumental counts for the calibration star. This is the "cal" value

mentioned in the calibration section. The third term is the color correction term. It is this

factor that is applied to the data to account for the different spectral shapes of the

calibration star and program object through the bandpass of the filters.

To color correct our fluxes we used a program written for MATHCAD. Using our

raw 10 and 18 pLm flux density estimates as a starting point the code iterates to

simultaneously solve for T, T, and color correction. All of our flux density estimates

presented in this dissertation are color corrected. For the sample as a whole the mean

color correction for our broad band 10.8 gm flux density estimates was approximately

13%. The 18 gm estimates had a mean correction of -3%. The 18 gim value is lower due

to the much narrower passband of the 18 gim filter. Color corrections for 10.8 gim (N-

band) fluxes on the order of 15 to 20% are consistent with the work of Hanner et al.


Resolved or Unresolved: The Question

Perhaps the most fundamental question we attempted to answer in this

dissertation is, are our sample Vega-like sources resolved? For the most part, the answer

to that question is no. We find that -90% of our survey sources are unresolved at both

10.8 and 18.2 gm. However, in this section we only discuss how we answered the

question of resolution. The full results of the survey are presented in Chapter 3.

Our primary method of determining whether or not a survey source was extended

was close inspection of the reduced data by eye. We found early on in the reduction of

the survey data that the eye is the best detector for faint extended emission near bright

stellar sources. This is especially true when the data is reduced with the latest generation

of data reduction software. The ability to scale and color data in a point-and-click

environment is a very powerful tool at our disposal and we used it to the fullest extent we

could during this project. For each of our survey sources we carefully studied the data as

closely as possible by eye to look for hints of extended emission near the science sources.

Visual comparison of the science sources and PSF stars was also performed for each

survey source.


ar 1.0
a b

0.8 i 0.8
S0.6 0.6


S/ d

0 0.2 O 2 -, 0 2:
.. yr. / .....

I t] .- : I :, -

Are~-)ds AnrAsewnds

Figure 2-3: Scans of a resolved (a & c) and unresolved source (b & d). (a and c) Scans of
HD 151469 and PSF star at 10.8 pm and 18.2 ptm respectively. HD 141569 is the solid
line, the PSF star is dashed. The dotted line is a Gaussian fit to the data. (b and d) Scan of
HD 16908 at 10.8 pm (b) and 18.2 p.m (d). 18.2 ptm data for HD 16908 and PSF star have
a 2-pixel (0.12" FWHM) Gaussian smooth applied to boost the signal to noise of the HD
16908 detection.

Another way we investigated the data to try to answer the question of resolution

was through scans (i.e. line-cuts) through the science objects and corresponding PSF

stars. We find the best way to visualize the data in this form is through plots like Figure

2-3. In Figure 2-3 we show scans that are representative of a resolved and unresolved

source. The resolved source (panel a & c) is HD 141569 which we observed at Keck II

in 1999 May. This source is discussed in detail in Chapter 5. We contrast HD 141569

with a representative unresolved source, HD 16908. Also observed at Keck, the HD

16908 data has the same resolution as the HD 141569 observations.
OAnohrwyw netgtdtedt otyt wrteqeto frslto
wa3hog cn ie iect)truh h cec bet n orsodn S
stars. We ind the best way to visualize the da4ta nti omi hog lt ieFgr
2-.IZ iue23w hwsasta r ersnaieo eovdadursle

One immediately sees the differences in the two sets of panels. HD 141569 is

clearly broader than the PSF star at both 10.8 and 18.2 pm while the HD 16908 and PSF

star scans lie on top of each other. It is the extended emission of HD 141569 that causes

this broadening of its scan with respect to the PSF star. Close inspection of panel (a)

shows that there is significant extended emission from HD 141569 detected out to a

radius of ~1". At 18.2 plm the noise in the source scan prevents a determination of the

exact extent of the extended emission, however, even with the relatively low signal-to-

noise of the HD 141569 18.2 pLm data the difference in width of it and the PSF star is

easily distinguishable in panel (c). Images of HD 141569 also exhibit a 'fuzziness' that is

not present when observing a true point source. This effect is difficult to describe, but it is

very evident when looking at the data on a monitor.

The scans of HD 16908 are representative of the unresolved sources in the survey.

There is virtually no difference in the FWHM measurements of HD 16908 and the PSF

star at either 10.8 or 18.2 gpm. There is also no evidence for faint extended emission at the

lowest levels of panels (b) and (d). Using these scans in conjunction with an in depth

inspection by eye, we can confidently conclude that HD 16908 is unresolved at this

angular scale at both 10.8 and 18.2 pm. In the next chapter we move onto the discussion

of our survey data.


In this chapter we discuss our survey data. We do not give further details on the

mechanics of making the survey, rather we discuss our overall interpretations of the data.

In the first sections of this chapter we present size limits on the disks of the unresolved

sources, and give our derived grain parameters for their dust. We then compare the

unresolved survey sources to p Pictoris and HR 4796A, and show that if the survey stars

were at the distance of 0 Pictoris (19 pc) about 50% would still have been unresolved in

our observations. This comparison allows us to answer the question of whether the

unresolved sources are unusual as compared to archetypal disk examples, and gives us an

idea of how the unresolved sources fit into the Vega-like class.

In the later sections we discuss trends seen in the data and show how they can be

used to form a probable evolutionary sequence for Vega-like stars. By breaking the

survey sources into three classes and using the H-R diagram, we can graphically see how

these stars may be evolving. We bring in evidence that supports such a sequence in the

form of the SEDs of the star/disk system and relate the evolution of those SEDs to the

evolution of the disks around the stars. Using one of the sub-classes as a pivot point, we

see how one group of stars is likely still moving onto the ZAMS, while the majority of

them have already made it there and are now moving across the main-sequence in the

middle stages of their lives. In general we have incorporated ideas that are already well

represented in the literature into our survey discussion while adding our own ideas to the

overall evolutionary scheme. We start the chapter with the presentation of our disk size

estimates and a discussion of their implications.

Disk Sizes

Perhaps the most fundamental piece of information we can gain from our

observations are size estimates for the mid-IR-emitting region of the disk around each of

the survey stars. Once we have these estimates we can then compare them to the disks of

the archetypal Vega-like stars to piece together a more complete picture of the scale sizes

of this type of disk. As was mentioned in Chapter 1, the majority of our survey sources

are unresolved at both 10 and 18 jim. In fact, we observe 34 out of 38 sources as

unresolved within the uncertainties associated with our observations. Only HR 4796A,

HD 169142 (Chapter 4), HD 141569 (Chapter 5), and possibly 49 Ceti were resolved

during the survey.

Since over 88% of the sources are unresolved we can only place upper limits on

the size of their disks in the mid-IR. We do this by using the observed FWHM of the

emission as the limiting size of the mid-IR-emitting region of the disk. Using this limiting

size in conjunction with distance estimates to the stars allows us to translate the size into

physical units. Since most of our Vega-like sample stars are have distances < 200 pc, and

are bright in the visible, most of them were observed with the Hipparcos satellite. Indeed,

we have Hipparcos measured parallaxes, and therefore distances, for all but 5 of the stars

in our sample. As we will see later in the chapter the fact that we have accurate distances

for most of the stars is one of the main reasons that we can put these stars into an

evolutionary sequence. For now we use the distance estimates to help us place limits on

the size of disks associated with the unresolved sources in the sample.

Our limiting disk sizes are listed in the last column of Table 3-1. For each star in

the survey we present the observed upper limit on the diameter of its disk. Although

using the observed FWHM to limit the disk size is the least constraining method, for the

majority of the sources these limits are the tightest existing constraints on the size of their

disks in the mid-IR. For sources that are resolved we list approximate limits on the sizes

of their mid-IR-emitting regions (denoted with '=' in the limiting size column). Our size

estimates for these disks are only approximate since there may be faint extended emission

from the disk that we did not detect in our survey-mode observations. An observational

program centered on deep mid-IR imaging of these systems will more accurately define

the limits of emission for these four disks. Many of the sources have not been observed in

the mid-IR since IRAS, and those that have been observed were mostly observed with

single element detectors or at low spatial resolution. Walker and Heinrichsen (2000)

observed nine of the sample with ISO, but only at 60 and 90 pLm and at relatively low

spatial resolution. The only other observations of a significant number of the survey stars

is that of Sylvester et al. (1996), in which mid-IR spectra of 12 of the sources were

presented, but no imaging.

Table 3-1 also lists our color corrected observed flux densities of the survey

sources at 10.8 and 18.2 pm. The flux densities presented in Table 3-1 are the sum of the

photospheric emission of the star and the thermal emission from the circumstellar dust.

We present these values here in this manner so that they can be directly compared to

other observations. For our analysis of the dust properties we remove the stellar

Table 3-1: Observed Properties of the Survey Sources

Source F10.8pm F18.2 pm IRAS IRAS IRAS IRS Limiting
Name (Jy) (Jy) 12m 25tm 6m 100m Size (AU)
(Jy) (Jy) (Jy) (Jy)

HD 432 13.5 3.59 11.7 2.87 1.00 <12.7 16
HD 8538 5.69 1.42 4.96 1.14 0.35 <6.09 27
HD 9672 0.36 0.14 0.33 0.38 2.02 1.88 =150
HD 16908 0.45 0.09 0.38 0.41L 0.36 <1.87 31
HD 17848 0.25 0.23L 0.46 0.25L <0.40 <1.00 51
HD 20010 3.09 0.77 2.85 0.74 0.23 <1.00 14
HD 22049 10.5 3.50 9.52 2.65 1.59 1.87 3
HD 27290 1.34 0.24 1.20 0.37 0.28 <1.00 20
HD 34282 0.63 0.32 0.70 1.63 10.8 10.7 173
HD 34700 0.48 1.40 0.57 4.42 14.1 9.38 900
HD 35187 3.86 6.02 5.39 11.5 7.95 5.00 200
HD 38678 1.80 0.63 2.12 1.15 <0.53 <1.00 35
HD 49662 0.19 0.13L 0.43 1.62 4.60 5.68 180
HD 74956 8.08 3.54 8.21 1.99 0.42 <8.76 40
HD 98800 1.58 5.09 1.98 9.44 7.28 4.46 4
HD 101584 96.5 113.0 92.6 138 179 99.1 800
HD 102647 7.45 2.70 5.26 1.75 1.02 <1.00 10
HD 104237 17.8 8.50 22.5 22.7 13.5 9.21 115
HD 109573 0.19 0.91 0.45 1.39 7.36 3.81 =200
HD 123160 0.66 0.14 0.60 0.38 3.11 4.41 10
HD 135344 1.58 2.69 1.57 6.91 25.6 25.7 60
HD 139614 4.26 8.05 4.11 18.14 19.3 13.9 68
HD 139664 1.11 0.61L 1.42 0.69 0.59 <1.00 15
HD 141569 0.29 0.65 0.53 1.82 5.54 3.48 =200
HD 142165 0.32 0.06L 0.25L 0.34 <2.77 <11.6 40
HD 142666 10.5 9.10 8.67 11.5 7.23 5.46 35
HD 143006 1.02 1.92 0.87 3.06 6.57 4.82 50
HD 144432 9.85 6.83 7.62 9.23 5.77 3.29 105
HD 155826 3.49 4.05 4.12 5.30 8.63 <260 20
HD 158643 20.0 11.8 15.6 10.1 1.06 <5.97 65
HD 163296 26.3 17.4 18.2 20.9 28.2 <40.6 60
HD 169142 2.72 8.14 2.95 18.4 29.5 23.4 =175
HD 172167 36.4 12.0 41.5 11.3 9.01 7.54 4
HD 188037 295 146 227 113 11.0 12.3 70
HD 207129 1.08 0.39 0.90 0.20 0.23 <1.00 12
HD 216956 17.1 4.62 18.5 4.79 9.11 11.1 21
HD 218396 0.37 0.08 0.40 0.24 0.41 <2.59 15
HD 233517 0.48 2.02 0.48 3.43 7.60 5.10 --

We adopt 10% errors on the OSCIR photometry at 10.8 and 18.2 Alm.
Denotes source with near-IR excess as listed by Sylvester et al. (1996)
L in flux column denotes upper limit used in temperature and optical depth calculation

contribution from the observed flux density. This procedure and our results are discussed

in the next section. For the few sources that we did not detect at either 10.8 or 18.2 Aim,

we list a 3c upper limit in the appropriate column in Table 3-1. These limits were

calculated using an aperture with radius = 1.22.-(/D). Table 3-1 also lists the IRAS 60

and 100 jim flux density estimates from the IRAS Point Source Catalog v2.0 in for each

source. At these longer wavelengths these is negligible contribution to the flux from the

star, it is comprised entirely of thermal emission from the circumstellar dust. All of the

sources in our sample have significant excess emission at these long wavelengths, and the

IRAS fluxes are included to give a more complete picture of the magnitude of the

excesses associated with each star. In Table 3-1 we also mark the nine systems that

exhibit near-IR excess. As we will see, these nine systems form one of the three sub-

classes we define when discussing our proposed evolutionary sequence for the survey

sources. In the next section we explain the procedure for removing the stellar contribution

to the flux density estimates presented in Table 3-1. This is necessary since this

procedure directly bears on which sources we could accurately model.

Photosphere Removal

As mentioned above the flux density estimates presented in Table 3-1 contain

both emission from the photosphere of the star and the thermal emission from the grains

in its disk. Since we are trying to investigate the properties of the dust, we must separate

these two contributions so we know as accurately as possible how much radiation the

grains themselves emit at 10.8 and 18.2 Lm. While this procedure is straightforward, the

uncertainties associated with both the predictions of the photospheric emission and our

observed flux density estimates have consequences on the final results. The procedure we

used to calculate the photospheric contribution of each star to our observed flux density

estimate is outlined below.

1. Use a published value or spectral type of the star to find the effective
temperature of its photosphere (Teff).
2. Model the SED of the stellar photosphere as a blackbody function with
temperature Teff, and normalize the blackbody function to observed 2.2 pm
(K-band) values.
3. Use the blackbody function to predict the fluxes emitted at 10.8 and 18.2 gm
for each Teff. Then form the ratios (F1o0.8s mF2.2 pm) and (F18.2 jm/F2.2 gtm) as a
function of Teff.
4. Multiply the observed 2.2 pm flux value by the correct ratios to predict the
photospheric flux at 10.8 and 18.2 pm.
5. Subtract the predicted flux estimates from our observed estimates, which
leaves only the contribution of the dust, defined as the excess emission.

Two questions immediately come to mind concerning this procedure: (1) is it legitimate

to use the ratios (FIo.8 im/F2.2 lpm) and (F18.2 pm/F2.2 pm) in the calculation and (2) is the

assumption the photosphere behaves as a blackbody accurate enough? In short, the

answer to both is yes. As illustrated in Figure 3-1 in the mid-IR we are working on the

Raleigh-Jeans tail of the blackbody distribution for even the coolest stars in our sample

(-5000 K), and the stellar photospheres follow the well-known v-2 falloff in this regime.

Close inspection of the 5000 K curve in Figure 3-1 shows there is s slight deviation from

Raleigh-Jeans falloff at 2.2 p.m, but this is a relatively small discrepancy, and only

applies to one of our survey sources. We address the second question by plotting three

quantities in Figure 3-2, the near and mid-IR portion a blackbody function with T =

10000 K, a model atmosphere for Vega from Cohen (1995), and flux density estimates

for Vega from OSCIR observations. All three curves have been normalized to unity at 2.2


.0 '1900MI K
:5004 K


001 -

1 10-5

1 10 -

1 10 -

S10 01 1 10 100 110
Waveengh (mrons)

Figure 3-1: Blackbody plots for temperatures that bracket the effective temperatures
(Teff) of the stars in our Vega-like sample. Our spectral types of our sample stars range
from a B3V (19000 K) to K2V (5010 K). The mean Teff of the sample is 8500 K. The
vertical dotted lines denote 2.2, 10.8, and 18.2 gm (from left to right).

gm. For the OSCIR observations we have also plotted error bars that mark the 10%

uncertainty in their values. Both the model atmosphere and the blackbody line lie well

within the photometric uncertainties associated with our OSCIR observations. The close

correlation of all three plots in Figure 3-2 illustrates that for the relatively hot

temperatures of the intermediate mass stars in our sample the approximation of the stellar

photospheric emission as a blackbody function is robust.

We used this method to remove the stellar contribution from our observed flux

estimates for all of the stars in our sample. For star/disk systems where the mid-IR flux

comes primarily from the dust emission the procedure is sound. However, for a subset of


...---- 'olen Dat* a
XXX BB appox.


WUy&,,t1, (ncrron)

Figure 3-2: Comparison of OSCIR data, 10000 K blackbody function, and model
atmosphere data for Vega. Error bars on OSCIR data are 10% and represent the
uncertainly in our measured flux values.

seventeen of the survey stars this procedure returns negative excesses at either 10.8 or

18.2 pm clearly not a legitimate result. We believe that these non-physical results are

primarily due to the 10% uncertainties in our OSCIR flux estimates. Since the

determination of the magnitude of the excess is directly related to our observed mid-IR

flux estimates, any uncertainty in those estimates translates directly into uncertainties in

our estimates of the excesses. For these seventeen sources the amount of their mid-IR

excess is within our observational uncertainties, and we therefore cannot calculate the

amount of excess they exhibit. Because of this, we are forced to remove these seventeen

stars from our analysis of the disk grain properties.

Clearly this has a dramatic effect on the results of our survey. Because of this

issue we have only modeled the grains associated with 22 of the observed systems, or

56% of our sample. However, we use the fact that some stars exhibit little or no excess in

the mid-IR in the discussion of our proposed evolutionary sequence presented later in this

chapter. Indeed, these sources with little or no mid-IR excess form one of the three sub-

classes we divide our survey into and represent the last stage of our proposed sequence.

We do have robust estimates for the 10.8 and 18.2 gtm excesses of the remaining

22 survey sources and we present them in Table 3-2. Listed by HD number, we present

our excess estimates in units of flux density, and as a percentage of our observed flux

estimate. A literature search and use of the SIMBAD database gave the spectral types for

each of the sources in the table. We also list the effective (Teff) we used for each star in

our calculations of the excess. The majority of these values were taken from Allen

(2000), while others were taken from the literature, in particular Dunkin, Barlow, and

Ryan (1997b).

One important conclusion that we can infer from Table 3-2 is that although our

sample is biased toward stars of A-type, within the sample there is no apparent

correlation between spectral type and amount of mid-IR excess. This is supported by the

literature where it has been found that at least 15% of nearby field stars of all spectral

types between A to K have circumstellar dust systems like those associated with our

sample (Lagrange, Backman, & Artymowicz 2000). We also see that percentage wise

there is varying amounts of 10.8 glm excess associated with our sample stars. The

excesses of twelve stars in Table 3-2 accounts for more than 95% of the observed 10.8im

flux, seven of the stars have intermediate values of 10.8 tim excess (between 10 to

Table 3-2: Excess Characteristics of the Survey Sources
10.8 gLm % Excess 18.2 tm % Excess
Source Name Sp. Type Tff Excess 8 Excess
(10.8 pLm) (18.2 I.m)
(Jy) (Jy)
HD 9672 A1V 9150 0.003 1.3 0.05 38.4
HD 34282 AOV 9480 0.49 98.6 0.31 99.2
HD 34700 GO 5930 0.37 88.8 1.34 98.8
HD 35187 A2V 8810 3.08 98.2 5.78 99.7
HD 38678 A2V 8810 0.13 7.2 0.05 7.7
HD 39060 A5V 8160 1.22 42.7 4.05 88.0
HD 98800 K2V 5010 1.12 79.8 4.82 98.0
HD 101584 FOI 7700 75.94 98.4 108.58 99.6
HD 104237 AO 9480 13.43 95.3 7.98 97.3
HD 109573 AOV 9480 0.03 1.6 0.84 93.1
HD 135344 AOV 9480 1.25 96.6 2.58 99.4
HD 139614 A7V 7930 3.50 99.0 7.76 99.8
HD 141569 B9Ve 10000 0.25 77.2 0.62 96.1
HD 142666 A8V 7630 8.26 99.3 8.75 99.8
HD 143006 G6 5620 0.82 97.4 1.84 99.6
HD 144432 A9V 7300 7.68 99.2 6.56 99.7
HD 155826 F7V 6240 2.10 75.7 3.67 94.1
HD 158643 AOV 9480 15.29 97.1 11.19 98.6
HD 163296 A1V 9150 20.19 97.5 16.60 99.0
HD 169142 A5Ve 8400 2.28 98.1 7.19 99.8
HD 188037 A2* 8810 167.94 72.2 95.47 80.8
HD 233517t K2III 4420 0.32 72.6 1.90 97.7
HD 233517t K2V 4900 0.32 72.6 1.90 97.7
SIn sp. Type column denotes multiple star system
t Ambiguous classification, see Chapter 6

95%), and three stars have little 10.8 gpm excess (< 10%). Coupled with the seventeen

sources that have no measurable excess and are not listed in Table 3-2 there are twenty

stars that exhibit <10% excess at 10.8 pim. This is our first look at the three sub-classes

we break the survey into with regards to our proposed evolutionary sequence. As we will

see, the amount of the 10.8 lnm excess exhibited by a star/disk pair is a good indicator of

the evolutionary state of the system.

Now that we have our best estimate at the values for the emission from the dust

around the stars in the survey we can proceed with our investigation of the properties of

that dust. In the following sections we discuss our modeling of these mid-IR-emitting

grains and present the results of that modeling.

Grain Models

In this section we discuss the properties of the radiative equilibrium models we

used to analyze our mid-IR data. The core of the model code used was developed by

members of the University of Florida Dust Dynamics Group. In particular, the version of

the code used in our analysis was written by Dr. Mark Wyatt. We begin this section with

a high-level overview of the operation of the modeling code, then present results for one

of the survey sources, HD 141569. Using the HD 141569 results as an example, we

describe the model output and show how it relates directly to the physical parameters of

the grains we are studying.

Before we discuss the operation of the modeling code, we need to address what

assumptions were made about the grains. The most fundamental assumption made is in

regards to the composition of the dust. For the models presented in this dissertation we

assume that the disk particles are comprised of the astronomical silicates of Draine and

Lee (1984) and Laor and Draine (1993). There is evidence for particles with this

composition in both interplanetary dust particles gathered high in the Earth's atmosphere

(Leinert and Griin 1990), and in disks similar to those studied here. Sylvester et al. (1996)

presented mid-IR spectra of 13 sources in our survey and definitively detected silicates in

seven of them. Other significant detections of silicate materials in Vega-like disks are

those in the archetypal Vega-like source P Pictoris (Telesco and Knacke 1991), and in the

source 51 Oph (Fajardo-Acosta, Telesco, & Knacke 1993).

We also make the assumptions that the model particles have a density p = 2.5 g

cm", and that they are spherical. We make the assumption about the density since

'astronomical silicates' is not a physical material and its measured density is undefined.

The value of 2.5 g cm-3 is used since it is representative of interplanetary dust particles

(Gustafson 1994). With regards to the assumption about the shape of the grains, one must

first realize that the methods for calculating the optical constants for grains are limited.

Rigorous solutions for light scattering from particles are only available for spherical

grains, using Mie theory, and infinite cylinders, as described in Bohren & Huffman

(1983). New solutions for more complicated aggregate particles have recently become

available (Xu & Gustafson 1999), but these methods are computationally intensive, and

beyond the scope of the modeling presented here. Recent advances in the work on the

composition and structure of particles in circumstellar disks has shown that these

assumptions likely oversimplify the grains (e.g. Gustafson 1994), however, this sort of

analysis provides a good starting point for more detailed models which incorporate more

complex grain structures. In a following section we discuss the limitations of our models,

and touch on how different compositions and/or grain morphology changes our results.

Using the optical constants for astronomical silicates from Laor and Draine

(1993) our radiative equilibrium models use Mie theory to calculate the absorption

coefficients (Qabs) of different sized particles at different distances from the central star.

Once we know how efficiently a particle of a given size (diameter, D) emits and absorbs

radiation we can work out what its temperature will be at any distance from the star (r) by

iteratively solving the equation from Gustafson (1994):

T(D,r) = [T* / r(D,r)]1/4 X Tbb (1)

Where T- is the absorption efficiency of the grain averaged over the stellar

spectrum (which has been approximated by a blackbody with temperature T.) and

T(D,r) is the same quantity averaged over a blackbody spectrum of temperature T.

Tbb is the equilibrium temperature of the grain in degrees Kelvin if it were a blackbody

and is given by:

Tbb = 278 x r-/2 x (L*/L)1/4 (2)

where r is the distance of the grain from the star in AU and L. is the stars luminosity in

units of L = 3.9 x 1033 erg sec'.

We can also calculate the thermal emission from the dust received at the Earth at

a given wavelength. In the optically thin case, which we assume for our survey sources,

this can be written as:

Fv(X, D, r) = Qabs(,, D) x Q(D) x Bv[X, T(D, r)] (3)

where Qabs(A, D) is the grains efficiency as a function of wavelength and diameter, Q(D)

is the solid angle subtended by the grain at the Earth, and Bv[A, T(D, r)] is the Planck

function for the temperature of the grain. Using the luminosity and effective temperature

of the stars along with the assumed optical properties of the grains, our models use

equations (1) and (2) to calculate temperatures for grains of different sizes at varying

distances from the stars.


I II 11l

200 10 AU

1 00 -
50 \ -- ...... 30 n -

5 0- ...... -. ..... -
0.01 0.1 1 10 100 1000
Particle diameter, /m

Figure 3-3: Model temperature calculations for spherical Mie particles 10 to 50 AU
from HD 141569. The shape to the temperature curves is due to the changes in the values
of the grain emission and absorption efficiencies. The shading marks the three regions
described in the text.

To show an example of our model temperature calculations in Figure 3-3 we plot

the calculated temperature against grain diameter for particles between 10 to 50 AU from

the star HD 141569, a source that we resolved at Keck II (see Chapter 5). We see that

grains at these distances reach temperatures of 150 to 250 K. This is the temperature

range we derive for the majority of the survey sources that have well defined excess

values, and is one of the conclusions of our work. That is, the mid-IR excess emission

from Vega-like sources traces material that is at temperatures between 150 to 300 K.

Once we have calculated the temperature of the grains as a function of distance

from the star and diameter T(D,r) we can use equation (3) in conjunction with the grain

3 1+

40 AU, \ --"5o

*o^:______V r7' 30:4_
Part'ci d omneter ur Poat'cle d3 o etr ira

Figure 3-4: Plots of emitted flux density per unit solid angle at 10.8 g.m (left) and
18.2 plm (right) for grains 10 to 50 AU HD 141569. Regions shaded as described in text.

efficiencies to calculate the amount of emission from a grain as a function of wavelength,

Fv(X, D, r). In Figure 3-4 we plot the emitted flux density per unit solid angle (i.e.

specificintensity) at 10.8 and 18.2 gm for grains near HD 141569. This is the quantity

Qabs(A, D) x By[X, T(D, r)]

from the right hand side of equation (3) and has units of Jy/ster. This is the first point in

the modeling process where we have output that is in terms of physical units which is

important since we need to compare our model output to our observations.

To make the direct comparison between our models and observations we form the

ratio of the two plots shown in Figure 3-4 by dividing the 10.8 gm curves by the

corresponding 18.2 gm ones. The resulting flux-ratio plot is shown in Figure 3-5. The

plotted curves represent the flux ratio predicted for a particle of a given size at a given

distance from the star. The horizontal line plotted in Figure 3-5 marks the observed flux

ratio of the excess emission associated with HD 141569. It is where this horizontal line

intersects the model curves that returns a particle size at a given distance from the star


I 1.00 ---------. \

10 A

S0.10 20 AU
S\ \ \\". .
30 AU-

40 AU

0.01 50 AU

0.01 0.1 1 10 100 1000
Particle diameter, p/m

Figure 3-5: Flux density ratio plot for HD 141569. The curved lines represent the flux
density ratio emitted by model grains between 10 to 50 AU from HD 141569. The
horizontal line marks our observed (10.8 pm/18.2 p.m) flux ratio. The intersection of the
model curves and horizontal line returns grain sizes at the given distance from the star
that are consistent with our observations.

that is consistent with our observed flux ratio. Note that there is a double-valued solution

for grains at 30 AU from the star. For most of the survey source model results we see this

behavior where two different values of the diameter for a particle at a given distance give

the same flux ratio. This results from the fact that the temperature of the particle depends

on the particle size, which in turn depends on the ratio of the absorption and emission

efficiencies. Since we cannot distinguish between the two values, we use the largest

consistent size to place an upper limit on the diameter of the mid-IR-emitting grains in

the disks.

It is the interaction of the absorption and emission efficiencies of the grain that

causes the shape of the curves in Figures 3-3, 3-4, and 3-5. This 'S' shape is

characteristic of our models, and is seen to some extent in all of the results for the survey

sources. This is because it is ultimately the absorption and emission efficiencies that

determine the equilibrium temperature of the grain with respect to the radiation field. The

temperature of the grains then determines the amount of emission at a given wavelength

(Figure 3-4), and the ratio of the emission at two wavelengths is how we connect the

models to the observations (Figure 3-5). Therefore, the effects of the varying efficiencies

are seen directly in all of the various model outputs. This is not surprising since these are

radiative equilibrium models, which means that the interaction of the grains with

radiation is the most fundamental physical process with which we are dealing.

To discuss the reasons behind this characteristic 'S' shape of the curves we will

use Figure 3-3 as an example. We divide the figure into the regions represented by the

three shades of gray where each of the regions defines a regime where certain interactions

between the grain efficiencies are determining the shape of the model output curves. At

the leftmost side of the plot the grains have a diameter of 0.01 p.m, similar to the

wavelength of UV radiation. At this point the small grains do not effectively absorb the

stellar radiation and have relatively low temperatures. As you move from the leftmost

side of the plot through region I to larger grain sizes, both the absorption and emission

efficiencies increase; however, the absorption efficiency increases more quickly than the

emission efficiency and we see a rise in grain temperature. At the boundary of regions I

& II the absorption efficiency reaches a maximum resulting in the maximum grain

temperature for a given distance from the star. As we move through region II toward

region III the absorption efficiency of the grains stays relatively constant but the emission

efficiency increases which results in cooler temperatures for the grains. Near the

boundary of regions II & III the grain temperatures actually dip below the blackbody

temperature for grains at a given distance from the star. This seemingly incongruous

result stems from the fact that for grains with D=20 iim their emission efficiency can go

up as high as 2 due to emission in silicate resonances and they become super efficient

emitters (Wyatt et al. 1999). In region III we move out of the regime where they grains

are super emitters and their temperature climbs back up toward the blackbody

temperature for grains at that distance from the star. The same overall shape can be seen

in the curves in Figures 3-4 and 3-5. We have shaded these figures similarly to illustrate

the three regions. The effects of the changing efficiencies are especially evident in Figure

3-4 since there we are looking at the quantity Qabs(k, D) x Bv[X, T(D, r)] and the

variations in Qabs directly change the shape of the plots.

Color Temperatures and Optical Depths

We can also use equation (3) to calculate a value for the mid-IR optical depth of

the material around the survey sources. If we use our size limits in Table 3-1 to calculate

a value for the solid angle (Q(D)) subtended by the entire mid-IR-emitting region, and

note that Qabsc t v through the relation:

T= Xt(D/2)2 Qabs X nd L (4)

where rt(D/2)2 is the cross sectional area of a grain with diameter D, nd is the number

density of the grains along the line of sight, and L is the line-of-sight path length we can

then solve for a consistent T, t pair. Using our observations at the two different

wavelengths (10.8 and 18.2 im) and equation (3) we then have two equations and two

unknowns for the temperature and T for the mid-IR-emitting dust:

Fi0.81m= x Ltio x By[10.8 um, T] (5)


F18.2 m= x 8.2 x Bv[18.2 im, T] (6)

We can then solve these equations iteratively for the color temperature of the dust, T, and

the corresponding depth of the dust, T. To absolutely scale the optical depth we use X =

9.7 gm as our reference wavelength. The resulting color temperatures and optical depths

are listed in Table 3-3. There is good agreement between this method of temperature

calculation and the grain temperatures returned by the modeling code. If we calculate the

average grain temperatures returned by the models for each source, and compare that

temperature to the calculated color temperature for the same source the mean difference

between the two methods is < 30 K for the survey as a whole.

We next investigate the dynamical properties of the thermally emitting grains

using our models. A particles radiation pressure efficiency (Qpr) can be calculated from

its optical constants using Mie theory. Qpr is related to its absorption and scattering

efficiencies through the relation

Qpr = Qabs + Qsca (1 )

where the term accounts for the fact that radiation is scattered asymmetrically

from the particles (Wyatt et al 1999). Once we have Qpr calculated we can consider the

ratio of the radiation forces acting on the particle to the gravitational forces as a function

of particle diameter. This ratio, called B is given by the relation below



01 j

0.01 0.1 1 10 100 1000
Particle diameter, imr

Figure 3-6: The quantity 3 as a function of grain diameter for particles near HD 141569.
Horizontal line marks a value of 1. Particles with p > 1 will be removed from the system
on hyperbolic orbits in < 104 yr.

(D) = (C/pD) T* (L./L,) (M*/Me) (7)

where C = 1.15.10-3 kg m-2, p is the particle density, D is the particle diameter, and

T. = J Qpr(D,X) Fx dX / J FX dX

is the particle's radiation pressure efficiency averaged over the stellar spectrum with

assumed temperature T. (Wyatt et al. 1999). By calculating p as a function of particle

diameter, we can determine what sized particles will be removed from the system through

radiation expulsion. Particles with 0 > 1 will be removed from the system by radiation

expulsion on very short timescales. On average, the time for removal for the survey stars

is <104 yr. Figure 3-6 shows p(D) for HD 141569. The horizontal line marks p = 1. Using

Figure 3-6 we see that for HD 141569 all particles smaller than D = 5 L.m have P > 1 and

will be removed from the system quickly. We can therefore make the claim that grains

with D < 5 Clm are likely not primordial, and have been replenished by some mechanism.

The determination of whether or not the mid-IR-emitting grains in the survey disks is an

important conclusion of the modeling since the presence of "second-generation" dust in

these disks places them firmly in the Vega-like class (Lagrange et al. 2000). Plots similar

to Figure 3-6 for all the modeled survey sources allow us to set a limit on the minimum

size of grains that would not be removed from the system through radiation expulsion.

These "p-Limiting" sizes are given in the last column of Table 3-3. To reiterate, all grains

smaller than the 'p Limiting' size (i.e. the blow-out size) will be removed from the

system by radiation expulsion.

Another dynamical property we investigate is the lifetime of the disk particles

with respect to Poynting-Robertson drag (P-R drag). Once we have calculated p(D) we

use the following relation to calculate the P-R drag lifetimes of the grains as a function of


tpR(D) = 400(M./M.)(r2/p(D))

where r is the distance of the grain from the star in AU. Using the upper limits for grain

diameter and distance from the star, we calculate a maximum P-R drag lifetime for the

mid-IR-emitting grains in each disk. This maximum value represents the longest time

any mid- IR-emitting grains would survive in the disk before spiraling into the star. If the

grains are smaller than our upper limit, or closer to the star, their P-R drag lifetimes

would be less than this value. We list the calculated maximum values for the survey stars

in Table 3-3. The longest P-R drag lifetime in the survey is 3.106 yr, for grains near HD

9672. Although it is not certain, we suspect that all of the survey sources are older than

this value. It therefore seems likely that the mid-IR-emitting grains in these disks

Table 3-3: Derived Grain Parameters
MidR Grain Grain
e Cr size Distance Max. [-Limiting
Source Color
Name Tmid-IR upper Upper P-R Drag size
(K) limit Limit Lifetime (yr) (PLm)
S(m) (AU)

HD 9672 107 6.30E-03 7.00 111.0 3.0E+06 6.0
HD 34282 375 3.80E-05 10.00 10.0 3.0E+04 9.0
HD 34700 165 5.60E-03 8.00 9.0 9.0E+05 0.9
HD 35187 211 9.90E-02 5.00 31.0 3.0E+05 6.0
HD 38678 588 2.80E-06 4.00 5.0 5.0E+03 5.0
HD 39060 171 6.80E-02 4.00 31.0 5.0E+05 3.5
HD 98800* 157 1.84E-01 2.50 5.0 2.0E+05 ---
HD 101584 237 5.50E-02 7.00 630.0 1.0E+05 >100
HD 104237 390 1.40E-03 8.00 7.0 2.0E+04 9.0
HD 109573 100 5.80E-03 10.00 78.0 1.0E+06 9.0
HD 135344 203 7.50E-03 15.00 43.0 5.0E+05 9.0
HD 139614 197 4.10E-02 2.50 15.0 2.0E+05 3.5
HD 141569 187 2.33E-03 4.00 30.0 2.4E+04 5.0
HD 142666 274 3.40E-02 1.50 6.0 2.0E+04 3.0
HD 143006 196 6.10E-03 10.00 4.0 5.0E+05 4.5
HD 144432 307 1.00E-02 7.00 4.0 1.0E+04 2.5
HD 155826 213 8.80E-03 3.00 7.0 1.0E+05 1.5
HD 158643 338 1.70E-02 2.00 13.0 5.0E+04 9.0
HD 163296 315 1.10E-02 3.00 13.0 5.0E+04 6.0
HD 169142 174 9.60E-02 4.00 30.0 4.0E+05 3.0
HD 188037 410 1.27E-01 6.00 5.0 9.0E+03 5.0
HD 233517 142 6.50E-02 10.00 81.0 1.0E+06 30.0
values: 232 K 0.042 6 tm 25 AU 4.4E+05 yr 6.0 pm
3 < 1 for all grain sizes

are not primordial material and have been replenished in some way. One method of

replenishment is through the collisions of larger bodies in the disks. Replenishment of

material in this manner is apparently occurring in the archetypal Vega-like disks of both

p Pictoris (Artymowicz 1997), and HR 4796A (Wyatt et al. 1999; Telesco et al. 2000).

The last row of Table 3-3 contains the mean values of each of the quantities listed

in the table. We calculate these numbers simply to give an idea of the properties of these

disks in the broadest sense. Certainly these values are only useful as benchmarks for the

survey sources, however, it is interesting to compare them with values from other Vega-

like disks. In particular the mean upper limit to the mid-IR-emitting grains size is very

consistent with the grain sizes in other Vega-like disks. Similar modeling of the HR

4796A disk by Wyatt et al. (1999) and Telesco et al. (2000) shows that the grains in that

disk are approximately 2 gtm in diameter. Mid-IR aperture photometry of p Pictoris

implies there are 1 to 3 ptm grains in that archetypal disk (Telesco et al. 1988; Telesco &

Knacke 1991; Aitken et al. 1993). Vega itself is also reported to have 1 to 10 gim grains

as the dominant population in its disk (Van der Bliek, Prusti, & Walters 1994).

The fact that our estimated grain sizes for the 22 survey sources we studied have mid-IR-

emitting grains that are similar in size to the archetypes is one of the main conclusions of

our work. In fact, the mid-IR emission detected around this class of source is tracing this

size range of grains in their disks. This is more or less expected since this is thermal

emission from the dust we are detecting and the dust grains emit most effectively when

their size is comparable to the wavelength of the radiation they are emitting. In the next

section we continue with our comparison of the survey sources to some of the archetypes.

Scale Size Comparison

An obvious question to ask is, would the unresolved sources be resolved if they

were at the distances of p Pictoris or HR 4796A? The average distance of the modeled

Table 3-4: Comparison of Unresolved sources to Archetypal Disks

Predicted Size of
Size at Size at Size of
Source Name angular DHR4796A D Pic at HR
size 4796A at
(arcec) (arcsec) (arcsec) Dsource Dso
(arcsec) Dsource

HD 9672 3.64 3.31 11.68 3.11 3.30
HD 34282 0.12 0.30 1.05 1.16 1.23
HD 34700 0.02 0.27 0.95 0.16 0.17
HD 35187 0.41 0.93 3.26 1.27 1.34
HD 38678 0.45 0.15 0.53 8.64 9.14
HD 39060 3.26 0.93 3.26 10.00 10.58
HD 98800 0.21 0.15 0.53 4.04 4.28
HD 101584' 1.55 18.81 66.32 0.23 0.25
HD 104237 0.12 0.21 0.74 1.64 1.73
HD 109573 2.33 2.33 8.21 2.84 3.00
HD 135344 2.15 1.28 4.53 4.75 5.03
HD 139614 0.19 0.45 1.58 1.21 1.28
HD 141569 0.61 0.90 3.16 1.92 2.03
HD 142666 0.11 0.18 0.63 1.67 1.76
HD 143006 0.09 0.12 0.42 2.02 2.14
HD 144432 0.03 0.12 0.42 0.75 0.79
HD 155826 0.45 0.21 0.74 6.13 6.48
HD 158643 0.20 0.39 1.37 1.45 1.53
HD 163296 0.21 0.39 1.37 1.56 1.65
HD 169142 0.41 0.90 3.16 1.31 1.39
HD 188037 0.04 0.15 0.53 0.83 0.88
HD 233517t 0.27 2.42 8.53 0.32 0.34
SLikely not a main-sequence star

survey sources is 95 pc, while P Pictoris is at 19 pc, and HR 4796A is at 67 pc. To

answer this question we 'moved' the unresolved sources to the distances to those

archetypes and calculated what their model predicted size would be. In this analysis we

assume that the maximum extent of the mid-IR-emitting grains from the stars is given by

the upper limits returned by our models, which are listed in Table 3-3. Using these values

in conjunction with the distance estimates for the stars given in Table 2-1 we can

calculate a predicted angular size for each of the disks. We then scaled those predicted

sizes to what they would be at the distances of 0 Pictoris and HR 4796A. These predicted

and scaled sizes are listed in Table 3-4. Assuming that we are using an 8 to 10 m class

telescope to make the observations, by analogy with HD 169142 (Chapter 4), we set the

threshold of detection at an angular size 0.7". While this limit is somewhat arbitrary, our

experience with the survey data has shown that this is about the limit at which a face-on

disk source can be distinguished from PSF measurements. This represents a worst-case

scenario though since edge-on disks may be detectable at smaller angular scales.

What we see is that for the most part the predicted disk sizes at the real distances

of the stars are in the tenths of arcsecond range, below the current limit of detectability.

This changes somewhat when we move the survey stars to the distance of the archetypes.

At the distance of 3 Pictoris, the closer of the two, we predict that we could have resolved

twelve, or 54% of the disks. This seems to imply that it is only an observational limitation

that prevents us from resolving them. This is clearly a very rough way to estimate the

detectability of these sources. However, it does give us a starting point for the planning of

more in-depth observational programs.

Conversely, Table 3-4 shows that if we move the two archetypes to the distances

of the survey sources, we would be able to resolve both the disks of both P Pictoris and

HR 4796 at effectively all the distances of the survey stars. This implies that the disks of

the unresolved sources have a smaller scale size then the archetypes. While the cause of

this scale size discrepancy is still not fully understood one idea that may explain it deals

with the size of the mid-IR-emitting grains in the disks. If the grains in the P Pictoris and

HR 4796 disk are smaller then those in the disks of the unresolved sources, they will be

detected farther from the stars since smaller grains in equilibrium with a radiation field

are hotter than larger ones at a given distance from the star. If the grains in the unresolved

disks are larger then those in the archetypes, even by a factor of a few in diameter, they

will have to orbit their parent stars within a few AU for a late A-type star to emit strongly

in the mid-IR. This idea is supported by the results presented in Table 3-3 where we see

the model results imply that the grains around most of the unresolved sources orbit within

-20 AU of their stars and have diameters of -6 p.m. Compared to the -2 to 3 p.m

diameter grain in the 3 Pictoris and HR 4796 disks, the larger grains size may account for

the unresolved nature of some of the survey sources. The idea that the mid-IR-emitting

grains in the disks of the unresolved sources are larger also has implications for the

evolutionary status of those disks, and therefore the stars. Since it is believed that the

grains in these disks grow through aggregation, disks with larger grains in them are likely

older than those with small grains. In a later section of this chapter we investigate the

evolutionary status of these sources in detail, and propose an evolutionary sequence for

our survey sources.

Here we touch on two limitations of our modeling that are related to the basic

assumptions we made about the composition and morphology of the grains. Obviously,

the dust grains may not be simple Mie spheres. For example, the zodiacal particles in our

solar system seem to have a much more complex "bird's nest" structure. The "bird's

nest" like appearance of these grains occurs because the grain itself is actually an

aggregate of tens or hundreds of smaller particles that have stuck together through

collisions (Gustafson 1994). As noted, we have assumed that the observed grains are

solid and spherical. However, if the grains are actually aggregates and complex in

appearance ("fluffy"), our overall conclusions about particle lifetimes are not

significantly changed. Gustafson (1994) points out that Mie calculations for

homogeneous spheres consistently underestimate the value of 3 for aggregated dust

models by a factor of two. For our survey sources, this means that even larger particles

would have P > 1, and would be expelled from the star/disk systems in a short time. Since

the P-R drag lifetime of a particle is proportional to P-' (Wyatt et al. 1999), the larger

value of p for a grain of a given size would drive its P-R drag lifetime down, implying

that the grain would spiral in to the star in a shorter time. We therefore see that if the

grains in the disks of the survey sources are porous or fluffy, it does not change our

overall conclusion that most of them are second-generation grains, perhaps collision


The modeling presented here has shown that the mid-IR-emitting grains in the

disks of the survey sources are on the order of a few microns in diameter, similar to the

grains in the archetypal disks of 3 Pictoris, Vega, and HR 4796. It has shown that these

grains orbit at distances of tens to a hundred AU from the central stars, and have

temperatures in the range of 150 to 250 K. Considerations of the dynamical properties of

the grains like their P-R drag lifetime and the parameter 3 imply that these grains are

likely not primordial, and have been replenished by some mechanism. We also present

rough calculations which show that about half of the disks could possibly resolved at the

distance of 3 Pictoris, but that the dust lies too close to the other half of the sources to be

seen even at this relatively close distance of 19 pc. The main conclusion that we draw

from our models is that the mid-IR properties of the survey sources are, in fact, similar to

those of the archetypes. The fact that the mid-IR-emitting grains in the survey disks are

probably replenished places them firmly into the debris disk class, and the similarity in

their grain sizes, temperatures, and distances from the central stars to those of the

archetypes implies that they are at a similar evolutionary state. In the following sections

we expand on this idea of source evolution and present an evolutionary sequence for the

survey stars.

A Proposed Evolutionary Sequence

In this section we present an evolutionary sequence we have constructed for our

survey sources using the properties of their mid-IR excesses in conjunction with the

characteristics of the stars. The idea that the Herbig Ae/Be stars are the progenitors of the

Vega-like stars is well represented in the literature (cf. Lagrange et al. 2000; van den

Ancker et al. 1998). Trends seen in the characteristics of properties of disks also support

such an evolution. For example, Holland et al. (1998) sees a strong trend of decreasing

disk mass with age. Their Figure 2 clearly shows that the mass of the dust associated with

Herbig Ae/Be stars is orders of magnitudes greater than that in the tenuous disks in our

sample (Holland et al. 1998). More relevant to the discussion of our evolutionary

sequence is the trend seen in the evolution of the spectral energy distributions (SED) of

sources with associated circumstellar dust. Below we discuss how this idea is directly

related to our survey sources.

SED Evolution

For a graphical representation of what SED evolution looks like we reintroduce

the scheme of Andre (1994) in Figure 3-7. This figure shows how the SED of a star/disk


2 II I09
V cc

M Body

1 2 10 100

12 10 100
A un

SStI kr
1 2 10d t .

1 2 10 100
) (WW'

Figure 3-7: SED evolution of a star/disk system moving onto the ZAMS. Most of our
survey sources fall into Class III, but a small number of them exhibit some characteristics
of a Class II source. Using our survey sources we refine the Class III group here by sub-
dividing it into three smaller parts.

system evolves during the first few Myr of its existence. The most important trend to

notice in regards to our proposed additions to this sequence is how the overall amount of

excess emission decreases as the system evolves. That is, as the system evolves, less of

the overall infrared luminosity comes from the circumstellar dust, and the SED of the

system approaches that of the star itself, a simple blackbody curve. By the time the

source can be classified as Class III in this scheme the dust associated with the star has

become optically thin and only contributes a fraction of the overall system luminosity.

The low optical depths we derived for them in Table 3-3 support the idea that the dust

around our survey stars is optically thin. As we will see, a subset of our survey stars have

SEDs that have characteristics of an object transitioning from Class II to Class III.

However, the majority of our survey stars can be classified as true Class III objects.

To introduce our survey sources into the context of this sequence we present

Figure 3-8. In this figure we show three examples of SEDs of sources in our survey. In

the top panel we show the SED of HD 143006. Note the strong near-IR excess associated

with this source. We suggest that the presence of such an excess can be taken as a sign of

relative youth within the survey sources. This excess comes from dust that is close to the

star, within a couple AU in the case of HD 143006. The fact that there is dust so close to

the star implies that the processing of the inner disk has yet to commence. Indeed, the last

stages of active accretion may still be occurring in this system. Nine of our survey

sources exhibit excess emission in the near-IR, and these nine star/disk systems form the

first of three sub-groups we break our survey sources into. We name this sub-class "Class

IIIa" since the SED of these sources is similar to that of a source that has just evolved

into Class III in the Andre (1994) scheme. The middle panel of Figure 3-8 shows the

SED of HD 141569, a source that falls into our second sub-class (named Class IIIb).

Along with HD 141569, the other three members of this group represent an intermediate

stage of evolution within our survey. The four sources in Class IIIb exhibit no near-IR

excess, however they all have well defined excesses in the mid-IR. Processing of the

inner disk regions is most likely still ongoing in these sources. Interestingly, three out of

le-12 "




1 10 100 1000
Wavelength (rm)


le-14- N

1 10 100 1000
Wavelength (pm)





1 10 100 1000
Wavelength (ptm)

Figure 3-8: SED evolution within our survey. (top) SED of HD 143006. Note the strong
near-IR excess associated with this source. (middle) SED of HD 141569. This source
represents an intermediate evolutionary step. There is no near-IR excess associate with
this star. (bottom) SED of HD 142764, a source that represents the last stage of evolution
within our survey. No near-IR or mid-IR excess is associated with this star.

the four sources in Class IIIb have been resolved in the mid-IR. The bottom panel of

Figure 3-8 shows the SED of HD 142764, a source representative of the last of the three

sub-groups, Class IIIc. HD 142764 and the other sources in this group exhibit no excess

emission in the near-IR, and little or none in the mid-IR either, indicative of them being

the most evolved. By the time a star/disk system evolved to this point, processing of the

inner portions of the disk has taken place, and a central hole has formed leading to the

paucity of excess IR emission. In the next section we discuss the three sub-classes in

more detail, and show how each group is tied together by common characteristics.

Class liIa: The near-IR Nine

The first sub-class we discuss is the group the represents the earliest stage in the

evolutionary sequence. Class Ilia contains nine of our survey sources that have a

common trait in that they all exhibit significant near-IR excess. Because of this the SEDs

of these sources all resemble that of HD 143006 (Figure 3-8 top), which exhibits a strong

near-IR excess as seen in Figure 3-8. Listed in Sylvester et al. (1996) we see that the nine

sources have an average of 0.6 mag of excess at 1.7 glm, and 1.1 mag of excess at 2.2 p.m.

Sylvester et al. (1996) also calculated blackbody fits to the near-IR excesses of the

sources and found that grain temperatures in the range of 1500 to 2000 K fit the excess

well. In our own analysis of these sources we find that the Class Ilia stars exhibit 7 out of

the top 10 highest mid-IR optical depths (see Table 3-3). The fact that these stars have

near-IR excess, and relatively high optical depths makes them similar to Herbig Ae/Be

stars, and they are therefore most likely young. Indeed, at least 3 of the 9 Class IIIa stars

show hydrogen emission lines in their spectra (Sylvester et al. 1996), another sign of


the r i \ F) '- .()
\ %
-I O-2--------------------\-^^^------

I O E16S

Figure 3-9: SED of HD 169142 with ( )4/3 plotted for comparison. The diamonds are

fluxes. The solid line is a blackbody function with Teff = 8400 K. The dashed line follows
the relation F(X) a ()-4/3.

youth as compared to the other survey sources. Since it seems these sources are the

youngest in our survey, we use these sources as the starting point in our evolutionary


One of the main distinctions between Class II and Class III sources is whether

their disk is optically thick or thin. Class II sources have optically thick disks, while those

associated with Class III stars are optically thin. As noted by Hillenbrand et al. (1992)

one of the most powerful diagnostics of a circumstellar disk is its SED in the near-IR

(1 pm < X < 2.2 plm). Objects with flat, optically thick disks have SEDs that follow a

4/3 falloff wavelength. This fact gives us a criterion we can employ to place our survey

sources into the correct class. We know that any optically thick disks will follow that

functional relationship, and would belong in Class II. A steeper than X-4/3 falloff in

indicitave of emission from an optically thin disk, and warrant a Class III designation of

any source whose SED follows this trend. To answer the question of whether the sources

we designate as Class liIa are actually Class II sources we present Figure 3-9. In this

figure we plot the SED of one of the Class liIa sources, HD 169142. Its SED is

representative of all nine of the Class Ilia sources showing a strong near-IR excess

beginning at approximately X = 1.4 lim. HD 169142 is an A5 V star with a Teff = 8400 K

(Dunkin, Barlow, & Ryan 1997) and we plot a blackbody function with this temperature

as a model photosphere in the figure. Also plotted is a dashed line designating a X-4/3

falloff with wavelength. Inspection of the figure shows that in the near-IR regime, which

is bracketed by the two vertical lines, the SED of HD 169142 has a much steeper falloff

than :-413. This implies that HD 169142 and the other eight sources with similar SEDs are

likely to have optically thin disks, and should rightfully be classified as Class liIa objects.

Class IIIb: ZAMS Sources

The next step in our sequence brings us to Class IIIb. The four sources in this

class do not exhibit any near-IR excess, implying that they are somewhat older than the

Class IIIa stars. These stars do have a common trait though, they all have low

luminosities for their colors, which places them very near the ZAMS in the H-R diagram

(Jura et al. 1998). Interestingly, 3 out of the 4 sources (p Pictoris, HR 4796A, and HD

141569) in this class have been resolved in the mid-IR.

With respect to how these sources fit into the evolutionary sequence, we can ask

the question: how young are they? Luckily, for this sub-class we have an answer for that

question. One of the Class IIIb members is HR 4796A, which has a well-constrained age

of 8 3 Myr (Stauffer et al. 1995). Recent work by Barrado y Navascues et al. (1999)

uses Hipparcos data and evolutionary tracks to estimate the age of the archetype P

Pictoris at 20 10 Myr, similar to the age of HR 4796A. The third resolved source, HD

141569 has no published age values. However, we show in Chapter 5 that it is likely a

source that is transitioning from the Herbig Ae/Be class to the Vega-like class. Lagrange

et al. (2000) coin the term "old-pre-main-sequence" star when describing HD 141569,

indicating that it is more or less finished with its PMS evolution and is approaching the

ZAMS. The age estimates for HR 4796A and p Pictoris, along with their position on the

ZAMS seem to imply that these stars are relatively young, perhaps 10 to 30 Myr old.

Since it seems that they are finished with their PMS evolution, we suggest that they are

older then the Class liIa sources, and form the second step in our sequence.

Class IIIc: Post ZAMS Sources

The last sub-group in our sequence is Class IIIc. The remaining twenty-six stars in

our survey fall into this class. They represent sources that are the most evolved and have

most likely already moved onto and then off of the ZAMS and are now crossing the main

sequence in the middle stage of their lives. Class IIIc sources show no near-IR excess,

and little if any mid-IR excess. In fact, all of the sources that we could not model due to

uncertain excess determination fall into Class IIIc. None of the sources in this class

exhibit any signatures of youth, like emission lines in their spectra. We take the paucity

of near and mid-IR excess coupled with the lack of any signatures of youth to mean that

these are the most evolved sources in the sample. In reality, these sources could likely

Table 3-5: Evolutionary Class characteristics

Class Excess Age H-R Diagram Class Charactristics
Name Wavelength (Myr) Position

Ha emission in spectra
Class IIIa 1.2 to 1.7 < 10 Pre-ZAMS Li seen in source spectra
Strong near-IR excess

no near-IR excess
Class IIIb 10- 20 10 to 20 ZAMS strong mid-IR excess
Good age estimates

little or no mid-IR excess
Class IIIc > 20 > 20 Post-ZAMS
no signatures of youth

form a new class, "Class IV" in the schemes of Andre (1994) and Hillenbrand et al.

(1992). A class that would represent sources already on the main sequence, completely

finished with the formation and PMS evolution.

In Table 3-5 we summarize the characteristics of the new sub-groups. To quantify

the idea of SED evolution we list the wavelengths at which the excess 'turns-on' for each

class and show that it increases with age. We also list approximate ages for each of the

groups and their corresponding position in the H-R diagram. Finally, we comment on

some of the common characteristics for each of the sub-classes that we use to tie each

group together. In the next and final section of this chapter we use the H-R diagram to

graphically illustrate our proposed evolutionary sequence.

Survey Sources in the HR Diagram

It is useful to investigate the position of our survey sources in the H-R diagram.

Such an investigation allows us to graphically see how the survey fits into the overall

scheme of stellar evolution, and it gives us another way to present our evolutionary

sequence. For this purpose we use a H-R diagram comprised of =41000 stars taken from

the Hipparcos database (Perryman et al. 1995). We have Hipparcos measured distances

for 90% of the stars in our survey, and we use them in conjunction with their Tycho

measured B-V values to place them in the diagram as shown in Figure 3-10. For most of

the sources this is the first time their newly measured distances have been used to place

them in the diagram. In fact, this is the first time ever that Hipparcos and Tycho data has

been used in conjunction with a significant number of Vega-like stars to illustrate their

position in such a diagram. Inspection of Figure 3-10 shows that several interloping

giants made their way into the survey, one in particular, SAO 26804, Is discussed in

detail in Chapter 6. However, generally the sources are on, or near the main sequence.

While it is instructive to see the whole survey in the diagram, the real value of this

kind of presentation comes into view when we discuss our proposed evolutionary

sequence. We present the sequence as a series of three panels in Figure 3-11. In Figure 3-

1 la we show our so-called Class Ilia sources. These represent the youngest of our survey

sources, and we propose that they are still moving onto the ZAMS. In Figure 3-1 Ib we

show our Class IIIb stars. These are the sources that have relatively well defined ages at

-10 Myr. Notice that we can now see that they all have low luminosities for their colors,

as reported by Jura et al. (1998). Since they are very close to the ZAMS we use them as a

useful fiducial point in the sequence. These sources are somewhat of a fulcrum point,

S ,

*~ 4 +




0.5 1.0

B- V [mag]


1.5 2.0

Figure 3-10: The Vega-like survey sources in the H-R diagram. This figure shows the
sources of the survey plotted in a HR diagram comprised of -41000 stars from the
Hipparcos catalog. The four squares are the position of the disks that have been resolved
in the mid-IR, P Pictoris, HR 4796A, HD 141569, and 49 Ceti. The arrow is a reddening
vector. Hipparcos data from Perryman et al. (1995).






S *

5 5

10 01 10 -

(a) (b) (c)
0.5 00 05 0 1.5 00 05 1.0 1.5 00 05 1.0 1.5 2.0
B V mi agl- Hm B-V ],i,

Figure 3-11: Three steps in an evolutionary sequence on the H-R diagram. (a) Step 1:
Class IIIa sources, stars moving onto the ZAMS. (b) Step 2: Class IIIb sources, stars
which are on the ZAMS. (c) Step 3: Class IIIc sources, stars moving away from the

about which the other stars in the survey pivot. On one side of the fulcrum are the Class

IIIa sources, which are younger and still moving onto the ZAMS. On the other side, we

have the Class IIIc sources that are most likely older and have already reached and

moved off of the ZAMS.

To investigate the timescales involved with the first stage of this sequence, we

present Figures 3-12, and 3-13. In Figure 3-12 we have plotted the Class IIIa and IIIb

stars in the H-R diagram with the PMS evolutionary tracks of Palla and Stahler (1993)

overlaid. Using the tracks we determine age estimates for the Class IIIa sources in the

range of 3 to 10 Myr, consistent with them being younger than the Class IIIb stars.

Indeed, the Class IIIb stars lie at the end of the PMS tracks while the Class IIIa sources

look to be still following them onto the ZAMS. The isochrones of Iben (1965) agree with

the age estimates given by the PMS tracks. In Figure 3-13 we see that most of the Class

IIIa sources lie between the 3 Myr and 10 Myr isochrones.


**,4 4''


0.5 1.0

B- V [mag]

Figure 3-12: H-R Diagram with the PMS evolutionary tracks for intermediate mass stars
of Palla and Stahler (1993) overlaid. The stellar birth line is the dotted line. The mass
associated with each evolutionary track is given on the left in units of M0. Tick marks on
each track correspond to the times listed in the legend. Age estimates for these sources
are in the 3 to 10 Myr range.

Time (yr)






. m

, I *l I I n I .


0.5 1.0
B V [mag]


Figure 3-13: H-R Diagram with the isochrones of Iben (1965) overlaid. Also plotted are
the Class Ilia and Class IIIb stars. Class IIIa (circles) and Class IIIb (squares) stars are
plotted for comparison The times to reach each isochrone is given on the left side of each.

15 L


. .

~- .*

'" '.

As noted by Lagrange et al. (2000) a problem with using this method for age

estimates is that isochrones are normally bi-valued. That is, under normal circumstances

you cannot distinguish between a source moving onto the ZAMS and one evolving away

from it. Our sequence eliminates this issue since we use the shape and apparent evolution

of the SED of a source to determine whether or not is moving onto or away from the

ZAMS. Hence, we believe our method solves this problem for our survey.

That is in essence the extent of our proposed sequence. By sub dividing the

survey into three sub-classes we have come to one of the main conclusions of this work.

That is that we feel the evolution of the SEDs of the sources, coupled with the positions

of the sources in the H-R diagram is strong evidence that Vega-like sources in our sample

evolve in this manner. In the next chapters we move to talking about individual survey

sources. Chapter 4 discusses the face-on disk of HD 169142, Chapter 5 presents our work

on HD 141569, and in Chapter 6 we present results that 'de'-resolve the source SAO

26804. In Chapter 7 we discuss the OSCIR camera system and its important role in this



In this chapter we present Keck imaging of the source HD 169142. Our images

show the source to be resolved at both 10 and 18 Jm with strongly detected extended

emission seen in the form of a seemingly face-on disk with diameter approximately equal

to 1.2". At a distance of 145 pc (Sylvester et al. 1996) this corresponds to a radial extent

for the mid-IR emitting dust of -85 AU, which is very similar to the size of the mid-IR

emitting regions of the archetype P Pictoris and the source HD 141569 (Lagage & Pantin

1994; Fisher et al. 2000). Although we have a high degree of confidence in our detection

of extended emission associated with the source, we have no significant spatial

information on the distribution of the dust itself. This is primarily due to the fact that it is

extremely difficult to accurately remove the contribution of the star from the central

region of the emission. As we will discuss, the photospheric contribution to the total flux

of the source is minimal, however, the spatial deconvolution of that flux from the image

itself it not an easy task due to the relatively low signal-to-noise of the outer regions of

the extended emission. We can however use the global characteristics of the disk to infer

some of its properties. Using our images we place strong limits on the size of the

extended emission and show that the orientation of the disk is nearly face-on.

We also use our observations to model the mid-IR emitting dust as silicate Mie

spheres and determine a characteristic size and temperature for the grains. Since we have

no little spatial information on the grain distribution, we use our images to derive 'global'

dust parameters. We then use our derived grain parameters to discuss the possible

Table 4-1: Characteristics of HD 169142

Spectral Type: A5 Vet Teff: 8400 Kt

Distance: 145 pct log g: 4.2t
(B V): 0.29* M/Me: 2.0'
My: 2.3t L/Le: 14'
v sin(i): 55 2 km/sect R/Re: 1.7'

t Dunkin, Barlow, & Ryan (1997), T Sylvester et al. (1996), Allen (2000)

origin, and eventual fate, of the grains in the disk of this star. We also place HD 169142

into an emerging evolutionary sequence for this class of source and discuss how it may

be a bridge between the well-known class of Herbig Ae/Be stars and the Vega-like


Accurate stellar parameters are needed to confidently model the grains near HD

169142 and to interpret our new mid-IR images of the source. A literature search gives

the relevant characteristics of HD 169142 listed in Table 4-1 above. Note that HD

169142 was originally classified as a B9 V source in the list of Walker and Wolstencroft

(1988) but has since been re-classified as A5 V using high resolution optical

spectroscopy (Dunkin, Barlow, & Ryan 1997). Recent photometry has also refined the

rotation speed of the source and polarimetry of the source and has revealed that the

source exhibits variable polarization in the near-IR (Dunkin, Barlow and Ryan 1997).

These facts are important to our discussion since they give us clues to the nature of the

circumstellar disk. In the rest of the chapter we present our new observations, discuss the

results of our modeling of the grains in the disk, and comment on the evolutionary status

of HD 169142.

As we hinted at, the evolutionary status of HD 169142 is still very much in

question. It has been classified as a Vega-like star by many authors including both

Walker and Wolstencroft (1988) and Sylvester et al. (1996). However, HD 169142 is also

a member of various Herbig star lists like those of Malfait et al. (1998) and Corporon &

Lagrange (1999). The reason that HD 169142 is on both kinds of lists is because it is one

of several stars in our survey sample that exhibit characteristics of both a Herbig star and

a Vega-like source. HD 169142 meets two of criteria to be defined as a Herbig star by

having a spectral class earlier than FO and exhibiting single peaked Balmer emission line

profiles. HD 169142 also has Hel [X5876] in emission (Dunkin, Barlow, and Ryan 1997),

a fact that is relevant to its evolutionary status. The star does exhibit some Vega-like

characteristics though. In particular, both the existence of the large mid and far-IR excess,

and its characteristic shape are very similar to those of other Vega-like sources.

We believe that the possible transitional nature of HD 169142 is evident in its

spectral energy distribution (SED), which is shown in Figure 4-1. The optical-to-mm

SED shows that the source exhibits a large excess for all X > 1.7 plm. First discovered by

IRAS, the excess peaks between 30 to 40 gtm, which is very similar to the SEDs of all of

the Vega-like archetypes. Observations by Sylvester et al. (1996) revealed a strong near-

IR excess associated with the source in the H (1.7 gpm) and K bands (2.2 gtm) where they

measure an excess of 0.4 and 0.8 mag in the H and K bands respectively. The excess

becomes more pronounced at longer wavelengths and reaches 1.7 mag in the M-band

(Acen = 4.8 gim) (Sylvester et al. 1996). As we discuss in the observation section, the

excess is very pronounced in the mid-IR. In fact, the photosphere of the central star




is 7
10 \ -

1 10 100 1000
Wavelength (pm)

Figure 4-1: SED of HD 169142. The solid line is a model photosphere the points are
observed fluxes. The IR excess of HD 169142 is seen at all wavelengths > 1.7 pnm. The
near-IR excess can be interpreted as a sign of youth since it implies there is hot material
very close to the star and that the inner disk region has yet to be processed or cleared.
SED from Sylvester et al. (1996)

only accounts for a couple percent of the total flux at 10.8 gmm, and an order of magnitude

less than that at 18.2 pim.

Using the IRAS long wavelength data as a diagnostic Walker & Wolstencroft

(1988) assigned a temperature of T = 115 K to the circumstellar dust around HD 169142.

As we will see this value is low compared to the temperature we derive from our mid-IR

data (T = 175 K). This is not an unexpected result though since Walker & Wolstencroft

fit the 25 and 60 pm fluxes in the SED, while we use 10 and 18 tm data.

Mid-Infrared Observations at Keck

We observed HD 169142 as part of our Vega-like survey at the Keck II telescope

in 1999 May using OSCIR. On the night of 03 May 1999 UT we first detected the

extended emission associated with the source in the N-band (ko = 10.8 um, AX = 5.2 utm)

and IHW18 (Xo = 18.2 glm, AX = 1.7 gm) filters. Our observing sequence consisted of

first observing an 18 gim PSF star (a Boo). Then we observed HD 169142 for 20 minutes

chopped integration (10-min on-source) at 10.8 and 18.2 gtm. After the science

observations we observed a 10 lim PSF star (19 Sgr). Observations of an infrared

photometric standard to absolutely calibrate the HD 169142 data finished the observing

sequence (a Boo, and yAql). For the HD 169142 observations a Boo was used as the 18

mm PSF star since the chosen star was not bright enough to give a good signal-to-noise

detection in a reasonable integration time.

On Keck OSCIR the 128 x 128 pixel detector of OSCIR has a plate scale of

0.062"/pixel which gives a field of view of 7.9" x 7.9". The observations were made

using standard chop/nod techniques with an 8" chopper throw in declination. The Keck

autoguider system was used for all of the Keck observations. We estimate that there is 1

pixel (-0.06" on Keck) of "guiding jitter" inherent in any OSCIR observations made

while using it. On the Keck run in 1999 May the weather was non-photometric with thin,

but uniform, light cirrus clouds present during the night that HD 169142 was observed.

During post-processing of the data, chopped image pairs that were obviously

compromised by the cirrus were discarded. However, there is still some uncertainty in the

photometry of the source and the standard stars, which we take to be 10%. This

translates into the major component of the uncertainties associated with our flux


Measurements of the full-widths at half maximum (FWHM) intensities of

comparison stars were approximately 0.45" and 0.49" at 10.8 and 18.2 glm, respectively.

Quadratic subtraction of the diffraction limits (A/D) of 0.22" at 10.8 gim and 0.37" at 18.2

p.m from these values from implies seeing of -0.3 to 0.4" in the 10 to 20 .m spectral

region. It is our experience that these values are typical for non-photometric nights on

Mauna Kea.

Our measured flux estimates of HD 169142 contain both the thermal emission

from the circumstellar grains, and emission directly from the photosphere of the central

star. To calculate the amount of photospheric emission the star emits in the mid-IR we

normalized the J-band (ken = 1.22 gm) flux estimate of Sylvester et al. (1996) to that of a

blackbody function with effective temperature, Teff = 8400 K. We then predicted that the

photospheric contribution to the 10.8 .m flux density is 41 mJy, or 1.7%. At 18.2 4im, the

photosphere only accounts for 0.2% (14 mJy) of the observed flux. Our photosphere-

subtracted flux estimates for HD 169142 are: 2283 228 mJy at 10.8 jm, and 7187

718 mJy at 18.2 mrn. These flux estimates are in agreement with those of Sylvester et al.

(1996). The 10.8 jm point is also consistent with the IRAS PSC 12 jm value of 2.95 Jy,

however our 18 jm point is low compared to the IRAS 25 jLm estimate of 18.4 Jy. Even

accounting for the spectral shape of the dust SED in this regime the IRAS point seems to

overestimate the flux from the source. The IRAS 25 jpm point also lies above the overall

level of the CGS3 spectrum of HD 169142 presented in Sylvester et al. (1996). We can

only speculate as to the reason for this and an obvious hypothesis is contamination of the

IRAS beam. Since it is the ratio of the fluxes (F10.8/F18.2) that lets us place an upper limit

to the grain sizes in our modeling, we believe the non-photometric quality of the sky will

not fundamentally change our modeling results.

HD 169142 Source Size

In this section we present our new mid-IR observations of HD 169142. These

observations represent the highest resolution mid-IR images currently possible. We

measured a FWHM for HD 169142 of 0.57"at 10.8 gm, and 0.62" at 18.2 gim. Measured

FWHM values of the PSF stars are 0.45" at 10.8 gim, and 0.48" at 18.2 g.m. Subtracting

these values quadratically from the HD 169142 data gives an inherent source size of

0.35" at 10.8 gAm and 0.39" at 18.2 g.m. Interestingly, these values match the peaks of the

intensity of the residual emission of the disk. Our mid-IR data is presented in Figure 4-2

in the form of normalized radial scans along the cardinal directions through HD 169142

and a nearby PSF star (19 Sgr). Also plotted in each of the panels is the residual emission

in the scan direction. Each of the scans through the source and PSF star was normalized

to have a maximum of one and the peaks of the scans were aligned. The residuals were

then calculated by differencing the two scans. An inherent assumption in this method is

that the star sits in the exact center of the emission, and therefore the exact center of the

disk. Recent work on the disk of HR 4796A by Wyatt et al. (1999) has shown that this

may not be true. In that work they introduce the concept of "pericenter glow" where the

long term perturbation of a disk by an object in orbit inside or outside the disk may shift

the disk with respect to its central star (Wyatt et al. 1999).


(a) N band (b) N-band
north-south east- west




-2 -1 0 1 -1 0 2

S (c) 1HW18 (d) IHW18
., north-south east est



0.2 /

-2 -1 0 1 -1 0 1 2
Offset (arcseconds)

Figure 4-2: Normalized scans through HD 169142 and a nearby PSF star. In all plots
HD 169142 is the solid line, the PSF star is the dashed line. The dotted line is the residual
emission, which is the difference of the source scan and the PSF scan. The four panels are
as follows: (a) 10.8 gim north-south direction. (b) 10.8 im east-west direction. (c)
18.2 gm north-south direction. (d) 18.2 pm east-west direction.

In the case of our HD 169142 data small (1 pixel) changes in the position of the

PSF star scan with respect to the source scan produced significant changes in the shape of

the residual emission, but under no circumstances could we make the residuals disappear.

We therefore are confident that the residual emission is present, even within a couple

tenths of an arcsecond from the peak of the emission, but we cannot say confidently say

anything about the structure of the residuals. The scans in Figure 4-2 are our best estimate

at the alignment of the source and PSF star, and therefore our best estimate at the

residuals. The residuals are present in scans through the source at any position angle at

approximately the same level of brightness, which is consistent with the disk being

almost face-on to our line of sight. Additionally, the overall shape of the residuals stays

relatively constant with scan position angle. The approximate face-on orientation of the

disk is also evident in our images. Both the images of the source and of the residuals are

close to circular although there is a slight asymmetry in the outermost region of the 18.2

gm residuals. This asymmetry is detected at a very low level and may well be an artifact

of the PSF removal. In a later section we discuss other evidence that predicts that the disk

should be face-on.

The difference in the widths of HD 169142 and the PSF star is very evident at

both 10.8 and 18.2 gim. Careful inspection of Figure 4-2, especially panels b and c shows

that HD 169142 is broader than the PSF star even in the very core of the scans. We

interpret this as meaning that there is excess emission from the dust coming from very

close to the central source. We see evidence for dust emission within 0.2" in the scans

and images, and it probably reaches in closer to the central source than that. This

interpretation is supported by the fact that HD 169142 exhibits a near-IR excess since

such an excess is caused by hot dust at temperatures of 500 to 1000 K. Calculations

show that blackbody grains need to be within a couple of AU of the star to reach these

temperatures. The fact that there may be dust this close to the star also has implications

for the evolutionary status of the disk that we address in a later section.

Modeling of Characteristic Grain Parameters

Here we present and discuss our modeling of the grains in the HD 169142 disk.

We can use our observations at two wavelengths to estimate basic parameters of the mid-


5 6 7 8 9
Wu cllcngh (.tm)


I6 3

5xI 4

8 10 12 14 16 18
%'a ielcngth (|Im)

20 22 24

Figure 4-3: Ground-based and satellite spectra of HD 169142. This combination of
observations gives complete coverage from 5 to 24 plm, except for the region blocked by
atmospheric CO2 between 13.75 and 16 pm. In particular, note the "UIR" emission bands
between 7 and 9 pLm, and the sharp UIR peak at 11.3 p.m. Also present in the CGS3
spectrum is a broad silicate feature centered around 20 plm. Top panel: ISOPHOT spectra
from Walker & Heinrichsen (2000). Bottom panel: CGS3 spectrum from Sylvester et al.

IR-emitting grains in the disk such as their temperature and approximate size. To derive

these properties we make an assumption about the composition of the grains.

To address the question of the composition of the mid-IR-emitting grains in the

disk of HD 169142 we present the spectra shown in Figure 4-3. Here we see two

moderate resolution spectra of the source that give complete wavelength coverage of the

thermal infrared region. The top panel of Figure 4-3 is an ISO-SWS spectrum reproduced


10 11

I < l l i I I I I

I I X ^ l l l I I

from Walker & Heinrichsen (2000). The bottom panel is a CGS3 spectrum of HD 169142

from Sylvester et al. (1996). In these spectra we see compelling evidence that there are

both silicate grains and polycyclic aromatic hydrocarbons (PAHs) present in the disk.

The features seen between 7 and 9 p.m, and the sharp peak at 11.3 pm are normally

attributed to PAH grains. In fact, HD 169142 was the first Vega-like source to have these

features detected in its 10 p.m spectrum (Sylvester et al. 1994). Also present in the HD

169142 spectrum is a broad silicate feature that dominates the 20 p.m region, which is

caused by the stretching of Si:O bonds in the silicate molecules. Another clue to the

existence of silicate disk material is the fact that optical spectroscopy has shown that HD

169142 has a large under abundance in both Si and Mg. HD 169142 is deficient by 0.86

dex in Si, and 0.56 dex in Mg as compared to the Sun (Dunkin, Barlow, & Ryan 1997).

These values are by far the lowest abundance of these elements of any star in the 13

Vega-like sources studied by Dunkin, Barlow, & Ryan (1997). Dunkin, Barlow, & Ryan

(1997) suggests that the photospheric depletions of these elements is indicative of the

material residing in circumstellar grains, and indeed the presence of the 20 pm silicate

feature seems to support this hypothesis.

These results imply that there are at least two different grain populations in the

disk. This has consequences for our modeling of the grains since our models assume that

the grains are composed of the astronomical silicates of Draine and Lee (1984). However,

since we are working with the global flux estimates of the source it is impossible to de-

couple the emission of the silicate grains from that of the PAHs. To deal with this we

treat the PAH emission as a 'contamination' and run the models assuming that it

contributes varying amounts to the total 10.8 jim flux. Below we discuss our models and

this process in detail.

Model Results

Before we discuss the results of our modeling we introduce the model itself and

synopsize its operation. The University of Florida Dust Dynamics group developed our

model initially to work on the zodiacal dust of the Solar system. We have used these

thermal equilibrium models to analyze similar data on the disk of HR 4796A (Telesco et

al. 2000; Wyatt et al. 1999) and HD 141569 (Fisher et al. 2000; chapter 5). Our models

assume the grains are spherical Mie particles composed of the astronomical silicates of

Draine and Lee (1984) with a density p = 2.5 g cm3.

The results presented there we first need to discuss the operation of our models.

The primary inputs to the modeling code are the optical constants of the grains, and the

characteristics of the grain heating source. In this case the only heating source is HD

169142 itself and we use the data in Table 4-1 as our inputs. We also input a distance to

the source, a vector of grain sizes (diameters in gm) we want the code to model, and a

vector of wavelengths we want the code to calculate fluxes for. Our models then start

with the optical constants of astronomical silicates (Draine and Lee 1984) and use Mie

theory to calculate the absorption efficiencies (Qabs) for different sized particles and

different wavelengths. Once we know how efficiently a particle of a given size absorbs

and emits radiation, we can calculate what its temperature would be at a given distance

from the star. To do this we use the effective temperature of the star to work out the

stellar spectrum, and the luminosity of the star to work out the equilibrium temperature of

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ECPI6PRK0_QJ0RBB INGEST_TIME 2013-11-15T23:33:23Z PACKAGE AA00014222_00001