THE ASTRONOMICAL JOURNAL, 116 : 16261642, 1998 October

( 1998. The American Astronomical Society. All rights reserved. Printed in U.S.A.

BULGE-DISK DECOMPOSITION OF 659 SPIRAL AND LENTICULAR GALAXY BRIGHTNESS PROFILES

W. E. BAGGETT
Science Programs, Computer Sciences Corporation, 3700 San Martin Drive, Baltimore, MD 21218

S. M. BAGGETT
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218

AND
K. S. J. ANDERSON
Department of Astronomy, New Mexico State University, P.O. Box 30001, Department 4500, Las Cruces, NM 88003
Received 1996 October 21 ; revised 1998 June 29

ABSTRACT
We present one of the largest homogeneous sets of spiral and lenticular galaxy brightness prole
decompositions completed to date. The 659 galaxies in our sample have been tted with a de Vaucouleurs law for the bulge component and an inner-truncated exponential for the disk component. Of the
659 galaxies in the sample, 620 were successfully tted with the chosen tting functions. The ts are
generally well dened, with more than 90% having rms deviations from the observed prole of less than
0.35 mag. We nd no correlations of tting quality, as measured by these rms residuals, with either morphological type or inclination. Similarly, the estimated errors of the tted coefficients show no signicant
trends with type or inclination. These decompositions form a useful basis for the study of the light distributions of spiral and lenticular galaxies. The object base is sufficiently large that well-dened samples
of galaxies can be selected from it.
Key words : galaxies : photometry galaxies : spiral
1.

INTRODUCTION

where I (r) is the surface intensity of the bulge at radius r,

B
r is a characteristic
radius dened to be the radius within
e
which half the total light is emitted, and I , the eective
intensity, is simply the surface intensity at r .e(In this paper,
we will consistently use the term surface eintensity, I, to
mean linear intensity units per square arcsecond, and
surface brightness, k, to mean the same quantity in magnitude units.) Similarly, the inner-truncated exponential is
dened by Kormendy (1977) as

In order to compare the large-scale characteristics of galaxies objectively, quantitative measures of the structural
components are necessary. There are many schemes for
describing the structure of galaxies, including Hubble classication (Sandage 1961), isophotal radii (Holmberg 1958),
concentration parameters (Kent 1985 ; Kodaira, Okamura,
& Ichikawa 1990, hereafter PANBG), and the use of various
tting functions (de Vaucouleurs 1953 ; Freeman 1970 ;
Kormendy 1977). All these techniques specify parameters
that can provide insight into the formation and evolution of
galaxies. The use of standardized tting functions is arguably the most powerful method for measuring the largescale structure of galaxies, as the functions yield a variety of
parameters that can be easily compared with the results of
theoretical models. They also provide a reasonably detailed
description of the radial light distribution with a small
number of parameters.
Ideally, tting functions would be based upon the physics
of the formation and evolutionary processes. Unfortunately,
these processes are neither simple nor well understood, so
the most commonly used functions are derived empirically.
Traditional tting functions for elliptical galaxies and spiral
galaxy bulges include the Hubble law (Hubble 1930), the
King model (King 1966), and de Vaucouleurs law (de Vaucouleurs 1953). Recently, there has been some work that
suggests that a generalized version of the de Vaucouleurs
prole (r1@n) provides for better bulge ts (Andredakis, Peletier, & Balcells 1995), and that late-type spirals often have
bulges that are best tted by exponentials (Andredakis &
Sanders 1994). Exponentials (Freeman 1970) and innertruncated exponentials (Kormendy 1977) work well for the
disk components of spiral galaxies. Overall, the de Vaucouleurs law seems to be quite eective as a tting function for
bulges ; it can be written as
I (r) \ I 10~3.33*(r@re)1@4~1+ ,
B
e

I (r) \ I exp M[[r/r ] (r /r)n]N ,

(2)
D
0
0
h
where I (r) is the disk surface intensity at radius r, I is the
D
central intensity
of the disk, r is the disk scale length,0 and r
0 of the disk ( hole radius )h;
is the radius of the central cuto
the pure exponential disk is the same as equation (2) with
r \ 0. Kormendy (1977) found that a value of n D 3 in the
h
truncation
term works well, and we have adopted n \ 3 for
all of the ts presented here. Figure 1 illustrates the usefulness of including a truncation term in the tting function,
using NGC 3145 as an example.
Others have used these tting functions to obtain the
relevant structural parameters for spiral galaxies in order
to compare galaxies of dierent types, luminosities, and
environments. For example, Boroson (1981) tted brightness proles for 26 nonbarred spiral galaxies in order to
determine how the bulge-to-disk ratios are related to the
Hubble types and to investigate the relationship between
spiral and S0 (lenticular) galaxies. Kent (1985) performed a
similar analysis using 105 intrinsically luminous galaxies of
all types. Kormendy (1977) decomposed the brightness proles of seven compact S0 galaxies and one normal galaxy
to check Freemans (1970) hypothesis of a constant central
disk surface brightness. More recently, de Jong (1996) investigated the Freeman (1970) result of a constant central
surface brightness of disks, and some other relationships
between the tting parameters and the Hubble sequence,
using B- and K-band brightness proles of 86 face-on disk

(1)
1626

BULGE-DISK DECOMPOSITION

1627

FIG. 1.Brightness prole ts to the NGC 3145 data. Both plots show the bulge component, the disk component, and the sum of the two. The horizontal
line at 25.3 V -mag arcsec~2 indicates the range of radius included in the t. L eft : Fit without a truncation term in the disk component ; the rms deviation of
this t is 0.28 mag. Right : Fit including the truncation term in the disk component ; the rms deviation of this t is 0.12 mag.

galaxies.
A relatively recent innovation is to t a surface to the
entire galaxy image (Byun & Freeman 1995) using the
above tting functions and also solving for the center and
ellipticity of the projected distributions. A general advantage of this approach is that the bulge and disk components
can be allowed to have dierent ellipticities, which alleviates
the problem of projection eects for moderate- to highinclination systems : because the rounder bulge is typically
sampled at a smaller galactocentric radius than the inclined
disk for a given position in the image, the derived bulge
parameters are systematically too large when estimated
from brightness proles obtained by azimuthal averaging
techniques. However, the proles used here are major-axis
cuts (see below), so this difficulty should not aect our tted
parameters (Burstein 1979). The cost of using major-axis
cuts is that of throwing away much of the information in the
images.
All of these programs except Kormendys (1977)
employed a simple exponential to describe the disk light
distribution. As part of a study of the origin of innertruncated spiral galaxy disks, or type II brightness proles
(Freeman 1970), we have used the de Vaucouleurs law and
the inner-truncated disk (ITD) tting function from
Kormendy (1977) for the bulge-disk decomposition of 659
spiral and lenticular galaxy brightness proles. Our preliminary study (Baggett, Baggett, & Anderson 1993) indicated
that a substantial fraction of all spiral galaxies exhibit an
inner truncation, so the inclusion of such a term in the
tting function seems justied with this large set of bright-

ness proles. Furthermore, the data set used in this study is

extremely homogeneous, all images having been obtained,
reduced, and analyzed in the same way. Thus, the results of
our tting should be a useful resource for many studies of
the large-scale properties of disk galaxies. The following
sections will describe the data and the bulge-disk decomposition procedures and will present the tting results together
with a discussion of the associated errors.
2.

DATA

The brightness proles used for this study were obtained

from the PANBG in a machine-readable form. The initial
PANBG sample of galaxies was selected on the basis of
being included in the Revised Shapley-Ames Catalog of
Bright Galaxies (Sandage & Tammann 1981, hereafter
RSA) and being north of declination [25. Of the 911 such
galaxies in the RSA, 791 are included in the PANBG, and
659 of those have Hubble types (T -types) from the Third
Reference Catalogue of Bright Galaxies (de Vaucouleurs et
al. 1991, hereafter RC3) in the range [3 to 9, which indicates that they are spiral or lenticular galaxies. These 659
galaxies form the basis of our study.
Galaxies in the PANBG were observed photographically
over a period of almost two decades (late 1970s through
1988) with the Kiso Observatory 1.05 m Schmidt telescope,
using Kodak IIa-D plates and a Schott GG 495 glass lter
to dene the photographic V band. Exposure times
ranged from 30 to 60 minutes, with 50 minutes being standard, and the plates were developed in Fuji Pandol or
Kodak D-19 developer. The plates were then digitized with

1628

BAGGETT, BAGGETT, & ANDERSON

the Kiso Observatory PDS microdensitometer utilizing a 1A

square aperture, except for NGC 224 and NGC 598, which
were measured with a 10A square aperture. Each plate
included a step wedge image that was scanned in the same
fashion as the galaxy images. Measured densities were converted into relative intensities via the wedge calibrator, and
the aperture photometry from Longo & de Vaucouleurs
(1983) was then used to transform the resulting magnitudes
to a standard photometric system. The stated internal
photometric accuracy in the PANBG is about 0.1 mag
(standard deviation) and is dominated by errors in the
absolute calibration.
The brightness proles in the PANBG were obtained
from the resulting calibrated images by taking a cut along
the apparent major axis of each galaxy. The major axis was
dened by tting the 25 V -mag arcsec~2 isophote with an
ellipse whose center was xed at the center of gravity of a
21 ] 21 pixel region around the apparent nucleus of the
galaxy. The surface brightness was then sampled along this
axis using a circular aperture that was stepped outward
from the galaxy center in 2 pixel steps (2A for all but NGC
224 and NGC 598, which used 20A steps). The aperture used
was 2 pixels in diameter at the galaxy center, and the diameter was varied in such a way as to be tangent to a sector
with a 5 vertex angle centered on the major axis. This
scheme of varying the aperture size was chosen to compensate for the decreased signal-to-noise ratio in the outer
portions of the galaxies. As a result, there is some radial
smearing of the intensity information at large galactocentric
radii, smoothing structure in the outer portions of each
prole. Further smoothing results from our averaging of the
two halves of the major-axis cut to produce the nal proles. For full details of the data acquisition and reduction
processes, the reader is referred to the PANBG.
3.

FITTING

3.1. Procedure
The major-axis brightness proles were tted using a
combination of a de Vaucouleurs (1953) law (eq. [1]) and an
inner-truncated exponential (Kormendy 1977 ; eq. [2]). The
interactive STSDAS task NFIT1D was used for all of the
tting ; this task uses the downhill simplex algorithm for
performing a nonlinear least-squares t of the data to a
specied function and allows interactive control over the
inclusion of the various parameters and the range of the
data to be tted. It is a very exible routine, and we found
that it accurately returns the values of the tting parameters
in a number of test cases. Fitting is performed on the surface
intensity data and is accomplished by minimizing the
weighted rms deviation of the data from the t.
The most appropriate weighting function, w , is one that
i
uses the variance of the intensity measurement
for each
point as its basis, with the weight of the ith point being
1
w\ ,
(3)
i p2
i
where p2 is the variance of the ith point (Bevington 1969,
i chose to use a weighting based on the Poisson
p. 184). We
distribution, where p P I1@2, as this was consistent with the
i
fashion in which the intensity
measurements were obtained.
An unfortunate side eect of this 1/I weighting function
i
is that it destroys the usefulness of the weighted
rms residual
as a goodness-of-t measure between galaxies. The value of

Vol. 116

the weighted rms residual is highly dependent upon the

tting range, with ts to lower surface intensities being virtually guaranteed a lower weighted rms value than ts stopping at higher surface intensities. As a result, the weighted
rms residual is a useful diagnostic only during the tting
process for a given galaxy, and, as such, the weighted rms
values for each t are not reported here.
However, one can compute the unweighted rms residuals,
([; (k [ k )2]/N)1@2, after the fact, and we have tabulated
i
fit
these values, expressed in magnitude units, as a basis for
assessing the relative quality of the ts. These residuals were
computed from the portions of each brightness prole at
radii larger than 3A and out to the point where the prole
rst drops to a surface brightness fainter than 25 V -mag
arcsec~2. Hence, all of the calculations avoid the portion of
the prole most aected by seeing and reach the same limiting surface brightness. Also, any structure present in the
prole contributes to this measure, and thus galaxies with
signicant structure will be recognizable by their correspondingly larger rms value. In this way, the unweighted
rms residuals are directly comparable from galaxy to galaxy
and reect more accurately the quality of the ts than do
the weighted rms residuals.
Aside from the weighting function, the only other controlled aspect of the tting was the choice of components to
include in each t. In general, all ts were attempted with
both a bulge and a pure exponential disk, resulting in the
estimation of four quantities and their uncertainties : I , r ,
e e
I , and r ; ITDs, with the additional parameter, r , were
0
0
h
utilized only if the prole had the suggestion of a plateau
near the center. If any component (bulge, disk, ITD) was
tted with a negative value for a coefficient, that component
was deemed nonphysical and was removed from the tting
function. In cases where the need for a specic component
was not obvious, ts with and without it were obtained, and
the component was included if the weighted rms value was
smaller by at least 10% than the t without the component.
A sample of the results of the tting are presented in
Table 1, which provides for each galaxy its NGC/IC designation, its revised Hubble type, T -type, and axis ratio from
the RC3, and the t parameters determined here. The t
parameters are given as the tting range in arcseconds (in
the format minimummaximum), the bulge eective surface
brightness (k ), the bulge eective radius (r ), the disk central
e
e
surface brightness
(k ), the disk scale length
(r ), and the
0
disk hole radius (r ). The table also includes the0 seeing as
h
reported in the PANBG,
and the seeing-corrected values of
the tted bulge parameters (k0 and r0 ; see Appendix). The
e
last two columns contain the eunweighted
rms deviation of
the t from the prole in V -mag arcsec~2 and a column of
notes. All surface brightness quantities are in units of
V -mag arcsec~2, and all radii are in units of arcseconds. No
corrections for Galactic extinction, internal extinction, or
inclination have been applied ; the ts are for the observed
major-axis prole. We have chosen to present the results for
the observed proles in order to allow the reader the
opportunity to apply the corrections deemed most appropriate ; we therefore avoid the necessity of uncorrecting our
ts and subsequently applying a dierent correction.
The tting procedure did not include any allowance for
the eects of seeing on the brightness prole. In the Appendix, we describe an experiment that was designed to quantify the impact on the tted bulge parameters of excluding
seeing from the tting ; the net result is that our tted I
e

No. 4, 1998

BULGE-DISK DECOMPOSITION

1629

TABLE 1
SAMPLE GALAXY FIT PARAMETERS
Galaxya
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC

16 . . . . . . .
23 . . . . . . .
148 . . . . . .
151 . . . . . .
157 . . . . . .
224 . . . . . .
237 . . . . . .
245 . . . . . .
253 . . . . . .
254 . . . . . .
255 . . . . . .
268 . . . . . .
274 . . . . . .
278 . . . . . .
289 . . . . . .
309 . . . . . .
357 . . . . . .
404 . . . . . .
428 . . . . . .
473 . . . . . .
514 . . . . . .
521 . . . . . .
524 . . . . . .
598 . . . . . .

Typeb

Tc

.LX.-./
.SBS1..
.L..0*/
.SBR4..
.SXT4..
.SAS3..
.SXT6..
.SAT3P$
.SXS5..
RLXR]*.
.SXT4..
.SBS4*.
.LXR-P.
.SXT3..
.SBT4..
.SXR5..
.SBR0*.
.LAS-*.
.SXS9..
.SXR0*.
.SXT5..
.SBR4..
.LAT]..
.SAS6..

[3
1
[2
4
4
3
6
3
5
[1
4
4
[3
3
4
5
0
[3
9
0
5
4
[1
6

R d
25
0.27
0.19
0.4
0.34
0.19
0.49
0.24
0.06
0.61
0.21
0.08
0.14
0.01
0.02
0.15
0.08
0.14
0
0.12
0.2
0.1
0.04
0
0.23

Rangee
370
358
368
365
311, 92148
1005600
38, 2750
346
3276, 725936
382
370
352
440
3102
360
374
319, 5780
3288
492
358
394
394
4136
32256

kf
e
20.83
7.18
12.3
17.95
20.85
19.89
...
20.11
21.19
15.94
21.98
...
16.67
20.36
...
11.88
18.5
22.15
...
19.22
24.17
21.26
21.47
22.71

rg
e
13
0.1
0.4
2.4
7.2
282.2
...
3.9
18.7
1.4
12
...
2
17.1
...
0.2
3.5
63.8
...
4.2
38.5
10.9
33.6
109.6

k h
0
22.41
19.83
19.24
20.93
20.56
20.58
19.99
19.24
19.51
20.09
19.97
19.86
16.79
...
18.81
20.9
22.35
23.02
20.19
20.44
20.97
20.9
...
20.3

r i
0
18.8
11.5
12.5
34.4
35.8
1781.2
10.4
8.8
192.3
18.5
14.7
11.2
5
...
14
21.4
31.5
129.5
20.7
13.8
24.4
26.6
...
533.3

rj
h
...
...
...
...
...
...
...
10.1
...
...
12.5
...
17.5
...
...
...
...
...
...
...
13.2
20.8
...
...

rmsk

Seeingl

0.07
0.21
0.11
0.54
0.51
0.13
0.17
0.12
0.4
0.12
0.12
0.25
0.22
0.19
1.12
0.25
0.38
0.09
0.12
0.1
0.12
0.1
0.11
0.12

1
1
5
2
2
4
2
2
5
5
2
3
3
3
5
3
3
2
5
5
5
5
3
2

k0 m
e
20.67
...
13.64
17.88
20.74
...
...
20.02
21.3
16.93
21.87
...
16.87
20.32
...
...
18.6
22.05
...
19.69
24.21
21.46
21.41
...

r0 n
e
12.37
...
0.5
2.37
7.01
...
...
3.82
19.73
1.75
11.61
...
2.11
16.95
...
...
3.62
60.86
...
4.85
39.51
11.82
32.92
...

Noteso
D
A, E
A
A
A
D
I
A, D
A, E

NOTES.A, Prominent arm/bar/ring/dust lane removed from t ; B, much structure in prole ; C, truncated disk t to a bright arm or lens ; D, bar near
major axis ; E, bar near minor axis ; F, bar between axes ; G, edge-on ; H, box/peanut bulge evident in published image ; I, interacting ; J, NGC 4891 is not
included in the RC3, but it appears to be listed as NGC 4897. The RC3 data listed are those for NGC 4897. Table 1 is presented in its entirety in the
electronic edition of the Astronomical Journal. A portion is shown here for guidance regarding its form and content.
a Galaxy ID.
b Revised Hubble type from RC3.
c T -type from RC3.
d R [\log(a/b)] from RC3.
25
e Fitting
range(s), in units of arcseconds.
f Bulge eective surface brightness in V -mag arcsec~2.
g Bulge eective radius in units of arcseconds.
h Disk central surface brightness in V -mag arcsec~2.
i Disk scale length in units of arcseconds.
j Disk truncation radius in units of arcseconds.
k Unweighted rms deviation of the t in V -mag arcsec~2.
l Seeing in arcseconds from the PANBG.
m Fully corrected bulge eective surface brightness in V -mag arcsec~2.
n Fully corrected bulge eective radius in units of arcseconds.
o Notes about individual galaxies.

tends to be too large by up to an order of magnitude when

the seeing is much larger than the input eective radius, and
our tted r tends to be slightly too small in the same cases.
This is whate is intuitively expected, of course.
Although the prole denition procedure introduces a
signicant level of radial smearing, structure is still apparent in many of the proles. The tting process nominally
made no allowance for the presence of structure, tting
across arms and bars as if they were simply noise in the
data, except in some specic instances. These instances
occur when a very strong, isolated feature is present ; then
the radial range occupied by the feature was excluded from
the t. If a strong feature is present at the end of the tting
range, the tting range was suitably shortened. The existence of such a condition was manually determined, and is
indicated in Table 1 by the presence of multiple ranges in
the tting range column and/or by a note (e.g., A ) in the
last column of the table.
Figure 2 presents the proles and ts for the galaxies
from Table 1. Each plot shows the observed prole from the
PANBG as crosses, and the tted bulge and disk components and their sum as solid lines. In addition, the range
of radius included in the t is indicated by the horizontal

line(s) at the 25.3 V -mag arcsec~2 level, and the t parameters are provided near the top of each plot. This selection
of objects includes some very good ts as measured by the
rms deviations (NGC 16, NGC 404), some typical-quality
ts (NGC 224, NGC 237), some ts that avoid strong structure in the prole (NGC 151, 157, 253), and one of the worst
ts in the entire sample (NGC 289).
There were 39 galaxies in the sample for which no ts
were obtained ; these objects are listed in Table 2, with their
revised types and axis ratios from the RC3. These galaxies
are particularly ill-suited for tting by the selected tting
functions, as many of them have pronounced concavities
toward low surface brightness or multiple exponential components in their brightness prolesthe chosen functions
simply do not represent their light distributions in any
meaningful way.
Finally, there are also several ts (e.g., NGC 16, 628, 890,
and 5033) where the disk component is everywhere fainter
than the bulge, and other ts where the disk becomes
brighter than the bulge only at intermediate radii (e.g.,
NGC 670, 955, 1090, and 1187). In both of these situations,
the bulge t might have beneted from an alternative functional form, perhaps the generalized de Vaucouleurs law as

FIG. 2.Brightness prole ts to the galaxies presented in Table 1. Each plot shows the observed prole (crosses), the bulge and disk components and
their sum (solid lines), and the range of radius included in the t (horizontal line(s) at bottom). Also shown are the tted parameters, where the surface
brightness parameters are in units of V -mag arcsec~2 and the size parameters are in arcseconds.

1630

FIG. 2.Continued

1631

1632

BAGGETT, BAGGETT, & ANDERSON

Vol. 116

TABLE 2
NONFITTED GALAXIES
Galaxy

Type

NGC 150 . . . . . . .
NGC 275 . . . . . . .
NGC 337 . . . . . . .
NGC 908 . . . . . . .
NGC 941 . . . . . . .
NGC 1035 . . . . . .
NGC 1358 . . . . . .
NGC 1667 . . . . . .
NGC 2146 . . . . . .
NGC 2633 . . . . . .
NGC 2793 . . . . . .
NGC 2976 . . . . . .
NGC 2990 . . . . . .
NGC 3003 . . . . . .
NGC 3021 . . . . . .
NGC 3043 . . . . . .
NGC 3067 . . . . . .
NGC 3312 . . . . . .
NGC 3432 . . . . . .
NGC 3455 . . . . . .
NGC 3556 . . . . . .
NGC 3628 . . . . . .
NGC 3664 . . . . . .
NGC 3717 . . . . . .
NGC 3810 . . . . . .
NGC 4013 . . . . . .
NGC 4085 . . . . . .
NGC 4302 . . . . . .
NGC 4487 . . . . . .
NGC 4517 . . . . . .
NGC 4618 . . . . . .
NGC 5112 . . . . . .
NGC 5170 . . . . . .
NGC 5301 . . . . . .
NGC 5474 . . . . . .
NGC 5746 . . . . . .
NGC 5775 . . . . . .
NGC 5949 . . . . . .
IC 764 . . . . . . . . . .

.SBT3*.
.SBT6P.
.SBS7..
.SAS5..
.SXT5..
.SAS5$.
.SXR0..
.SXR5..
.SBS2P.
.SBS3..
.SBS9P.
.SA.5P.
.S..5*.
.S..4$.
.SAT4*.
.S..3*/
.SXS2$.
.SAS3P$
.SBS9./
PSXT3..
.SBS6./
.S..3P/
.SBS9P.
.SA.3*/
.SAT5..
.S..3./
.SXS5*$
.S..5*/
.SXT6..
.SAS6*/
.SBT9..
.SBT6..
.SAS5*/
.SAS4*/
.SAS6P.
.SXT3$/
.SB.5$/
.SAR4$.
.SAS5$.

25
0.32
0.14
0.20
0.36
0.13
0.48
0.10
0.11
0.25
0.20
0.07
0.34
0.27
0.63
0.25
0.48
0.42
0.42
0.66
0.21
0.59
0.70
0.03
0.73
0.15
0.71
0.55
0.74
0.17
0.83
0.09
0.15
0.91
0.69
0.05
0.75
0.62
0.33
0.47

discussed by Andredakis et al. (1995), although no attempt

has been made to investigate this in the present study.
3.2. Fitting Errors
We have chosen three methods of estimating the tting
errors : (1) unweighted rms residuals, (2) error estimates provided by the tting software, and (3) comparison with the
results of other workers.
The unweighted rms residuals for each t are tabulated in
Table 1. Because we have computed them in a uniform
fashion for all objects, including all structure in the proles,
these values provide an unbiased and consistent measurement of how well the tting functions and the determined
parameters describe the observed brightness distribution.
From these values we nd that the median rms deviation of
the ts is only 0.15 mag and that more than 90% of the ts
are better than 0.35 mag ; Figure 3 shows the distribution of
the rms deviations using bins of width 0.05 mag. The overall
ability of the ts to describe the brightness distributions is
quite good, given that the unweighted rms residuals include
all of the structure present in the prole.
In an eort to quantify the value of the unweighted
residuals as a measure of the goodness of t, we have examined plots of the proles and selected the best examples of
proles that are without signicant structure and appear to
be well represented by the chosen tting functions. Note

Comments
Very bright arms at r D 50A
Concave to low k (ITD with no bulge ?)
Concave to low k
Exponential disk truncated at D150A
Strong structure in outer prole
Concave to low k
Large, bright plateau with sharp outer cuto
Three exponentials ?
Very disturbed galaxy
Strong structure throughout prole
Very asymmetric prole
Concave to low k
Concave to low k
Strong structure throughout prole
Concave to low k
Concave to low k
Concave to low k
Large, bright plateau with sharp outer cuto
Strong structure throughout prole
Concave to low k, faint outer extension
Strong structure throughout prole
Disturbed edge-on system
Strong structure throughout prole
Double exponential, edge-on
Strong structure in outer prole
Concave to low k, edge-on
Concave to low k
Extremely edge-on
Double exponential
Extremely edge-on
Strong structure in outer prole
Strong structure in outer prole
Concave to low k, edge-on
Strong structure in outer prole
Very asymmetric prole
Concave to low k, edge-on
Concave to low k, edge-on
Concave to low k
Strong structure throughout prole

that this selection did not involve consideration of the computed unweighted rms residuals. There were 16 objects
included in this selection, and they have a mean unweighted
rms deviation of 0.05 mag and a range of 0.030.10 mag ;
this suggests that ts with unweighted rms residuals greater
than about 0.10 mag are aected to some degree by signicant structure and/or poor tting quality. The full sample
contains 161 galaxies with rms residuals of 0.10 mag or less.
At the other end of the distribution, the worst-t galaxies
have been investigated to attempt to nd out why they were
so poorly tted. We have inspected the 18 galaxies that have
unweighted rms residuals greater than 0.5 mag to search for
common characteristics such as morphology and inclination. The T -type distribution of these 18 galaxies is essentially the same as for the sample as a whole, so there appears
to be no correlation between poor t and T -type. There also
appears to be no serious trend with inclination : the mean
inclination of the group is 57, consistent with a random
distribution of tilts. The most common characteristic is a
low surface brightness extension to the brightness prole,
such that the t falls below the measurements in the outer
portion. The extension is sometimes featureless, and sometimes it contains a distinct bump (as if an outer ring or arm) ;
sometimes it is nearly constant brightness, and other times
it is more or less parallel with the inner prole. There is only
one case (NGC 157) where the problem region is in the

No. 4, 1998

BULGE-DISK DECOMPOSITION

1633
TABLE 3
PARAMETER FRACTIONAL ERRORS

Galaxy
(1)

FIG. 3.Distribution of rms deviations of the ts from the data. The

rms deviations were computed for all data points in each prole between 3A
and the rst point at which the surface brightness drops below 25 V [ mag
arcsec~2. The median value of the distribution is at 0.15 mag. There are six
objects with rms deviations greater than 1.0 mag that are not included in
the gure.

main portion of the disk, and this galaxy looks disturbed,

almost as if undergoing a collision. The apparent overriding
reason for the poor ts is simply that the chosen tting
functions do not work well for some galaxies. This same
conclusion holds for the 39 galaxies that were not tted in
this eort : common characteristics of these objects are
extremely strong, large-scale structures, concavity of the
brightness prole toward faint surface brightness, and
apparent multiple components of the prole, usually with
more than one exponential. This occasional inappropriateness of the tting functions suggests that careful consideration of the tted parameters and their error estimates
should be exercised before judging whether a specic t is
truly meaningful for a detailed study of any individual
galaxy.
The second method of judging the tting errors was the
use of estimated coefficient errors as produced by the tting
software. The NFIT1D task estimates the errors in each
parameter by a process of bootstrap resampling, with a
choice of distribution functions for use in the procedure. We
used the Poisson distribution as the distribution function
for the parameter error estimation since this reects the
photon statistics expected to aect the measured relative
intensities in the brightness proles. In order to minimize
the eects of comparing parameters that vary wildly in
value from galaxy to galaxy, we have computed the fractional error in each tted parameter. Table 3 lists these
fractional error estimates for a sample of the galaxies :
column (1) gives the galaxy identication and columns (2)
(6) the fractional error of each of the parameters. Note that
since the tting was performed in the surface intensity scale
(not surface brightness), the relative errors are computed in
linear units, not magnitude units.
The fractional errors for the full set of galaxies are summarized in Table 4, which gives for each parameter the

NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC
NGC

16 . . . . . . .
23 . . . . . . .
148 . . . . . .
151 . . . . . .
157 . . . . . .
224 . . . . . .
237 . . . . . .
245 . . . . . .
253 . . . . . .
254 . . . . . .
255 . . . . . .
268 . . . . . .
274 . . . . . .
278 . . . . . .
289 . . . . . .
309 . . . . . .
357 . . . . . .
404 . . . . . .
428 . . . . . .
473 . . . . . .
514 . . . . . .
521 . . . . . .
524 . . . . . .
598 . . . . . .

*I
e
(2)

*r
e
(3)

*I
0
(4)

*r
0
(5)

*r
h
(6)

0.212
0.365
0.520
0.255
0.601
0.078
...
0.337
0.671
0.258
0.250
...
0.117
0.026
...
4.950
0.257
0.057
...
0.365
0.227
0.070
0.024
0.202

0.125
0.094
0.157
0.099
0.380
0.047
...
0.165
0.350
0.110
0.174
...
0.053
0.014
...
0.620
0.121
0.052
...
0.307
0.248
0.042
0.014
0.162

0.447
0.023
0.031
0.044
0.117
0.018
0.023
0.053
0.010
0.038
0.035
0.012
0.154
...
0.007
0.021
0.230
0.158
0.010
0.114
0.032
0.038
...
0.011

0.087
0.008
0.012
0.035
0.035
0.010
0.010
0.018
0.007
0.014
0.018
0.008
0.028
...
0.004
0.014
0.115
0.080
0.007
0.014
0.028
0.017
...
0.006

...
...
...
...
...
...
...
0.019
...
...
0.021
...
0.015
...
...
...
...
...
...
...
0.036
0.021
...
...

NOTES.Col. (1), galaxy ID ; col. (2), fractional error in I ; col.

e (5),
(3), fractional error in r ; col. (4), fractional error in I ; col.
e (6), fractional error in r . Table
0
fractional error in r ; col.
3 is
0 in the electronic edition of theh Astronomipresented in its entirety
cal Journal. A portion is shown here for guidance regarding its
form and content.

number of error estimates, the mean fractional error, the

standard deviation of the fractional errors, the median fractional error, and the minimum and maximum fractional
errors. As can be seen in Table 4, every parameter except I
is signicantly aected by outliers ; histograms of the frac-0
tional error of each parameter are shown in Figure 4, which
demonstrate this problem. Removing the most discrepant
outlier from the statistical calculations (second part of
Table 4) improves the results considerably. Unfortunately,
there is not just one bad t in the sample causing all of
these outlying points : the bulge parameter errors are both
dominated by the t for NGC 2441, while the disk central
surface intensity error is aected by the t for NGC 2541,
the disk scale length error by NGC 5033, and the hole
radius error by NGC 2997.
The bulge component of NGC 2441 dominates the disk
only at radii smaller than about 6A, so there are very few
data points dening the bulge, and the coefficient uncertainties reect this fact. The prole for NGC 2541 is tted
with a faint ITD (the peak disk brightness is only about 24
V -mag arcsec~2), and the hole radius is more than six disk
scale lengths from the galaxy center. The southern half of
the brightness prole contains a relatively bright spiral arm
(about 23 V -mag arcsec~2 at its brightest), and this asymmetry results in a bump in the averaged prole, which has
been tted with the inner truncation. NGC 5033 was tted
with a very faint (k \ 25.1 V -mag arcsec~2), very at disk
(r \ 732A) that was0 never comparable in brightness to the
0
bulge.
As a result, the disk parameters for this galaxy are
not well constrained by the data, and the estimated errors in
the coefficients are correspondingly large. Finally, NGC
2997 has been tted with a hole radius just smaller than the

1634

BAGGETT, BAGGETT, & ANDERSON

Vol. 116

TABLE 4
FRACTIONAL ERROR SUMMARY
Parameter

Number

Mean

Standard
Deviation

Median

Min

Max

0.126
0.082
0.040
0.016
0.019

0.014
0.010
0.000
0.003
0.005

20508
19.4
1.03
917.9
0.160

0.014
0.010
0.000
0.003
0.005

1241.3
5.0
0.767
1.304
0.068

All Fitted Galaxies

I .........
e
r ..........
e
I .........
0
r .........
0
r ..........
h

507
507
559
559
156

43.2
0.194
0.075
1.67
0.022

912.3
0.910
0.110
38.8
0.015

Most Discrepant Galaxy Removed

I .........
e
r ..........
Ie . . . . . . . . .
0
r .........
0
r ..........
h

506
506
558
558
155

2.79
0.156
0.073
0.027
0.021

55.2
0.317
0.102
0.064
0.010

radius of the innermost data point, so the value of the hole

radius is, again, not really constrained by the data.
Because of the presence of these extreme outliers, the
median fractional errors are much more useful than the
mean for examining the tting errors of the sample as a
whole. The summary in Table 4 shows that the median
coefficient fractional errors range from a low of 1.6% for the
disk scale length to more than 12% for the bulge eective
intensity. From these data, it is clear that the bulge parameters are the least well determined quantities in the ts,
while the disk parameters are generally well determined.
This is not unexpected : since the bulge coefficients are
usually dominated by a relatively small number of data
points, the constraints on them are not very strong.
Finally, a few of the galaxies included in this study have
been previously tted by others with the same tting functions, providing a completely independent check of the
results of our tting procedure. There are ve galaxies in the
current study that are also included in Boroson (1981), and
12 are in common with Kent (1985), making a total of 17
measurements available for use in this comparison. Prior to
making any comparisons between the various works, the

0.125
0.082
0.040
0.016
0.019

surface brightness parameters reported in Boroson (1981)

and Kent (1985) have been uncorrected for the eects of
inclination and galactic absorption as applied in each study,
and the length parameters have been converted from kiloparsecs to arcseconds using the distances adopted by those
authors. Furthermore, Boroson (1981) lists values for the
disk B[V color, which have been used to convert his disk
central surface brightnesses from the B bandpass to the V .
Finally, there are two galaxies in common between Boroson
(1981) and Kent (1985), and the same analysis has been
applied to them. The percentage dierences between the
tting parameters from these works and the present study
have been computed and tabulated in Table 5. In this table,
we list for each galaxy the percentage dierence between the
parameters in the referenced work and this study, and also
between the two reference works ; the last column identies
the reference work. The dierences were computed in the
sense reference work minus this study, and Boroson minus
Kent. These data are also presented in Figure 5, which
shows the tting parameters from Boroson (1981) and Kent
(1985) plotted against the values obtained in the present
study.

TABLE 5
FITTING COMPARISONS
Galaxy
NGC 16 . . . . . . . . . . .
NGC 628 . . . . . . . . . .
NGC 670 . . . . . . . . . .
NGC 2268 . . . . . . . .
NGC 2639 . . . . . . . .
NGC 2683 . . . . . . . .
NGC 2776 . . . . . . . .
NGC 2782 . . . . . . . .
NGC 2880 . . . . . . . .
NGC 3627 . . . . . . . .
NGC 3898 . . . . . . . .
NGC 5380 . . . . . . . .
NGC 5533 . . . . . . . .
NGC 5676 . . . . . . . .
NGC 5970 . . . . . . . .
NGC 6340 . . . . . . . .
NGC 7331 . . . . . . . .
Comparison :
NGC 488 . . . . . . .
NGC 2967 . . . . . .

P.D.(I )
(%) e

P.D.(r )
(%) e

P.D.(I )
(%) 0

P.D.(r )
(%) 0

Reference

80
146
35
88
[114
[125
87
8
[180
41
[168
[154
0
[200
[142
[193
[23

[56
[169
[11
[108
73
105
[89
[13
142
[46
76
69
17
167
69
154
[49

159
[8
[12
25
11
30
55
171
[135
118
[65
82
34
43
[2
[53
[30

[14
42
18
5
[4
[13
5
[120
27
[42
50
[23
[9
[2
21
24
16

1
2
1
2
1
1
1
1
1
1
2
1
1
1
1
2
2

112
195

[127
[180

[12
[37

[7
11

2
1

REFERENCES.(1) Kent 1985 ; (2) Boroson 1981.

No. 4, 1998

BULGE-DISK DECOMPOSITION

1635

FIG. 4a

FIG. 4b

FIG. 4c

FIG. 4d

FIG. 4e
FIG. 4.Histograms of the fractional errors. The distributions of the estimated fractional errors of each tted parameter are shown to illustrate the
problem with outliers : (a) I , (b) r , (c) I , (d) r , and (e) r . Note the smaller bin size for r and r .
e
e
0
0
h
0
h

In general, the ts from the various studies do not agree

very well, although the disk ts are typically more similar
than the bulge ts. The mean of the absolute values of the
percentage dierences are 105% for I , 83% for r , 61% for
e all parameters.
e
I , and 26% for r , with 17 objects for
Our
0
0
bulge ts agree somewhat better with those of Kent (1985),
while our disk central surface brightnesses are closer to
Borosons (1981) resultsthis reects the conversion of
Borosons surface brightnesses to the V band. Interestingly,

our values of r agree equally well, on average, with both

reference works,0 leading us to believe that a 25% scatter in
the disk scale length is to be expected under the circumstances of this comparison (diering bandpasses).
The bulge t dierences are dominated by the eects of
the prole acquisition procedures used in the dierent
studies : this work used a wedge-shaped major-axis cut,
while Boroson (1981) and Kent (1985) used two variations
on azimuthal averaging. In particular, it has been pointed

1636

BAGGETT, BAGGETT, & ANDERSON

Vol. 116

FIG. 5.Comparison of tted parameters. The t parameters for the galaxies in common with Boroson (1981) (plus signs) and Kent (1985) (asterisks) are
shown along with the 45. line. The surface brightness parameters are expressed in units of mag arcsec~2, and the length parameters are given in arcseconds.
There is a reasonable correlation for all of the parameters, although the disk parameters have tighter correlations than those for the bulge.

out by Boroson (1981) that azimuthally averaged proles

will be systematically too bright in the bulge-dominated
regions owing to sampling the rounder bulge at a smaller
galactocentric radius than the disk, at a given position in
the image. Inspection of Table 5 reveals that every instance
of a large departure in I shows a large departure in r of the
e with
opposite sign. That our ebulge parameters agree better
Kent (1985) than with Boroson (1981) probably reects the
use of xed ellipticities and position angles by Boroson,
while Kent allowed those quantities to vary with radius. As
a result, the mean surface brightness around ellipses in
bulge-dominated regions are more representative of the
major axis with Kents prole acquisition procedure than
with Borosons, hence the slightly better agreement with our
major-axis cuts.
4.

FITTING CHARACTERISTICS

The basic result of the prole decomposition process is

that 620 of the 659 proles were successfully tted with the

chosen tting functions. In this section, we will discuss some

of the characteristics of the prole decomposition and the
errors as they appear with this sample of galaxies. Our
intent is to provide information regarding the characteristics of the tting ; we will present our analysis regarding
the structure of disk galaxies elsewhere.
4.1. Fitting Parameters and Morphology
All galaxies included in this sample have lenticular or
spiral RC3 Hubble types ; however, 61 of the ts have been
made with no disk component, indicating a lack of any
appreciable exponential component to their brightness proles. Visual inspection of these galaxies in the PANBG
shows many of them to have what appear to be disks even
though the prole shows none (e.g., NGC 1784, 1961, and
3370) ; others are possibly misclassied elliptical galaxies
(e.g., NGC 2655, 3998, and 5485). The galaxies with no disk
component span a range in T -type from [3 to 9, the entire
range of T -type included in this sample. Early-type galaxies

No. 4, 1998

BULGE-DISK DECOMPOSITION

1637

TABLE 6
MEDIAN FRACTIONAL ERROR MORPHOLOGY DEPENDENCE
T -Type Range
[3.5 T
[0.5 T
2.5 T
5.5 T

*I

\ [0.5 . . . . . .
\ 2.5 . . . . . . . .
\ 5.5 . . . . . . . .
\ 9.5 . . . . . . . .

0.09
0.12
0.14
0.24

e
(131)
(114)
(204)
(58)

*r
e
0.05 (131)
0.07 (114)
0.10 (204)
0.17 (58)

*I
0.06
0.05
0.03
0.02

0
(119)
(100)
(246)
(94)

*r
0
0.03 (119)
0.02 (100)
0.01 (246)
0.01 (94)

*r
h
0.02 (41)
0.02 (27)
0.02 (71)
0.02 (17)

NOTE.The numbers in parentheses are the number of galaxies contributing to the median
fractional error.

are somewhat more likely to be tted without a disk than

are late types, but the trend is not very strong. There is no
signicant dierence in the t quality (as measured by the
tabulated rms deviation) between the galaxies without disk
ts and the sample as a whole.
A total of 113 galaxies were tted without a bulge component, and although these galaxies range in T -type from
[2 to 9, they are mostly late-type galaxies. This, of course,
is consistent with the basic behavior of the Hubble classication scheme, where bulges are less prominent in later
types. Again, the basic tting quality is the same for galaxies
without bulge ts as with the sample in general.
An investigation of the variation of the median fractional
errors with T -type is summarized in Table 6. This table
gives for each range of T -type the median fractional error of
each tting parameter, as well as the number of galaxies
included in each median determination. The errors in I , r ,
e e
and I show some slight trends that are in the expected
0
senses but which are small enough to be of questionable
signicance. The two bulge parameters seem to have somewhat larger median errors for later types, as would be
expected as the bulge contribution to the light distribution
decreases. We should also note the work of Andredakis &
Sanders (1994), who show that the inner regions of late-type
spirals are perhaps better represented by an exponential
light distribution than a de Vaucouleurs law. Similarly, the
disk central intensity error has larger values at earlier types,
when the disk contribution is generally lower.
We also looked at the possibility of a tting quality
dependence on the presence or lack of a bar, and we have
found nothing signicant. The median rms residual of the
ts on nonbarred galaxies (Hubble type in the RC3 contains
an A explicitly) is about 0.14 mag, for barred galaxies
( B ) it is 0.15 mag, for mixed types ( X ) 0.16 mag, and for
objects with no bar classication given in the RC3 we nd a
value of 0.14 mag. All of these values are sensibly the same
as the median value of 0.15 mag found for the full sample,
and there is clearly no trend apparent. Table 7 lists the
median fractional error of the individual tting parameters
for each bar class ; there are no signicant trends in these
results. We conclude that bars have no discernible eect on
the quality of the ts.

Objects that have been tted with an inner truncation

make up about 25% (156/620) of the sample. Remembering
that inner truncations were included only if they improved
the weighted rms by at least 10%, this serves to justify our
initial decision to use that function with this large data set.
Some of these inner truncations are probably caused by
arms or rings at large galactocentric radii (e.g., NGC 2859,
3368, and 5701)the arm/ring is bright relative to the local
disk and thus mimics an inner truncation. In these cases, the
arm/ring is typically faint (peaks near 24 V mag arcsec~2)
and has a short disk scale length. These galaxies are also
generally classied as having an outer ring. These objects
are identied in Table 1 with a C in the comment
column. A more detailed analysis of the presence of an inner
truncation is left for a later study.
4.2. Fitting Parameters and Inclination
An analysis of the unweighted rms residuals shows little
or no trend in the mean value with inclination. We divided
the sample into three inclination ranges based solely on the
R value listed in the RC3, assuming that this isophote
25
corresponds
to a at, circular disk : i 30 (R 0.0625),
30 \ i 60 (0.0625 \ R 0.301), and i [2560 (R [
0.301). Note that no T -type25dependence was included in25this
inclination estimate. The resulting mean rms residuals for
the low-, medium-, and high-inclination samples are 0.16,
0.19, and 0.21 mag, respectively. These three values are all
much less than 1 p from each other, so the trend is statistically meaningless. The median rms residuals for each
inclination sample are 0.15, 0.14, and 0.17 mag for the low,
medium, and high inclinations, respectively. We conclude
that the sample has no signicant inclination dependence
on the unweighted rms residuals of the ts.
The quality of the individual tting components is investigated by computing the median fractional errors within
each inclination range ; the mean fractional errors are not
useful because of the outliers discussed previously. Table 8
provides the results of this investigation, which shows that
there are no indications of any inclination dependencies. In
Table 8, the rst column lists the parameters, and the next
three columns give the median relative errors in each inclination range.

TABLE 7
MEDIAN FRACTIONAL ERROR BAR CLASS DEPENDENCE
Bar Class
SA . . . . . . . . . . . . . .
SB . . . . . . . . . . . . . .
SX . . . . . . . . . . . . . .
Not typed . . . . . .

*I
e
0.11 (173)
0.12 (154)
0.16 (148)
0.16 (32)

*r
e
0.08 (173)
0.08 (154)
0.09 (148)
0.07 (32)

*I
0.04
0.05
0.04
0.03

0
(173)
(176)
(170)
(40)

*r
0
0.02 (173)
0.02 (176)
0.01 (170)
0.01 (40)

*r
0.02
0.02
0.02
0.02

h
(42)
(61)
(44)
(9)

NOTE.The numbers in parentheses are the number of galaxies contributing to the

median fractional error.

1638

BAGGETT, BAGGETT, & ANDERSON

TABLE 8
MEDIAN FRACTIONAL ERROR INCLINATION DEPENDENCE
Error

i 30

30 \ i 60

i [ 60

*(I ) . . . . . .
0.12 (68)
0.12 (291)
0.14 (148)
*(r e) . . . . . .
0.09 (68)
0.08 (291)
0.09 (148)
e
*(I ) . . . . . .
0.05 (76)
0.05 (283)
0.02 (200)
0
*(r ) . . . . . .
0.02 (76)
0.02 (283)
0.01 (200)
0
*(r ) . . . . . .
0.02 (24)
0.02 (92)
0.02 (40)
h
NOTE.The numbers in parentheses are the number of
galaxies contributing to the median fractional error.

Finally, we check if the rate of occurrence of the inner

truncation has any inclination dependence by computing
the rate in each inclination range. The rates are
28% ^ 6.5% for the low-inclination group, 28% ^ 3.3% for
the medium inclinations, and 19% ^ 3.3% for the highinclination sample. Thus, there is no trend signicant at the
2 p level.
5.

CONCLUSIONS

We have presented one of the largest, if not the largest,

collections of spiral and lenticular galaxy brightness prole
bulge-disk decompositions yet completed. Of the 659
brightness proles in our sample, 620 were tted with the de
Vaucouleurs law plus inner-truncated exponential disk
function, while the remaining 39 proles could not be so
tted. The general quality of the ts is quite high, with
about 50% having an unweighted rms deviation from the
data (including real structures) of less than 0.15 mag and
more than 90% having unweighted rms residuals of less
than 0.35 mag. We nd no systematic trends in the tting
quality with either galaxy morphology or inclination. Comparison of our ts with those of Boroson (1981) and Kent
(1985) show discrepancies attributable to a number of
observational and data reduction factors.
Probably the most interesting result from this process of
tting is simply that we achieved a success rate of
620/659 (94%) in our ts, compared to success rates of 75/94
(80%) for Kent (1985 ; spiral and S0 galaxies only) and 16/26
(62%) for Boroson (1981). While these success rates are statistically similar, we wonder if the small number statistics
are the only dierences. An analysis of the 19 nonttable
disk galaxy proles from Kent (1985) shows that we found
ts for all 10 of those galaxies that were also in our sample.

Vol. 116

For these 10 objects, the mean inclination is 41 ^ 18, the

median is 42, and only one galaxy (NGC 5566) has an
inclination greater than 60. Thus, Kents nonttable galaxies do not tend to be high-inclination objects. Furthermore, the rms deviations of our ts for these same galaxies
are generally small, with a median value of 0.15 mag, the
same as for our full sample. The principal dierence seems
to be that we included the ITD factor in our ts : seven of
these 10 galaxies have ITDs in our ts, often with large r /r
h 0
ratios. It is also possible that Kents simultaneous tting of
the minor-axis proles made his results more sensitive to
deviations from the standard tting functions. A similar
analysis of the 10 nonttable proles in Borosons (1981)
sample shows us having ts for the nine that are included in
our sample. These nine galaxies are also of relatively low
inclination (the largest is about 58) and have small rms
deviations in our ts (median value of 0.13 mag). However,
our tted parameters show three objects with bulge only,
three with a bulge plus exponential disk, and three with a
bulge plus ITD ; the case for the inclusion of the ITD is not
as strong with this set of proles. We conclude, however,
that the inclusion of the ITD term in our tting functions
has allowed us to t 10%15% more galaxies than we
would have t without the inner-truncation term.
It is also interesting that about 25% of the proles in our
sample are tted better by an ITD function than with a
plain exponential. Some of these ts are certainly due to the
ITD being tted to outer rings (as indicated by a C in the
last column of Table 1), and others may be marginal
improvements (remember the requirement for a 10%
improvement of the weighted rms to include the innertruncation term), but clearly a signicant fraction of the
proles support the physical reality of the inner truncation
in the light distribution. A quantication of the strength of
the inner truncation and the search for the origin of this
feature is the subject of a future paper.
W. E. B. has been supported by STScI under contract
NAS 5-26555 for this work. The authors would like to
thank M. Hamabe for making the PANBG brightness proles available to us for this project. We also wish to thank
the anonymous referee for some useful suggestions. Part of
the data analysis for this paper used STSDAS, which was
developed by the Space Telescope Science Institute under
US government contract NAS 5-26555.

APPENDIX
EFFECTS OF SEEING ON THE FITTING RESULTS
The tting procedure described in 3.1 makes no allowance for the eects of seeing other than to start the tting at a radius
larger than 3A to avoid the most aected portion of the brightness prole. We describe in this appendix a set of experiments
that were used to derive estimates of the correction factors for the tted parameters to measure more accurately the true
parameters of the bulge light distributions.
There are four factors in the prole acquisition process that can inuence the tting results, all of which occur at distinct
stages of the process and which can be assumed to be separable. Seeing smears the galaxy light as it travels through the
atmosphere, while Poisson noise occurs during the photographic exposure physics and chemistry. Pixelation broadens sharp
features during the digitalization of the plate, as well as adding some additional Poisson noise, and smearing by the aperture
photometry happens as a result of the prole acquisition from the digitized data. Our experiments were designed to
investigate only those processes that broaden sharp features of the galaxy light distribution : seeing, pixelation, and aperture
photometry.
The PANBG provides seeing estimates for all plates in their Table 4.1 ; we assume that seeing in this case refers to the
FWHM of stellar proles. The tabulated seeing in the PANBG ranges from 1A to 7A and is included in our Table 1.

No. 4, 1998

BULGE-DISK DECOMPOSITION

1639

We attempted to replicate the prole denition process as accurately as possible utilizing computer-generated images.
These images consist of a perfectly circular de Vaucouleurs light distribution with input values of I \ 1000 in arbitrary
intensity units and r \ 0A. 5, 1A, 2A, 4A, 8A, 16A, and 32A. Poisson noise was included in the images. Noedisk component was
e
included in these images because the most pronounced eects of seeing will be on the bulge component, owing to its very steep
slope at small radii. We generated 512 ] 512 pixel images with these characteristics, assuming a pixel scale of 3 pixels
arcsec~1 (to simulate the photographic resolution), then smeared them with Gaussians of FWHM \ 1A, 2A, 3A, 4A, 5A, 6A, and
7A to simulate the eect of the atmosphere during a long exposure. The resulting images were then block-averaged with a
3 ] 3 pixel (1A square) block to replicate the plate scanning utilized for most of the galaxies in the PANBG. Finally, the
brightness proles from the simulated images were extracted using a set of variable-diameter circular apertures along a single
radius of the galaxies. These proles were then tted in the same fashion as the real proles in order to estimate the
functional parameters, using a xed tting range of 3A48A. A second set of images was generated using the same input
parameters, but these were not smeared by the Gaussians ; these images were used to determine the correction factors due to
the pixelation and prole acquisition aperture photometry prior to estimating the eects of smearing with a Gaussian.
The results of this procedure are presented in Table 9, which includes for each simulation the seeing size in arcseconds, the
input I and r , the tted I and r , and the ratio of each tted parameter to its input value, all based on the fully degraded
e
e
e
e
brightness proles. The I are all in arbitrary intensity units, and the r are all in simulated arcseconds. Figure 6 shows the
e
e of the ratio of the tted to the input r ; the two
relationship between the ratio of the tted to the input I as a function
e
e
quantities are highly anticorrelated, suggesting that the tted parameters are not truly independent of one another.
As can be seen from Table 9, there are some cases where the tted parameters are very dierent from the input values : the
worst cases, as would be expected, are where the seeing disk is large compared to the input r . For r , the percentage errors
e
range from less than 1% to almost 40% (seeing \ 7A, r \ 1A), while for I the range is from almost
0%e (seeing \ 4A, r \ 32A)
e
e
e
to more than 1000% (seeing \ 7A, r \ 1A). We have every reason to suspect that similar tting errors exist in the PANBG
ts,
e is highly desirable.
and a means of correcting these errors
The correction procedure has been separated into two stages : rst correct the tted parameters for the combined eects of
pixelation and aperture photometry, then correct these modied parameters for the eects of the Gaussian smearing. The
pixelation/aperture photometry correction for the eective radius, based on the simulations without Gaussian smearing, has
been found to be well represented by a power law of the form
log (r@ /r ) \ 0.005 log (b/r ) [ 0.018 ,
e e
e

(4)

or
r@ \ 0.959r (b/r )0.005 ,
(5)
e
e e
where r is the tted eective radius, b is the digitization aperture size (1A), and r@ is the eective radius corrected for the eects
e
e
of pixelation
and aperture photometry ; the units of all quantities are in arcseconds.
This function has a correlation coefficient
of 0.989, and a maximum percentage error of 0.2% within the parameter space studied.
The pixelation/aperture correction function for the eective intensity is similarly found to be
I@ \ [0.068 log (b/r ) ] 1.246]I ,
(6)
e
e
e
where I is the tted eective intensity, r is the tted eective radius, b is the digitization aperture size, and I@ is the eective
intensitye corrected for pixelation and eaperture photometry. This function has a correlation coefficient eof 0.995 and a
maximum percentage error of 0.6% within the parameter space studied. Figure 7 shows the ts of the pixelation/aperture
photometry correction functions for both r and I .
e
e

FIG. 6.Relationship between I tting accuracy and r tting accuracy. The apparent anticorrelation suggests some dependence of the tting parameters
e
e
on each other.

TABLE 9
MODEL PROFILE FITTING RESULTS (INPUT I \ 1000)
e
Input r
e
(arcsec)

Fitted r
e
(arcsec)

0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.51
1.03
2.08
4.18
8.39
16.85
33.81

Fitted r /Input r
e
e
Seeing \ 1A
1.02
1.03
1.04
1.04
1.05
1.05
1.06

Fitted I
e

Fitted I /Input I
e
e

856.7
843.6
842.3
848.4
858.7
871.1
884.2

0.86
0.84
0.84
0.85
0.86
0.87
0.88

1139.2
1009.4
947.8
917.9
905.2
902.4
905.2

1.14
1.01
0.95
0.92
0.91
0.90
0.91

1571.1
1437.7
1223.8
1080.4
1004.0
964.7
945.2

1.57
1.44
1.22
1.08
1.00
0.96
0.95

2260.0
2071.9
1665.8
1340.2
1153.8
1053.2
998.8

2.26
2.07
1.67
1.34
1.15
1.05
1.00

3440.4
3254.8
2396.8
1700.3
1336.4
1150.6
1052.4

3.44
3.25
2.40
1.70
1.34
1.15
1.05

5767.5
5709.5
3641.8
2176.4
1534.6
1241.8
1100.2

5.77
5.71
3.64
2.18
1.53
1.24
1.10

10658.8
11727.6
5652.2
2726.0
1716.7
1308.3
1117.8

10.66
11.73
5.65
2.73
1.72
1.31
1.12

Seeing \ 2A
0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.49
0.99
2.01
4.08
8.23
16.60
33.41

0.97
0.99
1.01
1.02
1.03
1.04
1.04
Seeing \ 3A

0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.46
0.92
1.89
3.88
7.93
16.13
32.69

0.92
0.92
0.94
0.97
0.99
1.01
1.02
Seeing \ 4A

0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.43
0.85
1.75
3.64
7.55
15.54
31.79

0.86
0.85
0.87
0.91
0.94
0.97
0.99
Seeing \ 5A

0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.40
0.78
1.60
3.40
7.18
14.99
30.99

0.81
0.78
0.80
0.85
0.90
0.94
0.97
Seeing \ 6A

0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.37
0.70
1.45
3.17
6.85
14.54
30.37

0.74
0.70
0.73
0.79
0.86
0.91
0.95
Seeing \ 7A

0.5 . . . . . . .
1.0 . . . . . . .
2.0 . . . . . . .
4.0 . . . . . . .
8.0 . . . . . . .
16.0 . . . . . .
32.0 . . . . . .

0.34
0.61
1.31
2.98
6.61
14.27
30.19

0.67
0.61
0.66
0.75
0.83
0.89
0.94

BULGE-DISK DECOMPOSITION

1641

FIG. 7a

FIG. 7b

FIG. 7.Pixelation/aperture photometry correction functions. (a) The data and t for the correction function for r shown in a log-log plot. There is
e
clearly a systematic residual function, but it is of insignicant amplitude. (b) The data and t for the I correction function
; again there is an insignicant
e
systematic residual, particularly at large values of b/r .
e

The seeing correction is applied by the linear interpolation of the galaxy parameters on the model parameters for the same
value of seeing. In Table 10 we list the model data used in the interpolation ; column (1) lists the seeing in arcseconds (S),
column (2) the ratio of the seeing to the pixelation-corrected eective radius (S/r@ ), column (3) the ratio of the input eective
e ratio of the input eective radius to the
intensity to the pixelation-corrected eective intensity (I0/I@ ), and column (4) the
e
e
pixelation-corrected eective radius (r0/r@ ). The interpolation procedure is performed by computing the S/r@ ratio for a t and
e for
e the tabulated seeing-disk size to derive the fully corrected I0e and r0 from the
interpolating the I0/I@ and r0/r@ ratios
e
e
e
e
e
e
previously computed values of I@ and r@ (eqs. [5] and [6]).
e
e
We include in Table 1 the corrected bulge parameters resulting from the described correction procedure. Corrections for
objects whose tted parameters lie outside of the interpolation parameter space are not provided in Table 1.
TABLE 10
INTERPOLATION DATA FOR SEEING CORRECTION
S
(arcsec)
1.......

2.......

3.......

4.......

S/r@
e
2.03
1.011
0.504
0.251
0.126
0.063
0.031
4.274
2.101
1.038
0.515
0.256
0.128
0.064
6.777
3.396
1.662
0.811
0.399
0.197
0.097
9.606
4.879
2.391
1.153
0.558
0.272
0.134

I0/I@
e e
0.922
0.952
0.969
0.979
0.984
0.987
0.99
0.692
0.795
0.861
0.904
0.933
0.953
0.967
0.501
0.557
0.666
0.767
0.84
0.89
0.925
0.348
0.386
0.488
0.618
0.73
0.815
0.875

r0/r@
e e
1.015
1.011
1.008
1.006
1.005
1.005
1.005
1.068
1.05
1.038
1.03
1.024
1.02
1.017
1.129
1.132
1.108
1.082
1.063
1.049
1.04
1.201
1.22
1.196
1.153
1.116
1.089
1.069

S
(arcsec)
5.......

6.......

7.......

S/r@
e
12.857
6.675
3.263
1.544
0.734
0.353
0.171
16.764
8.938
4.319
1.986
0.922
0.436
0.21
21.527
11.954
5.567
2.461
1.115
0.519
0.246

I0/I@
e e
0.228
0.245
0.339
0.486
0.63
0.745
0.83
0.136
0.139
0.222
0.379
0.548
0.69
0.793
0.073
0.068
0.143
0.302
0.489
0.654
0.781

r0/r@
e e
1.286
1.335
1.305
1.235
1.174
1.129
1.096
1.397
1.49
1.44
1.324
1.23
1.164
1.119
1.538
1.708
1.591
1.407
1.274
1.186
1.125

1642

BAGGETT, BAGGETT, & ANDERSON

REFERENCES
Andredakis, Y. C., Peletier, R. F., & Balcells, M. 1995, MNRAS, 275, 874
Holmberg, E. 1958, Medd. Lund. Astr. Obs., Ser. 2, No. 136
Andredakis, Y. C., & Sanders, R. H. 1994, MNRAS, 267, 283
Hubble, E. 1930, ApJ, 71, 231
Baggett, W. E., Baggett, S. M., & Anderson, K. S. J. 1993, BAAS, 24, 1223
Kent, S. 1985, ApJS, 59, 115
Bevington, P. R. 1969, Data Reduction and Error Analysis for the Physical
King, I. R. 1966, AJ, 71, 64
Sciences (New York : McGraw-Hill)
Kodaira, K., Okamura, S., & Ichikawa, S., eds. 1990, Photometric Atlas of
Boroson, T. 1981, ApJS, 46, 177
Northern Bright Galaxies (Tokyo : Univ. Tokyo Press) (PANBG)
Burstein, D. 1979, ApJ, 234, 435
Kormendy, J. 1977, ApJ, 217, 406
Byun, Y. I., & Freeman, K. C. 1995, ApJ, 448, 563
Longo, G., & de Vaucouleurs, A. 1983, Univ. Texas Monogr. Astron.,
de Jong, R. S. 1996, A&A, 313, 45
No. 3 (Austin : Univ. Texas)
de Vaucouleurs, G. 1953, MNRAS, 113, 134
Sandage, A. 1961, The Hubble Atlas of Galaxies (Washington : Carnegie
de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J.,
Inst. Washington)
Paturel, G., & Fouque, P. 1991, Third Reference Catalogue of Bright
Sandage, A., & Tammann, A. 1981, A Revised Shapley-Ames Catalog of
Galaxies (New York : Springer) (RC3)
Bright Galaxies (Washington : Carnegie Inst. Washington) (RSA)
Freeman, K. C. 1970, ApJ, 160, 811