Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stsci.edu/~bushouse/pubs/co_ii.ps.gz Äàòà èçìåíåíèÿ: Thu Sep 6 18:20:53 2001 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 07:27:27 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï ï ð ï ð ï ð ï ð ï ð ï ð ï

Molecular gas in strongly interacting galaxies: II. Global
properties of the sample
Ming Zhu 1 , Howard A. Bushouse 2 , E. R. Seaquist 1 , Emmanuel Davoust 3
and
David T. Frayer 4
Received ; accepted
1 Dept. of Astronomy, U. of Toronto, 60 St. George St. Toronto, ON M5S\Gamma3H8, Canada
2 Space Telescope Science Institute
3 UMR 5572, Observatoire Midi­Pyr'en'ees, 14 Avenue Edouard Belin, 31400 Toulouse, France
4 Astronomy Department, California Institute of Technology, 105­24, Pasadena, CA 91125 (USA)

-- 2 --
ABSTRACT
We have collected CO data on a sample of 95 strongly interacting galaxies
(SIG) and on comparison samples of 59 weakly interacting and 69 isolated spiral
galaxies. The statistical analysis of the samples shows that the SIGs, especially
the colliding and merging systems, have higher CO luminosity (per unit optical
area or luminosity) than isolated spiral galaxies. If this excess is interpreted as
excess H 2 , then we find no significant difference in the molecular to atomic gas
mass ratio between the samples; this indicates that the excess molecular gas is
not due to conversion of HI to H 2 or to the removal of HI gas from the galaxies
by interaction. Another possible interpretation of the excess CO luminosity and
of the normal gas ratio is that the CO­to­H 2 conversion factor is lower in SIGs
than in isolated starbursting galaxies. In agreement with previous studies, we
find that the star formation rate (estimated by the far infrared luminosity) is
higher in the SIGs than in the isolated galaxies. The star formation efficiency
(measured by the ratio of far infrared luminosity to inferred molecular gas mass)
is higher than average in the mergers and colliding systems only. Our results are
in agreement with a scenario in which gravitational interaction produces intense
star formation which becomes highly efficient in the late stages of evolution.
Subject headings: galaxies: interacting---galaxies: ISM: molecules :
ISM: galaxies --- starburst

-- 3 --
1. Introduction
The influence of gravitational perturbations on spiral galaxies has drawn much attention
in recent years. Galaxy collisions and mergers can have a dramatic impact upon the
morphology and subsequent dynamical evolution of galaxies. Interactions are thought
to be the major cause of the extensive starburst phenomenon seen at high redshift (e.g.
Ivison et al. 2000; Scoville 2000). In the local universe, starbursts induced by galaxy
interactions are believed to be the major source of the tremendous energy output from the
so called ultra­luminous IRAS galaxies (ULIRGs). Indeed, virtually all the ULIRGs with
L IR ? 10 12 L fi are found to be mergers (Sanders et al. 1991). The star formation activity
indicators such as Hff and far­infrared and radio continuum emission all point to a higher
level of star formation activity in interacting galaxies (IG's) compared to isolated spiral
galaxies (ISG's) (e.g, Kennicutt et al. 1987; Xu & Sulentic 1991; Hummel et al. 1990;
Bushouse 1987, 1988).
Molecular clouds are the birth places for star formation. Previous CO surveys of
external galaxies have established a close correlation between the CO luminosity and the
total far infrared (FIR) luminosity (c.f. Young & Scoville 1991). If this correlation holds for
IG's, their FIR enhancement would suggest excess CO emission in IGs compared to ISG's.
This prediction seemed to be confirmed by some CO studies of external galaxies which
reported that the MH2 =LB ratio is enhanced in IGs compared to isolated galaxies (Solomon
1988; Young 1996; Brains, 1993; Combes et al. 1994). However, most previous studies
were based on a relatively small sample and the interacting systems were often selected
according to their IRAS flux, which biases these studies toward gas­rich galaxies. The
reason for this selection is the relatively low sensitivity of millimeter radio telescopes which
made it difficult to detect the CO emission in a large number of gas­poor galaxies within
a reasonable observing time. Nevertheless, the CO data for individual interacting systems

-- 4 --
has been accumulating in the literature over the years. Ironically, interacting systems in the
southern hemisphere are studied more systematically and more comprehensively than the
IGs in the northern sky (e.g., Combes et al. 1994; Horellou & Booth 1997).
We have conducted observations with the NRAO 12m telescope and the IRAM 30m
telescope (Zhu et al. 1999, Paper I), and by combining these data with those obtained
by other investigators, we have compiled a large CO database comprising a complete
optically­selected sample of strongly interacting galaxies (SIGs) in the northern hemisphere.
Here we focus our study on SIGs for mainly two reasons: (1) SIGs can be easily identified
with minimal error because they exhibit obvious disturbances in morphology and most of
them have close companions (except for mergers); (2) the number of SIGs is relatively small
in the local universe so we can observe and sample them completely. Our sample contains
154 IGs (including 95 SIGs) and covers different galaxy progenitor types, interaction phases,
and encounter geometries. This enables us to arrive at a statistically meaningful conclusion
on the influence of galaxy interaction on the molecular gas properties and induced star
formation activity.
This is the second paper in a series devoted to a statistical analysis of the molecular gas
properties in IGs. Our goal is to make clear whether the CO emission is enhanced in IGs
and whether the enhanced star formation activity in IGs is due to a higher abundance of
molecular gas or to a higher star formation efficiency. In Paper I we presented most of the
CO data obtained by us. In this paper, we combine the available CO, HI and IRAS data
in the literature for our sample galaxies and conduct an analysis of the CO, HI and FIR
properties, including a comparison between SIGs and ISGs.

-- 5 --
2. The sample
2.1. A complete SIG sample in the northern sky
The complete sample of SIGs was compiled from the UGC (Nilson 1973), ARP catalog
(Arp 1966) and the Catalogue of Isolated Pairs of Galaxies in the Northern Sky (CPG)
(Karachentsev 1972). Bushouse (1986) has identified all the SIGs from galaxies that
have been labeled as disturbed, distorted, or with bridges or warped disks in the UGC.
By searching the Third Reference Catalogue of Bright Galaxies (RC3) (de Vaucouleurs
1991), we obtained the optical (blue) magnitude for each galaxy in Bushouse's complete
sample. Then all systems with at least one member brighter than B T = 14:5 were chosen,
yielding 164 systems. The ARP catalog contains 338 systems, but it also includes some
elliptical galaxies, isolated peculiar galaxies, and some galaxies in the southern sky. The
CPG contains 1206 galaxies in 603 isolated pairs (including elliptical galaxy pairs) and is
complete to m pg = 15:0 (Stocke 1977). Its selection criteria exclude compact group members
and also under­sample merger types. For our purpose, only the spiral pairs were selected
from the CPG. These three sources overlap with one another and together they cover most
of the field IGs in the northern sky (DEC ? \Gamma2 ffi ). In order to select our SIG sample, we
further inspected all the optical images from the Digitized Sky Survey and classified these
systems according to the following criteria:
ffl IG1: Weakly interacting galaxies (WIG's), including all the interacting galaxies with
component separation ? 1.5 D 25 and with little or no morphology disturbance. Here
D 25 is the diameter of the larger component in the pair, using the brightness contour
B T = 25 mag arcsec \Gamma2 .
ffl IG2: Galaxies with component separation ! 1.5 D 25 showing obvious morphology
disturbances. Those with strong morphology disturbances but with a component

-- 6 --
separation up to 2 D 25 were also included.
ffl IG3: Galaxies in projected contact with their companions and with severe morphological
disturbances, i.e. colliding galaxies. We also include in this category the so called ``ring
galaxies'' which are believed to be the remnants of a head­on collision.
ffl IG4: Merging systems, with a single amorphous body but double nucleus, and a pair of
remnant tidal tails.
ffl IG5: Late­stage mergers, with one or two tidal tails but only a single nucleus.
In this paper SIGs are defined as those in classes IG2--IG4. Since our main interest is the
SIGs, detailed classifications for the SIGs are given and all the paired galaxies not belonging
to SIGs were put into the WIG group (IG1) for comparison. The very late­stage mergers
were put into another group (IG5) because galaxies of this type should have consumed most
of their gas and thus there should be no significant CO enhancement. The very late­stage
mergers can be easily confused with isolated elliptical galaxies, so only those that have
some tidal features such as tails were retained in the sample.
Finally, our sample galaxies were selected according to the following criteria:
ffl (1) must be a SIG (IG2--IG4)
ffl (2) at least one member in the interacting system is brighter than B T = 14:5
ffl (3) declination DEC? \Gamma2 ffi .
Our final sample of SIGs contains 126 galaxies in 92 systems. This is by far the most
complete sample of SIGs in the northern sky with DEC ? \Gamma2 ffi . Table 1 lists the general
properties of our sample galaxies and their IG type. It is ordered by interaction type (IG),
and, within the type, by right ascension (RA). The names are in cols. 1, 2 and 3, the
interaction type (IG) in col 4, the coordinates (RA and DEC) in cols. 5 and 6, the optical

-- 7 --
radial velocity (cz, in km s \Gamma1 ) in col. 7, the corrected blue apparent magnitude (B 0
T ) in col.
8, the apparent diameter (D 25 in arcmin.) in col. 9, the minor axis (r b , in arcmin.) in col.
10 and the morphological type in col. 11. The data are from RC3.
To evaluate the completeness of our sample, we searched the literature for any obvious
SIGs with B ! 14:5 that had not been included in our sample, and none were found.
For example, Dahari (1985) compiled 209 peculiar galaxies with B ! 14:4 from the Atlas
and Catalog of Interacting Galaxies (Vorontsov­Velyaminov 1959, 1977). He also divided
his sample galaxies into six interaction classes (IAC). All IAC3­IAC6 galaxies in Dahari's
sample were inspected, and it was found that all the IAC4 to IAC6 galaxies had already
been included in our sample. All the IAC1, IAC2 and some IAC3 galaxies were not
included because their morphology distortion is not strong. Although the decision to call an
individual system ``strongly interacting'' is a subjective one, the IGs with close components
and severe morphology disturbance are easy to identify and their chance of being missed by
the UGC, ARP or CPG catalogue is small. Hence our SIG sample should be essentially
complete for paired galaxies. Most of the IGs in compact groups were not included because
they are not included in either the UGC or the CPG.
The only source of confusion in classifying the galaxies may be in the merger types.
For some mergers it is difficult to tell whether they contain double nuclei. To distinguish
between IG4 and IG5, use was made of the work by Keel & Wu (1995) who listed the
number of nuclei and tails for a sample of disk­disk merger remnants in the local universe.
We also used near IR images available in the literature (e.g. Stanford & Bushouse 1991;
Bushouse & Stanford 1992). It is possible that some IG4 mergers have been mis­classified
as IG5 if their double nucleus has not been detected. However, this should not seriously
affect our statistical study since the number of such ambiguous cases is very limited.
In summary, our SIG sample contains most of the strongly interacting systems in

-- 8 --
the northern sky. This sample is selected strictly according to optical brightness and
morphology, and without any a priori knowledge of the gas content or FIR emission level.
Thus it is unbiased and ideal for studying the molecular gas properties of IGs.
2.2. The control sample
The control sample of isolated spiral galaxies was obtained from the FCRAO CO survey
made by Young et al. (1996), which is the most homogeneous sample for nearby spiral
galaxies. Since the IGs are generally more luminous than the ISGs in the FCRAO sample
(Young et al. 1995), only the more luminous (LB ? 2 \Theta 10 9 L fi ) spiral galaxies were selected
for the control sample. We also excluded the galaxies belonging to the Virgo cluster. The
final control sample comprises 69 galaxies. Figure 1a shows the distribution of MB for the
SIG and control samples. The average LB of the SIG sample is 2:7 \Theta 10 10 L fi , which is
about 1.8 times higher than that of the control sample. This is primarily because the SIG
sample as a whole is biased towards more distant, and hence more luminous, galaxies, since
there are relatively few SIGs nearby. Moreover, some merging systems include emission
from more than one galaxy.
Figure 1b compares the morphological types of the SIG and control samples. There are
many more Sbc and Sc galaxies in the control sample than in the SIG sample, where the
galaxies are more uniformly distributed among all types. It is not surprising that there
are many Irr and Peculiar galaxies in the SIG sample. Furthermore, about 20% of the
SIGs, especially the merging systems, have a morphology so distorted that they cannot be
classified as any specific Hubble type. These spiral galaxies, which are designated S?, are
not shown in the figure. Since the primary concern is with the distorted galaxies, we have
not attempted to compare the morphology distributions of these two samples, nor did we
compare any parameters that are sensitive to galaxy morphology.

-- 9 --
3. CO fluxes and molecular gas masses
Since our earlier results on 80 interacting galaxies were reported in Paper I (Zhu et
al. 1999), another 30 galaxies were observed with NRAO 12m in March, May, and June
1999. The observing procedures used are identical to that outlined in Paper I. Another 66
galaxies with good CO measurements were obtained from the literature. In addition, all the
available CO data for weak IGs (IG1) and late stage mergers (IG5) were collected providing
a consistent CO data set for all types of IGs suitable for the statistical study.
In order to make comparisons between the IG sample and the ISG sample of Young et
al. (1995), we derived the global CO flux and the inferred total mass of the molecular
gas MH 2
using the same method as that of Young et al. (1995) (see also Paper I). For all
the galaxies larger than twice the telescope beam, several measurements along the major
axis were made and the global flux was derived with a Gaussian or exponential brightness
distribution model which best fit the data. For galaxies only slightly larger than the
telescope beam, measurements were made at a single position only and an exponential
model was used to derive the integrated flux from the observed flux. Similar procedures
were applied to the galaxies observed by other investigators to derive the global CO fluxes.
There are a total of 92 SIGs for which the total CO flux is derived consistently.
In Paper I it was shown that the uncertainty in the global fluxes derived in this way are
usually less than 40%, and the data taken by the IRAM 30 telescope and the NRAO 12m
telescope are generally consistent with each other after correcting for the flux outside the
telescope beam. Uncertainties for the data in the literature are more difficult to estimate.
Some galaxies, especially the typical interacting galaxies such as ARP220, IC883 have been
observed by different groups. Higher priority was given to data with high S/N ratio and
with larger beam sizes compared to the source.
The total MH2 may be derived from the global CO(1­0) flux by assuming a standard

-- 10 --
Galactic CO­to­H 2 conversion factor, (i.e. the X factor = N(H 2 )=I CO ). However, there is
strong observational evidence that the X factor is significantly different from one galaxy
to another (e.g Crawford et al. 1985; Stacey et al. 1991; Solomon et al. 1997; Downes &
Solomon 1998; Wilson et al. 1995), and a radial variation of more than a factor of 10 has
been reported in our Galaxy (Sodroski et al. 1995). Recent studies of the molecular gas in
ULIRGs have shown that the MH2 =L CO ratio may be 3­5 times lower in the centers of these
galaxies than in the Galactic molecular clouds (Solomon et al. 1997; Downes & Solomon
1998). Hence the application of the standard X factor leads to significant overestimates of
the MH2 in these galaxies. There is also a physical basis for such a variation in the X factor,
since the CO flux per unit mass of molecular gas is expected to be a function of gas kinetic
temperature, density, and chemical composition (cf. Young 1991; Wilson et al. 1995). In
particular, it has been shown to increase strongly with decreasing metallicity (Israel 2000),
which is consistent with the tenfold increase outward in our Galaxy (Sodroski et al. 1995).
Accordingly, we introduce a parameter H \Lambda
2 to distinguish the H 2 content derived from
the standard conversion factor from the actual H 2 content, bearing in mind that it is in
reality a reflection of the CO flux rather than a precise measure of the mass of molecular
gas. The total H \Lambda
2 mass may be derived from the CO fluxes using the conversion factor
X = N(H \Lambda
2 )=Ico = 2:8 \Theta 10 20 cm \Gamma2 [Kkms \Gamma1 ] \Gamma1 (Bloemen et al. 1986). Kenney & Young
(1989) have shown that this value of the conversion factor leads to H \Lambda
2 masses in solar units
given by MH \Lambda
2
= 1:1 \Theta 10 4 D 2 SCO , where D is the distance in Mpc and SCO is the CO flux
in Jy km s \Gamma1 .
Tables 2 and 3 lists the interacting galaxies with CO measurements made by us (Table
2) or available in the literature (Table 3). For consistency, the global masses MH \Lambda
2
based on
CO measurements in the literature were re­determined using Young's method as described
above. We list not only the SIGs (IG2 -- IG4), but also the available data for weak IGs
(IG1) and late stage mergers (IG5), in order to provide a comparison sample of weakly

-- 11 --
interacting systems. A total of 154 IGs are listed in Tables 2 and 3, including 95 SIGs, 44
IG1 and 15 IG5 galaxies. Out of the 126 SIGs in the complete SIG sample (Table 1), 75%
of them have CO data available (including non­detections).
4. HI fluxes and IRAS flux densities
HI fluxes for the galaxies in our sample have been taken from the RC3 and the catalogue
of Huchtmeier & Richter (1989), or from Bushouse (1987) and Davis & Seaquist (1981).
Since it has been shown that a significant amount of atomic gas in IGs may have been
dragged out of the optical disk along the tidal tails (Hibbard et al. 1999, 2000), we
preferentially selected measurements of the global HI flux made using telescopes with a
larger beam size, such as the NRAO 92m telescope. We also gave most weight to the more
recent and more sensitive measurements.
The mass of HI is given by:
MHI = 2:36 \Theta 10 5 D 2 S(HI) (1)
where D is the distance in Mpc and S(HI) is the HI flux in Jy km s \Gamma1 . The global HI
fluxes and the derived atomic gas masses are listed in Table 4. The uncertainties in these
values are approximately 30­40%.
The infrared luminosities (in solar units) integrated from 1 to 500 ¯m are based on IRAS
flux densities at 60 and 100 ¯m, denoted as S 60 and S 100 , and were computed using the
formula given by Lonsdale et al. (1985):
L(IR) = 3:94 \Theta 10 5 D 2 [2:58S 60 + S 100 ]; (2)

-- 12 --
where S 60 and S 100 are In Jy.
Xu and Sulentic (1991) and Bushouse (1988) have re­processed the IRAS data and have
derived the IRAS flux densities for their sample galaxies. Their data were used when
available. Other data have been obtained from the NED database. The IRAS 60 and 100
¯m flux densities and the derived L IR for the CO sample SIGs are given in Table 4.
Since the resolution is not high enough to resolve the interacting galaxies, except for a few
widely separated pairs of type IG1, all the IRAS measurements are totals for each system,
as are most of the HI fluxes. Therefore there is only one entry in Table 4 for most paired
systems.
5. Comparison with isolated spiral galaxies
Table 5 lists the statistical properties of all the interacting and isolated galaxies in our
samples. The mean properties were determined using the KMESTM program in the package
ASURV Rev 1.2 (see LaValley et al. 1992 and references therein). This program computes
the Kaplan­Meier estimator (Kaplan & Meier 1958) of a randomly censored distribution
allowing for the non­detections in the sample. The second row for each ratio in Table 5
lists the median properties. These two statistical properties have different advantages. The
median is less sensitive than the means to a few extreme values in the sample. However,
unlike the mean, it is impossible to estimate the uncertainties in the median. Knowing
the uncertainties allows us to test the statistical significance of any difference between the
samples. Therefore we based our discussion mainly on the mean values. However both the
mean and median properties point to the same trend.
The galaxies in the SIG sample on average have a higher LB and larger physical size. In
order to compare the gas content of galaxies between the two samples, we need to eliminate

-- 13 --
the effect of the size. This can be achieved by normalizing the molecular gas mass MH \Lambda
2
by
the optical lumminosity or by the optical area of the galaxy. We adopt the same normalized
quantities as Casoli (1998), namely the MH \Lambda
2
=LB and MH \Lambda
2
=D 2
25 ratios.
5.1. The MH \Lambda
2
=LB ratio
Figure 2 shows the distributions of the MH \Lambda
2
=LB ratio for the two samples. Most of the
normal spirals have a MH \Lambda
2
=LB ratio smaller than 0.3. M82 and NGC1055 are the only two
galaxies in Young's sample with a MH \Lambda
2
=LB ratio higher than 0.31, but they are not really
isolated galaxies. M82 is gravitationally interacting with M81, and N1055 is about 130
kpc from NGC 1068. Therefore essentially no strictly isolated galaxies have been found to
possess a MH \Lambda
2
=LB ratio higher than 0.31. On the other hand, high ratios of MH \Lambda
2
=LB are
frequently seen in IGs and mergers. Merger remnants such as ARP220 have MH \Lambda
2
=LB ratios
as high as 1.1. The average MH \Lambda
2
=LB ratio of the SIG sample is 0.22, which is 1.7 times
higher than that of the ISG sample.
The high MH \Lambda
2
=LB ratios suggest that SIGs, especially the colliding and merging systems,
have more molecular gas than isolated spirals. However, Perea et al. (1997) have argued
that the relation between LB and MH \Lambda
2
is not linear, indicating that normalizing the mass by
LB does not completely remove the size effect. To evaluate the latter, we plot MH \Lambda
2
against
LB in Fig. 3. The merging systems (IG3 and IG4) are shown as solid circles and IG2
systems are shown as open circles while stars represent the isolated spirals in the control
sample. The values of logL B range from 9:6 to 11. A linear regression applied to the ISG
sample, using the ASURV V1.2 BIVAR method which takes into account the upper limits,
yields

-- 14 --
logMH \Lambda
2
= (1:30 \Sigma 0:10)logL B \Gamma (4:04 \Sigma 0:98) (3)
This fit is indicated by a dashed line in Fig. 3. This line fits the ISG and IG2 systems
reasonally well, but fails to fit the merging systems. Examining the deviation of each galaxy
from the dashed line and applying a Peto­Prentice generalized Wilcoxon test (using the
ASURV package) for each subsample, we found that the merger type (IG3 + IG4) has a
residual significantly different from zero with a confidence level of 96%. Not surprisingly,
the residual for the IG2 subsample and the ISG control sample does not significantly deviate
from zero. Hence the high MH \Lambda
2
=LB ratios in merging galaxies are significant and cannot be
accounted for by the size effect. The fact that no isolated galaxies are found in the high
MH \Lambda
2
=LB regime strongly suggests that the high MH \Lambda
2
=LB ratio is the result of strong galaxy
interactions. However, not every interacting or merging system is associated with a high
MH \Lambda
2
=LB ratio.
5.2. The MH \Lambda
2
=D 2
25
ratio
Another way to eliminate the size effect is normalization by the optical surface area of
the galaxies. The mean and median values of MH \Lambda
2
=D 2
25 for the SIG sample as well as
for the individual IG subsamples are given in Table 5. The result is similar to that from
normalization by LB . The average MH \Lambda
2
=D 2
25 ratio is about twice as high in SIGs. The
most significant differences come from the merger types. A two­sample Wilcoxon test shows
that the IG3 and IG4 subsamples have a mean MH \Lambda
2
=D 2
25 ratio significantly higher than
that of the ISGs with a confidence level of 99%, while the confidence level drops to 94% for
the IG2 subsample. The histogram in Fig. 4 shows that the MH \Lambda
2
=D 2
25 distribution for the
SIG sample is more populated at the higher end, though the dispersion is also larger in the
SIG sample.

-- 15 --
Figure 5 shows a plot of MH \Lambda
2
versus D 2
25 for the subsample of IG2 (open circles), merger
type (solid circles) and ISGs (stars). A linear regression fit to the isolated galaxy sample
(dashed line) yields
logMH \Lambda
2
= (1:81 \Sigma 0:24)logD 25 + (6:85 \Sigma 0:32) (4)
or
logMH \Lambda
2
= (0:91 \Sigma 0:12)logD 2
25 + (6:85 \Sigma 0:32) (5)
This relation is similar to that derived by Casoli (1998) from a sample of more than 500
isolated galaxies. From this we can see that the relation between MH \Lambda
2
and D 25 is consistent
with a linear one and that the ratio MH \Lambda
2
=D 2
25 should completely remove the size effect.
Therefore, the high MH \Lambda
2
=D 2
25 ratio for IG3 and IG4 is obviously not due to their larger
size.
On Fig. 5 we can also see that the dashed line which best fits the ISG sample fails to
fit the SIG sample. Statistical tests show that the residuals are significantly different from
zero value at the confidence level of 99% and 95% for the merger type and IG2 subsamples,
respectively.
Due to the morphology distortion in virtually all SIGs, the size (D 25 ) may have a large
uncertainty. However, one is more likely to overestimate D 25 of a disturbed disk, and thus
underestimate the MH \Lambda
2
=D 2
25 ratio. Even if the errors on D 25 were random, this should
not result in a higher average MH \Lambda
2
=D 2
25 ratio for SIGs. Therefore the high MH \Lambda
2
=D 2
25 ratio
in some SIGs must be genuine and indicates an CO enhancement as a result of galaxy
interaction.

-- 16 --
Casoli (1998) showed that the MH \Lambda
2
=D 2
25 ratio depends on the Hubble type, with the Sa,
Sb and Sbc types containing ¸ 6 times more molecular gas than late­type spirals and Irr
galaxies. If a galaxy sample includes a majority of Sa­Sbc galaxies, it would have a higher
than normal average MH \Lambda
2
=D 2
25 ratio. However this is not the case for our SIG sample. In
Figure 1b we have shown that the SIG sample is not overpopulated with Sa­Sbc galaxies
compared to the control sample.
In summary, both the MH \Lambda
2
=LB and MH \Lambda
2
=D 2
25 ratios suggest either an enhancement in
molecular gas content or in the luminous efficiency of CO for some SIGs, especially for
IG3 and IG4. The choice between these two options will depend on careful modeling of a
number of individual systems.
6. Molecular versus atomic gas
Assuming MH \Lambda
2
represents the actual molecular mass MH 2
, we may compare the total
masses of molecular and atomic gases in our samples, in order to investigate the possibility
that some systems are enriched in molecular gas, either by phase conversion of HI to H 2 , or
by removal of HI from the system by tidal forces.
Figure 6 shows the distributions of the MH \Lambda
2
=MHI ratio for the SIG and ISG samples.
Both samples peak at MH \Lambda
2
=MHI ¸ 1 and there is no significant difference between them.
The average MH \Lambda
2
=MHI ratio for SIGs is only 1.3 times higher than that of the ISG sample
(Table 5). The merger type subsample has a higher MH \Lambda
2
=MHI mean value, but this is not
statistically significant, as the confidence level is only 83%.
Table 5 shows that the MHI =LB ratios of the SIGs and ISG samples are similar. This
means that the high MH \Lambda
2
=MHI ratios in some systems corresponds to an enhancement of
MH \Lambda
2
rather than to a depletion of MHI . This is a different conclusion from that derived for

-- 17 --
the IRAS­selected IG sample (e.g. Martin et al. 1991), which shows a depletion of HI gas for
the ULIRGs. This result also indicates that the single­dish HI data do not systematically
underestimate the HI gas even for the late­stage mergers (IG3 and IG4) which could have a
significant amount of HI gas ejected from the galaxy disks by tidal forces (e.g. Hibbard et
al. 1999).
Figure 7 shows plots of MH \Lambda
2
=LB versus MHI =LB for t