Submillimeter Interferometry of the Luminous Infrared Galaxy NGC 4418: A Hidden Hot Nucleus with an Inflow and an Outflow

1.5 MB·English

by  Kazushi Sakamoto

Accepted for publicationin ApJ
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
SUBMILLIMETER INTERFEROMETRY OF THE LUMINOUS INFRARED GALAXY NGC 4418:
A HIDDEN HOT NUCLEUS WITH AN INFLOW AND AN OUTFLOW
Kazushi Sakamoto1, Susanne Aalto2, Francesco Costagliola2, Sergio Mart´ın3,
Youichi Ohyama1, Martina C. Wiedner4, and David J. Wilner5
Accepted for publication inApJ
ABSTRACT
3 We have observed the nucleus of the nearby luminous infrared galaxy NGC 4418 with subarcsec
1 resolution at 860 and 450 µm for the first time to characterize its hidden power source. A ∼20 pc
0 (0′.′1) hot dusty core was found inside a 100 pc scale concentration of molecular gas at the galactic
2 center. The 860 µm continuum core has a deconvolved (peak) brightness temperature of 120–210 K.
The CO(3–2)peak brightness temperature there is as high as 90 K at 50 pc resolution. The core has
n
a bolometric luminosity of about 1011 L , which accounts for most of the galaxy luminosity. It is
a ⊙
J Comptonthick (N &1025 cm−2)andhas a highluminosity-to-massratio∼500L M −1 as wellas
H ⊙ ⊙
9 a high luminosity surface density 108.5±0.5 L⊙ pc−2. These parameters are consistent with an AGN
to be the main luminosity source (with an Eddington ratio about 0.3) while they can be also due to

] a young starburst near its maximum L/M. We also found an optical color (reddening) feature that
O
we attribute to an outflow cone emanating from the nucleus. The hidden hot nucleus thus shows
C evidence of both an inflow, previously seen with absorption lines, and the new outflow reported here
. in a different direction. The nucleus must be rapidly evolving with these gas flows.
h
Subject headings: galaxies: active — galaxies: evolution — galaxies: individual (NGC 4418) —
p
galaxies: ISM
-
o
r
t 1. INTRODUCTION the source inferred from the absorption depth. It has
s
a There are galaxies hiding a compact luminous source been argued that dust is unlikely to cover an extended
[ starburst so completely as to cause the deep absorption
in their nuclei and NGC 4418 is their local prototype.

1 The extremely extinguished power source in this early- features (Roche et al. 1986; Dudley & Wynn-Williams
1997; Spoon et al. 2001). On the other hand, the ar-
v type disk galaxy was first recognized by Roche et al.
guments for very young starburst are based on the low
8 (1986)whosingledoutthe galaxyforits infraredbright-
radio to far infrared ratio and the detection of a warm
7 ness. They found very deep silicate absorption at 9.7
molecular gas absorber toward the nucleus. The former
8 µmtowardthe infrared-brightnucleus anddeducedthat
1 the source of the infrared luminosity (∼1011L ) is com- ratio is lower than in radio-quiet AGNs and is consis-
⊙
tentwithastarburstinthepre-supernovaage(<5Myr)
. pletelyobscuredbyadustshroudofA ≫50mag. They
1 V (Roussel et al.2003). Theabsorbingmoleculesappearto
further inferredfromthe weaknessofline featuresin the
0 beintheregionwhereX-rayswouldhavedestroyedthem
infraredandopticalspectrathattheluminositysourceis
3 if the heating is due to an AGN (Lahuis et al. 2007).
an accreting massive black hole (i.e., active galactic nu-
1
The true nature of the luminous nucleus is important
: cleus or AGN), although they also mentioned a nascent
v starburst embedded within a dusty cocoon as a possible because this type of heavily extinguished nuclei could
Xi alternative. double the number of local AGN (Maiolino et al. 2003)
or it could represent a young phase of unusually in-
Subsequent observations to be reviewed in §1.1 have
r tense and compact starburst (≪ 100 pc). Luminous
a supportedthepresenceofadeeplyburiedcompactlumi-
hidden nuclei of this kind are also important because
nosity source, although its nature is still debated. NGC
they may be in a short transition phase to unobscured
4418 is the most absorption-dominated galaxy in the
AGNs, i.e., in the process of removing the obscuring
mid-infrared diagnostic diagram of Spoon et al. (2007).
gas(e.g.,throughanAGN-drivenoutflow,Sanders et al.
Its location on the diagonal branch in the diagram sug-
1988; Fischer et al. 2010). Such feedback is regarded as
geststhattheluminositysourceisalmostcompletelycov-
a key process in galaxy evolution and hence requires de-
ered by opaque dust. Usual spectroscopic signs of AGN
tailed study (Hopkins et al. 2006; Weiner et al. 2009).
canbecompletelyextinguishedbythisheavyabsorption.
Anewapproachtocharacterizeanddiagnosethecom-
Hence, the main piece of evidence for a hidden AGN,
pact, luminous, and extinguished nuclei of this kind is
among other indirect ones, has been the compactness of
through submillimeter-wave observations at high spatial
1Academia Sinica, Institute of Astronomy and Astrophysics, resolution. Sub-mmobservationsdirectlyprobethether-
Taiwan mal dust emission that constitutes most of the bolomet-
2DepartmentofEarthandSpaceSciences,ChalmersUniver- ric luminosity of such a nucleus. Unlike optical and in-
sityofTechnology, OnsalaSpaceObservatory,Onsala,Sweden
3EuropeanSouthernObservatory, Santiago,Chile frared observations, one can look deep into the obscur-
4LERMA&UMR8112duCNRS,ObservatoiredeParis,61 ingmaterialsincedustopacitydecreasesatlongerwave-
Av. del’Observatoire,75014,Paris,France lengths. Radio interferometry is available in the sub-
5Harvard-Smithsonian Center for Astrophysics, Cambridge, mmtoachievesub-arcsecresolutionthatiscurrentlyun-
MA,U.S.A.
2 SAKAMOTO et al.
available in far-IR where the radiation of such a nucleus 1993). (2)NGC4418haswarmmid-IRandfar-IRcolors
peaks. It is possible to estimate the bolometric lumi- S(25µm)/S(60µm) = 0.220 and S(60µm)/S(100µm) =
nosity and luminosity density of the nucleus using the 1.37 (Sanders et al. 2003). The former puts the galaxy
Stefan-Boltzmann law by measuring the source size and in the ‘warm’ population that often host active nuclei
brightnesstemperatureatafrequencywherethenucleus (de Grijp et al. 1985; Sanders et al. 1988). The latter
has a photosphere. One can also probe the nucleus by also suggests warm dust (color temperature of ∼50 K)
measuringitsmassfromrotationofcircumnuclearmolec- indicativeofreprocesseddustemission. (3)Thedeepab-
ulargas,whichcanbeobservedusingsub-mmmolecular sorptionhasbeenconfirmedbyfurtherspectroscopyand
lines. The luminosity-to-mass ratio, L /M , so ob- modeling; the estimated opacity is τ ≈ 7 or A ∼
bol dyn 9.7µm V
tained will further constrain the nature of the nucleus 100 mag (Roche et al. 1986; Dudley & Wynn-Williams
sinceanAGNcanhaveahigherL/M thanstars. Molec- 1997; Spoon et al. 2001; Siebenmorgen et al. 2008). A
ular lines canprovideadditional informationontemper- radial gradient of temperature in the nucleus may
ature, density, chemical composition of the interstellar also contribute to the deep 10 µm depression in the
medium (ISM), and on the radial motion of the gas af- same way Kwan & Scoville (1976) modeled for proto-
fectingtheevolutionofthedustyshroud. Wehadapplied stars (Dudley & Wynn-Williams 1997). It is also sug-
this approach to the ultraluminous infrared galaxy Arp gested from a shallow SED slope in the FIR that the
220, which shares some of the above mentioned charac- dust has a high enough column density to be optically
teristics ofNGC 4418including the deepsilicate absorp- thick even at or beyond 100 µm (Roche & Chandler
tion. We found that the brighter one of the twin nuclei 1993; Lisenfeld et al. 2000). Likely related to this is
of Arp 220 has Lbol≥2×1011 L⊙ and Lbol/Mdyn& 400 an unusually low ratio of the [C II] 158 µm line
L /M inthecentral80pc(Sakamoto et al.2008,here- to far-IR luminosity in NGC 4418 (Malhotra et al.
⊙ ⊙
after S08) and that each nucleus has a molecular out- 1997; Graci´a-Carpio et al. 2011). (4) The compact-
flowcausingmolecularP-Cygniprofiles(Sakamoto et al. ness of the infrared nucleus has been directly con-
2009). firmed with mid-IR imaging at subarcsecond reso-
In this paper, we apply the same approach to NGC lutions (Wynn-Williams & Becklin 1993; Evans et al.
4418toquantitativelyconstrainthenatureofthedeeply 2003; Siebenmorgen et al. 2008). According to the Keck
extinguished nucleus through sub-mm high-resolution imagingbyEvans et al.(2003),the10µmsizeofthenu-
observations. We made the first sub-arcsecond resolu- cleus is 0′.′3 and the 25 µm size is less than 0′.′6, which is
tionimagingofthegalaxyinthesubmillimeterband,ob- the diffraction limit of the 10 m telescope. (5) The nu-
serving860µmcontinuumandmolecularlines including cleus is also compact (∼0′.′1–0′.′4) and has a high bright-
CO(3–2), HCN(4–3), and HCO+(4–3) and also imaging ness temperature (& 105 K) at ∼2 GHz (Condon et al.
the nucleus with 450 µm continuum. We also analyzed 1990; Kewley et al. 2000; Baan & Kl¨ockner 2006). Such
optical multi-color images to probe circumnuclear dust a radio nucleus is either an active nucleus or a com-
distribution for signs of disturbance and feedback. The pact starburst or their hybrid (Smith et al. 1998). (6)
obtainedparametersofthenucleusaresimilartotheones Chandra X-ray observations “may imply the presence
in Arp 220. They are consistent with any of the follow- of a Compton-thick AGN” although the identification
ing being energeticallydominant: an AGN, a youngand is “somewhat tentative” because of the limited pho-
compactstarburst,andanAGN-compactstarburstcom- ton statistics (Maiolino et al. 2003). (7) NGC 4418
bination. Our results support the previous studies using has far-IR excess compared to its radio strength, at-
an independent method and indicate a possible path to tributedeithertoadust-enshroudedAGNortoayoung,
break the degeneracy about the luminosity source with compact starburst (Kawara et al. 1990; Yun et al. 2001;
future observations of higher quality. We also found a Roussel et al. 2003). (8) Molecular gas concentration to
sign of outflow from the nucleus in the optical data. In the central ≤ 5′′ has been observedwith interferometers
the following, we briefly review properties of NGC 4418 (Imanishi et al. 2004; Dale et al. 2005; Sakamoto et al.
and previous studies on the hidden nucleus in §1.1. We 2010;C12). The massofthemoleculargaswillbe about
describe our SMA observations in §2 and presentthe re- 1×109M iftheaverageconversionfactorinourGalaxy
⊙
sults in §3, where we also report our examination of op- (Hunter et al.1997)appliestotheobservedCO(1–0)flux
tical images of the galaxy. The results are discussed in of100±50Jykm s−1 (Kawara et al.1990;Sanders et al.
§4 and the conclusions are summarized in §5. Our com-
1991;Young et al.1995;Dale et al.2005;Albrecht et al.
panionpaperreportshigh-resolution1mmandcm-wave
2007). NGC 4418 has a high infrared luminosity-to-
observationsof the same galaxy (Costagliola et al. 2012, molecular gas mass ratio of L /M ∼ 100 L /M
IR mol ⊙ ⊙
hereafter C12).
for the above M (Kawara et al. 1990; Sanders et al.
mol
1991). High-excitation molecular lines with excita-
1.1. NGC 4418
tion temperatures above 100 K have been detected
StudiesonNGC4418anditsnucleussinceRoche et al.
toward the nucleus (Aalto et al. 2007; Lahuis et al.
(1986) are summarized below. (To guide readers, an
2007; Costagliola & Aalto 2010; Sakamoto et al. 2010;
optical image and key parameters of the galaxy are in
Gonz´alez-Alfonso et al. 2012a; C12). The last four pub-
Fig. 1 and Table 1, respectively. We adopt the galaxy
lications report vibrationally excited molecules and the
distance of 34 Mpc at which 1′′ is 165 pc.) (1) A
last three report lines involving energy levels as high
high infrared-to-optical luminosity ratio LIR/LB ≈ 20 as ∼1000 K. (9) Redshifted absorption lines have been
(Armus et al. 1987; Young et al. 1989) and the shape of
found toward the nucleus by Gonz´alez-Alfonso et al.
thespectralenergydistribution(SED)suggestthatmost
(2012a) in the mid-infrared O I and OH lines and by
of the luminosity of the galaxy is absorbed and rera-
Costagliola et al.(2012)intheHI21cmline. Theseare
diated by dust as thermal emission (Roche & Chandler
ApJ in press 3
attributed to a gas inflow. was significant at our observing frequency even though
theweatherconditionwasexcellentforMaunaKea. The
2. SUBMILLIMETERARRAYOBSERVATIONS mean zenith opacity was 0.038 at 225 GHz and 0.89 at
2.1. 350 GHz (860 µm) Observations 350 µm during our observations according to the mon-
itoring at the adjacent Caltech Submillimeter Observa-
We observed NGC 4418 in 2009, 2010, and 2012
tory. The opacity should be about the same at 450 and
at around 350 GHz (860 µm) using the Submillime-
350 µm. The median double sideband (DSB) system
ter Array (SMA)6. We used the subcompact (SC), ex-
temperature was about 1200 K toward NGC 4418. We
tended (EXT), and very extended (VEX) array configu-
used3C273forbothpassbandandgaincalibration. Both
rations. Observational parameters are in Table 2. The
350 GHz and 660 GHz data were used for the gain cal-
data fromthe SC configurationwerealreadyreportedin
ibration. We first applied the 350 GHz gain solution
Sakamoto et al. (2010, hereafter S10) and are combined
with its phase multiplied by the ratio of the observing
here withthe new data. The overallcoverageof baseline
frequencies, 1.9, and then derived another gain solution
lengthsis6–509mor7–590kλat350GHzforthegalaxy.
usingthe660GHzdataandappliedittoimprovethecal-
We simultaneously observed CO(3–2) and HCO+(4–3)
ibration. No strong line is in our passband and the data
lines in 2009 and CO(3–2) and HCN(4–3) in 2010 us-
are treated as continuum. Since Titan is too heavily re-
ing 2 GHz of bandwidth. We used in the EXT and SC
solved, flux calibration at 660 GHz is based on the 350
configurations 4 GHz bandwidth to simultaneously ob-
GHz flux density of 3C273 scaled to 660 GHz using the
serve CO(3–2), HCO+(4–3), and HCN(4–3). The SMA
spectral index of α = −1.0 (S10). We estimate our flux
primary beam covers about 34′′ (5.6 kpc) in full width
measurements at 660 GHz to be accurate only to 36%
athalfmaximum(FWHM)atourobservingfrequencies.
(for σ(α) = 0.3 and 30% error for other calibrations).
The SMA spectrocorrelator was configured to the reso-
lution of 3.25 MHz (about 3 km s−1) and recorded data
3. OBSERVATIONALRESULTSANDANALYSIS
from both upper and lower sidebands (USB and LSB).
Weused3C273forourgain,passband,andpointingcal- 3.1. Spectra and Line-Continuum Separation
ibrations. We also observed the quasar J1222+042 and Figures 2 and 3 show spectra7 made from the VEX
Titan in our 2010 VEX observations for further calibra- dataaloneandfromthe EXTandSC data,respectively.
tion and its verification. Flux calibration was made us- The SC+EXT data are two times more sensitive than
ing Titan and Ganymede in 2009, and Titan in 2010 the SC data alone reported in S10.
and 2012. The angular distances from our gain cali- Two lines among about ten in the spectra are
brator 3C273 to NGC 4418, J1222+042, and Titan in new detections in this galaxy. Both are from vibra-
our 2010 observations, were only 3.0, 2.7, and 4.0 de- tionally excited HC N. Namely, HC N(v =11f,J=39–
3 3 7
grees,respectively. This mitigatedcalibrationerrorsdue 38) is at f =356.072 GHz and has its upper energy
rest
todirection-dependentphasede-correlation,atmospheric level at E = 663 K. HC N(v =11e,J=39–38) is at
u 3 6
transmission, and pointing offsets. f =355.278 GHz and has E = 1059 K. The dou-
rest u
We reduced our data using MIR for calibration blet counterpart of the former line (v =11e,J=39–38)
7
(Scoville et al. 1993), MIRIAD for imaging (Sault et al. was already identified at f =355.566 GHz in S10.
rest
1995), and AIPS (Bridle & Greisen 1994) and stand- Lower J transitions of the latter line were detected
aloneprogramsforfurther dataanalyses. Thedatawere in Costagliola & Aalto (2010) and C12. Its doublet
firstimagedat 30km s−1 resolutionfor the entire band- counterpart (v =11f,J=39–38) at f =355.557 GHz
6 rest
widths to identify lines. Then the ‘line-free’ channels
is blended with the HC N(v =11e,J=39–38) mentioned
wereidentified and integratedto createcontinuum data. 3 7
above. Although unidentified, there may be more lines,
That continuum is then subtracted from the original to
e.g., at f ∼ 343.4 and 357.5 GHz; species having a
makecontinuum-subtractedvisibilities. Thespectraand rest
line at the former frequency include H CS andHOCO+.
the identified lines are presented in §3.1. The line data 2
The 350 GHz continuum in each sideband was
werefurtheranalyzedafterbinningto10–30km s−1 res-
made by averaging the following channels. For the
olutions. We did self-calibration of phase using the con-
VEX data we used channels outside of 1900–2300
tinuum data when making images from the VEX data,
km s−1forCO(3–2),HCN(4–3),H13CN(4–3),HC15N(4–
although we did not do that when measuring the source
3), HCN(v =11f,J=4–3), HCO+(4–3), CS(7–6), and
position or fitting the visibilities to determine the size 2
HC N(39–38)and(38–37). FortheEXTandSCdatawe
and flux density of the continuum source. Table 3 sum- 3
excludedchannelshavingvelocitiesof1850–2350km s−1
marizes the properties of the images and spectral data
for the same lines8 plus HC N(v =11f,J=39–38) and
cubes used in this paper. 3 7
(v =11e,J=39–38), channels between CO(3–2) and its
7
2.2. 660 GHz (450 µm) Observations nearest band edge, and the same for HCN(4–3). The
660 GHz continuum was averaged over the full band-
Wealsoobservedthegalaxyat660GHz(450µm)dur-
width and both sidebands.
ing our VEX 350 GHz observations in 2009and 2010by
allocating 2 GHz of correlator bandwidth to our high
7 Throughoutthispaper,velocitiesarewithrespecttothelocal
frequency receivers. Only the 2010 run resulted in use-
standardofrest(LSR)andaredefinedwiththeradioconvention,
able data from five antennas. Atmospheric absorption i.e.,vradio=c(1−(fobs/frest))wherefdenotesfrequencyandcthe
speedoflight. Theconversiontovelocity inthe optical definition
6 TheSubmillimeterArrayisajointprojectbetweentheSmith- isvopt=cz=vradio+15kms−1 forNGC4418.
sonian Astrophysical Observatory and the Academia Sinica Insti- 8 The velocity range used here is the same as the one used
tuteofAstronomyandAstrophysics,andisfundedbytheSmith- to analyze the SC data in S10 although the note in their Table 1
sonianInstitutionandtheAcademiaSinica. describeditotherwise.
4 SAKAMOTO et al.
3.2. Concentration of Gas and Dust in the Nucleus Figure 9 (a)–(c) show fits to the continuum visibilities
of NGC 4418,a test point source J1222+042, and Titan
Submillimeter emission is detected only from the cen-
tral kpc (∼6′′) with different spatial extents in various whose subarcsecond size is known. The fit results are
listed in Table 5. For NGC 4418, data from the three
emission components. We show this below using maps,
arrayconfigurations are combined. This cannot be done
aperture photometry, and visibility fitting.
for the calibrators because their flux densities and the
apparent diameter of Titan change with time. Since the
3.2.1. Maps
NGC 4418 data were taken in three different nights the
Figure 4shows CO(3–2)channelmaps to displayfaint three data sets should have independent flux calibration
extended emission at 0′.′6 resolution. The data cube errors, which are typically about 10–20% for the SMA
was made by combining all our data. We detected here 350GHz observations. However,wedidnotaddthe flux
797±80Jykm s−1 or84±18%ofthesingle-dishCO(3– scaling error in the plot and in our fitting because do-
2) flux observed in the central 14′′ of the galaxy by ing so would make the reduced χ2 well below unity. It
Yao et al. (2003). Despite the high recovery rate, the is most likely that we achieved better flux calibration
CO(3–2)emission was detected only in the central6′′ (1 than usual, presumably because we had accurate point-
kpc)ofthegalaxy,showingthattheextentoftheCO(3– ing from repeated pointing on the bright quasar 3C273
2)emissionisintrinsicallysmallerthanour34′′ (5.6kpc) nexttoNGC4418,ourprimaryfluxcalibratorTitanwas
field of view. The CO emission is not only compact but alsocloseto 3C273andNGC 4418,andatmospheric ab-
also strongly peaked toward the nucleus within the cen- sorption was low (225 GHz zenith opacity ≤0.05 for all
tral kpc. In this data set more than half (58 %) of the thethreenights). Thesizeof850µmcontinuumemission
CO(3–2) flux in the central kpc is in the central 165 pc of NGC 4418 obtained from our data is 0′.′10 (FWHM).
(1′′, FWHM). We will use this for the source modeling in §3.3.
Figures5and6respectivelyareCO(3–2)andHCN(4– We verified our visibility fitting using the quasar
3) channel maps at ∼0′.′3 (50 pc) resolutions showing J1222+042 and Titan. They were gain calibrated us-
that the line emission is strongly peaked at the nucleus ing3C273inthe same wayasNGC 4418wascalibrated.
atsubarcsecond(.100pc)scales. Thesedatacubeswere Our fit of J1222+042 using a Gaussian shows that blur-
madefromtheVEXdataalone. ThisCO(3–2)datacube ring of the point source was less than 50 milli arcsec
contains about 50% of the CO flux detected in our 0′.′6 (mas)in ourobservations. The diameter of Titanat860
resolution data. µmisknowntobe5230kmor0′.′8425forthe distanceof
Our continuum and line moment maps are shown in 8.5588 AU. This size includes the 40 km altitude of the
Figs. 7and8. The spatialresolutionsareabout0′.′6and tropopause,aroundwhich is the sub-mm photosphere of
0′.′3 for 350 GHz data and 0′.′2 for 660 GHz continuum. Titan. The distance at the time ofour 2010VEX obser-
In particular, Fig. 8 (k) and (l) show our detection of a vations is from JPL-Horizons. Fig. 9 (c) shows that our
compact continuum source at the galactic center. The data match the expected visibility curve very well with
continuum position is listed in Table 1 and agrees to only flux scaling. When the data were fit for both the
∼0′.′3withthe opticalandradiopositionsofthenucleus; source size and amplitude scaling, we obtained the size
see the caption of Table 1 for the optical position and ofthesub-arcsecondsourcewithanaccuracybetterthan
Condon et al. (1990) for the radio position. 3% or to 20 mas.
The flux density of Titan sets the absolute amplitude
3.2.2. Aperture Photometry scaling of all the visibility fitting. We adopt the con-
tinuum brightness temperature of 77 K as the Planck
The degree of central concentration, or compactness,
temperaturethatgivesthetotalfluxdensityofTitanfor
is compared among various lines and continuum at 350
the solid-surface diameter (5150 km) of the satellite (M.
GHzinTable4usingtheratioofflux(orfluxdensity)be-
tweenthecentral0′.′5(80pc)and6′′ (1kpc). Theformer Gurwell,privatecommunication). OurTitandatadonot
containCO(3–2)andHCN(4–3) lines, both of which are
photometry is made using our data from the VEX con-
knowntobebright. Still,ourimperfectknowledgeofthe
figurationandthelatterdataarefromS10thatusedthe
atmosphereofTitanmayintroduceacommonamplitude
SCconfiguration. Theratiosforvariousemissioncompo-
errortoourdataeventhoughtherelativecalibrationus-
nents indicate their relative compactness. The relatively
ing Titanis precise andconsistentamong the three data
extended nature of CO(3–2)emission is evident from its
sets. We therefore assign 5% to the overall flux scal-
low ratio. Only one third of the CO(3–2) emission is
from the central 0′.′5. Lines other than CO have higher ing uncertainty of our NGC 4418 fit, even though the
formal error in the fitting was about 1%. In addition
flux ratios, and hence are more concentrated toward the
to this, there is likely contamination of line emission to
nucleus. The ratio for 860 µm continuum, 0.95±0.07,
the ‘line-free’ channels used for the USB continuum of
is the highest among the ratios with S/N > 3 in Table
the galaxy. We observed a sign of that in our SC con-
4. Thus 860 µm continuum emission is more compact
figuration data in the form of an uncomfortably large
than both the CO and non-CO high excitation lines. At
spectral index α = 6.25±0.84 (S ∝ να) between the
the same time, the continuumemissioninthe VEXdata ν
is slightly moreextendedthan the 0′.′3synthesizedbeam twosidebands(S10). Asimilarsignwasseeninour2010
VEXdatatoo. Since weusedthe USB continuuminthe
as seen in the last two panels of Fig. 5. The detected
multi-configurationfitting as LSB data were noisier, the
450 µm continuum also looks compact but it has only a
line contamination is likely in the fitted amplitude. Its
minor fraction of single-dish measurement.
effect on the luminosity estimate is addressed in §3.3.3.
3.2.3. Visibility Fitting : 860 µm Continuum 3.2.4. Visibility Fitting : Lines
ApJ in press 5
Fig. 9 also shows visibility amplitudes of line emission case without mutual shielding is the flux-weightedmean
against baseline lengths. Line visibilities were averaged ofthespectralindexes,α=PS α /PS ,wheretheith
i i i
over1900–2300km s−1 aftercontinuumsubtraction,and componenthasthefluxdensityS andspectralindexα .
i i
then binned according to the baseline length. It is evi- For α =2 and α =4, the observed α corresponds to a
1 2
dent in the figure that the line visibility amplitudes de- fraction of the optically thick core of 73±9 %. Thus as
cline faster than the continuum amplitudes. In other longasweassumeβ ≈2,weneedanopticallythickcore
words, line emission is more extended than continuum. (τ & 1) that dominates the spectral energy distribution
Comparison of CO, HCN, and HCO+ data tells us that at λ ∼ 1 mm. In addition to the two models, the high
the CO emission is more extended than the latter two, opacity is also supported by the high brightness tem-
confirming our observation using the aperture photome- perature that sets a lower limit of about 0.1 to the 860
try. Moreover,theCOvisibilityplotshowsthatthedata µm opacity from the constraint that opacity-corrected
cannot be fit with a single Gaussian in the u–v domain. brightness temperature cannot exceed dust sublimation
Hence the spatial distribution of the CO(3–2) emission temperature.
is not a Gaussian. The line emission distribution has at There are other factors that depend on frequency and
least a compact subarcsecond peak toward the nucleus potentially affect the spectral index. They include the
(Fig. 8) and an extended component with the total ex- core size defined with the τ = 1 surface, the dust tem-
tent of ∼5′′ (Fig. 7). The extended component is non- perature at the photosphere, and absorption of the core
axisymmetric around the nucleus (Fig. 7). emission by dust around it. The last one is addressed in
§3.3.4.
3.3. Parameters of the Continuum Core
3.3.2. Gas Column Density, Mean Density, and Mass
3.3.1. Spectral Energy Distribution, Continuum Opacity
The dust continuum opacity τ ≈ 1 of the core
860µm
The continuum SED has a power-law slope of α = translates to a hydrogen column density N ∼ 1025.7
2.55±0.18(Sν ∝να)between1.3mmand0.85mm. This cm−2 for the dust opacity to column densiHty relation
is measured from the SMA data in Figure 10, namely N(H+H )/τ = 1.2×1025(λ/400 µm)2 H-atom cm−2
from this work, S10, and the 1.3 and 1.1 mm obser- 2 λ
(Keene et al.1982;Hildebrand1983). Since the corehas
vations used in the companion paper (C12). Since the asizeof∼20pc(0′.′1inFWHMfrom§3.2.4;seethenext
SMAcontinuumdataweretakenfromline-freespectrom-
subsection for two models), we obtain n ∼ 5 × 105
eter channels in each dataset, none of the SMA data H2
cm−3 for the mean gas density (i.e., number density of
is contaminated by strong emission lines such as CO
unlike wide-band bolometer data. The spectral slope hydrogenmolecules)andMmol(r ≤10pc)∼1×108 M⊙
that we obtained is shallower (i.e., α is smaller) than forthetotalgasmassinthecoreassumingthatthegasis
mostly molecular there. The total mass depends on the
those of other local IR luminous galaxies but is con-
sistent with the results of the FIR-to-mm SED surveys densitydistribution. Theestimateaboveisforauniform
of Lisenfeld et al. (2000) and Yang & Phillips (2007), in density andis an upper limit if the gasdensity increases
eachofwhichNGC4418hastheshallowestspectralslope toward the center.
in about 15 galaxies.
3.3.3. Brightness Temperature and Luminosity
Two simple models for the spectralindex suggestthat
the α measured at the nucleus is due to the dust opac- Table 6 lists the deconvolved (peak) brightness tem-
ity τ ∼ 1 at λ ∼ 1 mm. In the first model of a sim- perature and the bolometric luminosity of the nucleus.
ple one-zone slab, the spectral slope should be α = The luminosity is derivedfromthe 860µm core size and
2+βτ(eτ −1)−1 where β is the index of power-law fre- the temperature. Here we use two different models to
quency dependence of the dust opacity and τ is the slab estimate the brightness temperature and the bolometric
opacity at the wavelength of the α measurement. Here luminosity of the nucleus as we did in S08 for Arp 220.
we used the Rayleigh-Jeans approximation because the One is the circular Gaussian we used for the visibility
brightnesstemperatureofthe 0′.′1nucleusexceeds100K fit above and the other is a circular disk with uniform
as we will see next (§3.3.3). With optically thin emis- brightness. The Gaussian model describes dust that is
sion, this model gives the familiar form of α = 2+β. not fully opaque across the nucleus and that may also
The low α observed in NGC 4418 means low β or a have a radial temperature gradient. The circular disk
moderate opacity of dust emission, i.e., τ ∼ 1 at λ ∼ 1 modelcorrespondstoasphereofuniformsurfacebright-
mm. The dust opacity about 1 is favored, as was con- ness projected on the sky. Both models can fit the same
cluded by Lisenfeld et al. (2000), because most infrared data well when the visibilities are sampled only within
luminous galaxies have β ≈ 1.5–2 (Lisenfeld et al. 2000; the half-maximum baseline length of the visibility func-
Dunne & Eales 2001). The observed α corresponds to tion. The diameter of the model disk is 1.6 times the
τ = 2.2±0.5 and 1.8±0.5 for β = 2 and 1.5, respec- FWHMofthe Gaussianinsuchacase(S08). Forthe lu-
tively. Forthesecondmodel,wenotethattheSEDslope minositycalculation,itisassumedthatthe860µmemis-
does not become 2 at shorter wavelengths as expected sion is optically thick in the uniform disk/sphere model.
for τ ≫ 1. The shallower slope of the Planck function Since the deconvolved T was used without any opacity
b
compared to the Rayleigh-Jeans formula must partly be correction(i.e.,notdividedby1−e−τ),thederivedtem-
the reason. Another likely reason is that the dust dis- perature and luminosity are lower limits. We calculated
tribution is not just a 0′.′1 slab of uniform opacity but theluminosityoftheGaussianmodelbyusingitsFWHM
presumably has a low opacity halo as suggested by the size and its peak brightness temperature. This crudely
larger extent of line emission than the 860 µm contin- approximatestheeffectmentionedearlierthatthesizeof
uum. The spectralslopeinthe generalmulti-component the photosphere should increase toward higher frequen-
6 SAKAMOTO et al.
cies as dust opacity increases with frequency. ting gives a nominal deconvolved size of 0′.′10±0′.′04 in
The derived luminosity of the nucleus is about 1 × the major axis while the source is unresolved in the mi-
1011 L , which is comparable to the total luminosity of nor axis.) The 450 µm detection of the subarcsec-scale
⊙
the galaxy (1011.1 L , Table 1). We obtained nuclear nucleus is at 10σ and hence even more secure when the
⊙
luminositieslog(L /L )=11.3±0.3and10.8±0.3for data are convolved to 0′.′5 resolution (Table 4). (The
bol ⊙
the two models. Luminosity surface density and volume S/Nincreases because the convolutionlowersthe weight
density are also calculated for each model and listed in of noisy long baselines.) The peak and total flux den-
Table 6 for the central ∼20 pc. The errors here and in sities in both resolutions are about 370 mJy and mJy
Table 6 are random errors due to the 5% flux error and beam−1, respectively.
the 11% size uncertainty and do not include systematic The 450 µm continuum flux density that we detected
error due to the models. is significantly lower than that of the central 3 kpc of
There are two notable sources of systematic errors in the galaxy (see Fig. 10). The ratio is 0.31 ± 0.15
thederivedparametersinadditiontotheassumptionson between the SMA and single-dish measurements. The
thesourceshapeandonthe860µmopacity. (Theunder- latter was made by Roche & Chandler (1993) with the
estimating effect of the latter is minor as noted in S08.) UKT14 bolometer, the filter of which had the effective
One is the absorption of the core emission by the sur- frequency of 682 GHz (440 µm) and the bandwidth of
rounding dust and gas. It arises because the ∼0′.′1 (∼20 84 GHz (Duncan et al. 1990). We subtracted from the
pc)continuumcoreisembeddedinanorderofmagnitude bolometermeasurementthecontributionofCO(6–5)line
larger envelope as we saw in §3.2. The continuum emis- (∼0.1Jy,calculatedfromtheCO(3–2)single-dishfluxof
sionfromthecoremustsufferfromextinctiontosomeex- Yao et al.(2003)scaledbyafactorof4)andrescaledthe
tent by the envelope through continuum self-absorption continuumto ourobservingfrequency using the spectral
andlineabsorptionbyvariousmolecules. Ourcontinuum index α = 2.55. The single-dish flux density of the 450
brightness temperature derived without the extinction µm continuumestimated this way is 1.18±0.33 Jy. The
correction is therefore a lower limit of the true bright- low flux recovery ratio at 450 µm suggests that there is
ness temperature at the surface of the ∼0′.′1 core. The a significant 450 µm emission that is too extended for
correction for it would increase the core luminosity. We detection at our sensitivity and with the shortest base-
will analyze this submillimeter extinction a little more line of 240 kλ in the VEX configuration. We simulated
in §3.3.4 and find the extinction to be about 5% at 860 the degree of flux recovery with our actual u–v coverage
µm. The other source of systematic error is the possible for Gaussian sources of various sizes. (We did not fit
contamination of the 860 µm continuum by molecular the visibilities unlike our 860 µm data analysis because
lines. This effect can be estimated to be about 10% as- the 450 µm data have lower S/N ratio.) Fig. 11 shows
suming that the spectralindex betweenlowerandupper that the size of the 450 µm continuum source would be
sidebands obtained in the SC configuration, 6.25±0.84 0′.′24±0′.′07(FWHM)ifthe sourcewereapproximatelya
(§3.2.3), is due to line contamination to the USB data singleGaussian. Thepeakbrightnesstemperatureofthe
andthatthetruecontinuumspectralindexis2.55±0.18 sourcewouldbe70±20Kandthebolometricluminosity
(§3.3.1). The continuum brightness temperature of the of this source would be on the order of only 1010.2 L ,
⊙
core would be lower without the contamination. How- which disagrees with our estimate at 860 µm.
ever, this does not necessarily mean that the bolomet- A two-component model with a core and an envelope
ric luminosity of the core is overestimatedby this effect. agrees with the observations within the errors and is
Thermal lines can significantly add flux density to the more reasonable in our view. The two components are
860 µm continuum only when the line and continuum introduced in view of the compact continuum core and
opacities satisfy τ > τ and τ . 1, because no the more extended CO emission both observed at 860
line cont cont
thermalline willbe seenfromablackbody. Since weuse µm. They are treated separately to simplify our model-
thebrightnesstemperatureofthecontinuumcoreforour ing though they are probably not two discrete entities.
luminosity estimate assuming that the continuum is op- (Moredetailedmulti-componentmodelshavebeenmade
tically thick, the line emission, if it is from the 0′.′1 core by Gonz´alez-Alfonso et al. (2012a) and C12. Our inten-
itself, would (partly) compensate for the decrement of tion here is not to update their models but to apply a
the observed continuum brightness temperature due to simple model to understand our sub-mm continuum ob-
τ that is not ≫1. servations.) The core and envelope are assumed to have
cont
In addition to these two error sources, our visibility sizesof0′.′1and0′.′5(bothinFWHM),respectively. Also
fits are insensitive to a point-like source with a small the core is assumed to dominate the 860 µm emission
flux contribution. This is due to our u–v coveragelimit. with 85% of the total flux density. The extended enve-
For example, a 20 mJy source with a size of 20 mas (3 lope with the small remaining flux density has too low a
pc) can be undetected in our visibility fit (Fig. 9 a) and surface brightness to be detected in our high-resolution
would have a temperature of 700 K and luminosity of map. However, the amount of extended 860 µm con-
1012 L . The total luminosity of the galaxy limits the tinuum in this model is implied by our CO(3–2) data.
⊙
presence of such compact sources. Namely,ifone thirdofthe CO(3–2)emissionis fromthe
enveloperegion,thenitsfluxofabout300Jykm s−1and
3.3.4. 450 µm Continuum the typical CO(3–2) equivalent width of 1×104 km s−1
We detected 450 µm continuum at 5.8σ with the res- amonginfraredbrightgalaxies(Seaquist et al.2004)sug-
olution of 0′.′23 × 0′.′15 at the position of the 860 µm gest a 860 µm extended continuum of ∼30 mJy, which
continuum peak (Fig. 8 (l)). The 450 µm nucleus in is about 15% of the total continuum flux density. The
the map appears unresolved but this is largely due to analysis of spectral index in §3.3.1 also implied a small
the low signal-to-noise (S/N) ratio. (Image-domain fit-
ApJ in press 7
contribution to 860 µm emission from an optically-thin from the K band bulge luminosity 109.8L , which is
⊙,K
andextendedcomponent. We assumethe spectralindex calculated from the total galaxy luminosity in Table 1
between 860 µm and 450 µm to be 2 for the optically and the mean bulge-to-total luminosity ratio of 0.31 for
thick core and 4 for the optically thin envelope (i.e., the Sa galaxies (Graham & Worley 2008). The mass should
model assumes β = 2). In addition, since the core is have a factor of 2 uncertainty due to the scatter in the
embedded in the envelope, the former should be extin- L –M correlation (Marconi & Hunt 2003).
K,bulge bh
guished by the latter. The opacity of the envelope to We also infer from information of multiple lines that
coverthe core is assumedto be ∼1 at 200 µm, which we gasmotionisfastertowardthecenter. Fig. 14andTable
took from the model of Gonz´alez-Alfonso et al. (2012a) 7showthecentroidvelocitiesandwidthsofthemolecular
noting that it makes the envelope optically thin at 860 lines measured in the SC configuration data of S10. An
µm(τ =0.05)asweassumedabove. At450µmthecore interesting observationhere is that there are two groups
is extinguished by the envelopewith τ =0.2and the en- in the centroid velocity – line FWHM plane. The first
velope itself is resolved out in our VEX data. Overall, groupconsistsofCO(3–2),CO(2–1),andN H+(3–2)and
2
this model predicts our flux recovery rate to be about havecentroidvelocitiesV ≈2100km s−1andFWHM≈
c
45%,whichagreeswithour observedvalue of31±15 %. 142km s−1. The restof the lines form the secondgroup
The currentdata qualitydoes notjustify further param- at V ≈ 2088 km s−1 and FWHM ≈ 234 km s−1. Also,
c
eter tuning and model refinement. there is a trend that lines with higher critical densities
Thismodelillustratesthattwoeffectslikelycontribute are wider. This variation of line parameters again sug-
to the lower flux recovery rate at 450 µm than 860 µm. gests that the molecular gas in the center of NGC 4418
Oneisthatthecoreemissionissaturatedwhile emission has non-uniform properties or multiple components. In-
from the extended outer regionis not. The other is that deed, the latter trend can be easily explained if the gas
the emission from the hot nucleus is extinguished more motion around the nucleus is faster closer to the center
atshorterwavelengthsbythecoldercircumnucleardust. and if denser gas is more localized around the nucleus.
These effects can also explain the 60 µm peak of the This model agrees with our previous observation in Ta-
SED of NGC 4418 despite the energetically dominant ble4thatthelineswithhighercriticaldensitiesaremore
core emitting at > 100 K at 860 µm. The same effects concentratedinthenucleusthanCO.Iflineswithhigher
probably explain similar observations in Arp 220 where critical densities indeed trace gas closer to the nucleus,
the two nuclei contribute much less to the total at 435 andifwealsoassumethatthegasdistributionisaxisym-
µm than at 860 µm (Matsushita et al. 2009). metric around the nucleus, the true systemic velocity of
the nucleus would be about 2088 km s−1.
3.4. Kinematics of Molecular Gas We can also evaluate radial motion of molecular gas
fromourdata. HCNspectranearthecontinuumpeakin
The HCN velocity map Fig. 8(f) shows in the central
Fig.15andtheHCNPVdiagram(Fig.13)showaminor
0′.′5 (80 pc) a systematic gradient of mean velocity in
depression(2σ) at2080km s−1. This may be due to gas
the northeast–southwest direction, which is close to the
ataboutV andthroughlineself-absorptionorline ab-
galaxymajoraxis(p.a.=60◦,Table1). TheHCNchannel sys
sorptionofthe continuumfromthe core. Thevelocityof
maps in Fig. 6 show this velocity structure as the shift
this possible absorptionandthe absenceof clearabsorp-
ofthe emissionpeak acrossthe continuumposition from
tionoffthesystemicvelocityindicatelittleradialmotion
southwest(lower-right)tonortheast(upper-left)inchan-
of the HCN gas. Meanwhile, the CO emission is signifi-
nels between 1990 and 2200 km s−1. In Figure 12, mo- cantly brighter below about 2100 km s−1 than at higher
mentmapsofHCN,HCO+,andCSconfirmthepresence
velocitiesinthePVdiagram. SeealsotheCOlineprofile
ofthe gradientacrossthe nucleus althoughthe direction
at the continuum core position in Fig. 15 where the line
of the velocity gradient is closer to the north–south di-
profile is asymmetric. If this is due to self-absorption
rection in these lower resolution data. This is also the
of the CO emission, the foreground gas causing the ab-
case in the CO high-resolution data (Fig. 8 (b)) in the
sorption (and weaker CO emission) is redshifted with
central 0′.′5.
respect to the systemic velocity and hence is flowing in-
We attribute the central velocity gradient to rotation
ward. Although this asymmetry can be due to a chance
around the luminous core and estimate its dynamical
imbalance of gas distribution between approaching and
masstobeM (r ≤15pc)∼2×108M . Weestimated
dyn ⊙ receding halves around the galactic center, this is quali-
forthisavelocityshiftofabout400km s−1 in0′.′2across tatively consistent with the redshifted OI, OH, and HI
thenucleus fromthe COandHCNposition-velocitydia- absorption observed by Gonz´alez-Alfonso et al. (2012a)
gramsin Fig. 13, where almostthe full velocity rangeof and Costagliola et al. (2012). The plausible gas motion
eachline is observedatthe central∼0′.′3. The Keplerian mentioned here is opposite to that in Arp 220 where P-
dynamical mass above is for the 62◦ inclination of the Cygniprofilesofmolecularlinessuggestedmolecularout-
galaxy. Because of the insufficient spatial resolution to flows (Sakamoto et al. 2009).
obtain an accurate rotation curve, our Mdyn estimate is Gasmotionoutside the central0′.′5does notappearto
admittedlycrudeanditserrormaybeaslargeasafactor be ordinaryrotationinour data. The CO meanvelocity
of2eveniftheadoptedinclinationiscorrect. Iftherota- mapsFigs.7(b)and8(b)donotshowanordinaryspider
tion velocity increases toward the center at these scales, pattern of a rotating gas disk at r ≥ 0′.′25. The velocity
aswewillseepossiblebelow,thedynamicalmassismore mapsprobablycontainamixtureofrotationalandradial
likelytobeoverestimated. Forcomparison,weestimated motionsandappearcomplexbecauseofnon-uniformgas
in our companion paper the mass of the central black distribution and optical thickness of the CO line.
hole in this galaxy to be 9×106 M (Costagliola et al.
⊙
2012, for the distance of 34 Mpc). This value follows
8 SAKAMOTO et al.
3.5. Gas Properties tinuumcore(S08). Finally, the ratiobetweenHCN(4–3)
to HCO+(4–3) is 2.08±0.27 in the central 0′.′5 while it
Our high resolution observations also provide the fol-
is 1.65±0.07 in the central 6′′ (S10). Among other dif-
lowing new insights into gas properties in the nucleus.
ferences, HCN has an order of magnitude higher critical
First,theCO(3–2)linehasapeakbrightnesstempera-
tureof90Kat0′.′3resolution(Table4),indicatingwarm density for collisional excitation and is also easier to ra-
diatively excite than HCO+ (S10). See §4.1.8 for more
molecular gas around the nucleus. The brightness tem-
discussion on this line ratio.
perature of thermalized CO is the gas kinetic temper-
ature diluted by the CO opacity and by the observing 3.6. Optical Color Index
beamifitis largerthanthe source. Bothdilutioneffects
are probably small for the extended 12CO in this case. We found a U-shaped red feature in the multi-band
optical images of the galaxy from the Sloan Digital Sky
ToputtheobservedT intocontext,thepeakvalueof90
b Survey (SDSS). Fig. 16 shows the distribution of a flux
K at 50 pc resolution surpasses the peak CO brightness
ratioinvolvingfourSDSSbands,(i′+z′)/(g′+r′). Areas
temperatures of about 50 K observed in Arp 220 and
NGC 253 at 100 pc (0′.′3) and 20 pc (1′′) resolutions, re- brighter at longer wavelengths (i′ and z′) are shown in
red while areas of lower flux ratios are in blue. The U-
spectively(Sakamoto et al.2008,2011). Meanwhileboth
HCN(4–3)andHCO+(4–3)havepeakbrightnesstemper- shapedredfeature is alongthe northwesternsemi-minor
atures of about 30 K at the same 0′.′3 resolution (Table axis of the galaxy. No counterpart is visible around the
southeastern semi-minor axis. The reddest point is al-
4). Their ∼3 times lower T than that of CO must be
b most on our 860 µm continuum peak (i.e., the nucleus)
largely due to beam dilution effects for these lines be-
and is about 0.5 mag (i.e., a factor ∼1.6) redder than
cause they were found more compact than CO(3–2) in
the off-center region outside the U-shape. This feature
§3.2. Subthermal excitation and lower opacities of HCN
and HCO+ may be additional reasons for the lower T extends at least up to 10′′ (1.7 kpc on the sky) from the
b nucleus. The redder part of it near the nucleus has an
of these lines.
elongationalong the position angle of about 15◦, similar
Second, the detection of lines from vibrationally ex-
to the faint CO(3–2) emission in Fig. 7.
cited molecules is another indication of high tempera-
Weinterpretthe redfeaturetobeanoutflowconeem-
tures in the nucleus. The HCN and HC N lines that
3 anating from the galactic center along the rotation axis
we detected from their vibrationally excited states have
of the galaxy. The feature is very unlikely to be a struc-
upper energy levels 510–1070 K above the ground level.
ture on the galactic plane because such a feature would
Vibrational temperatures of these molecules have been
be tornby differentialrotationand also because the fea-
estimatedtobearound300K(Costagliola & Aalto2010;
ture is on the minor axis of the galaxy. The minor axis
S10;C12). Infraredradiationhasbeensuggestedtoplay
is determined by our viewpoint while a radial feature in
a significant role for the vibrational excitation in the
the galactic plane should have no preferred position an-
∼100–200 K gas and dust that we inferred from bright-
gle with respect to us. The outflow cone model does not
nesstemperatures. Ourcompanionpaper(C12)presents
have this problem because any feature along the rota-
amoredetailedanalysisanddiscussionontheexcitation.
Third, the mean gas density n ∼ 5×105 cm−3 of tion axis of the galaxy is always projected onto the mi-
H2 nor axis. Also, the outflow can be bipolar in this model
the ∼20 pc core (§3.3.2) is high enough for CO excita-
tion to J=3 but not for HCN and HCO+ to J=4. Their even though we only see half of it in the color index
map. The fact that we see the red feature only on the
critical densities for collisional excitation at 100 K are
∼104.5,107, and 108 cm−3, respectively. The mean den- northwesternsidesuggeststhesidetobethenearsideof
sity is also far short of the ∼1011 cm−3 critical density the outflow and the far side of the galaxy disk, against
which the outflow cone is visible as a red silhouette due
to vibrationally excite HCN with H collisions. There-
2 tocolor-dependentextinction. Dustintheconeprobably
fore the excitation of the lines with high critical den-
causes the extinction and the reddening. Our CO veloc-
sities are probably achieved through photon trapping,
ity data in Fig. 7 cover only a small fraction of the red
non-uniform density distribution to allow higher local
optical feature and do not reveal the gas motion in the
densities than the mean, mid-infrared radiation, and/or
redcone. Thedataonlytellthatthevelocityfieldwithin
electron collisions (S10; C12).
a radius of ∼500 pc is more complicated than expected
Fourth, the ratios between lines and between line and
from purely circular rotation around the center. How-
continuumvarywiththeareasizeofthemeasurementsin
ever, Lehnert & Heckman (1995) reported a 200 km s−1
thewayexpectedwhentheinterstellarmediumiswarmer
velocity shift along the minor axis from their [N II] line
and denser at smaller radii. This is a consequence of
measurementsforasurveyofstarburstwinds. Theveloc-
the size variation among various emission lines (§3.2).
ity shift suggests radial motion of the ionized gas along
For example the HCN(4–3)/CO(3–2) ratio of integrated
the polar axis. Although we cannot yet confirm an out-
brightnesstemperaturesis0.32±0.03and0.170±0.004at
0′.′5and6′′ apertures,respectively. Theratioisexpected ward radial motion because the direction of the [N II]
velocity shift was not given, the velocity shift along the
tobehigherwhenmoleculargasisdenser,warmer,orfa-
red cone better fits the outflow interpretation. For this
vorable for HCN line emission in other ways. Similarly,
radial motion to be an inflow, gas and dust need to be
the equivalentwidths oflines withrespectto the contin-
falling in the U-shape cone towardthe nucleus along the
uum depend on the size scale. The CO(3–2) equivalent
polar axis, which seems too much of a coincidence. Our
width changes by a factor of 3, between 1.4×103 and
outflow model predicts that the gas onthe northwestern
4.2× 103 km s−1, from 0′.′5 to 6′′ scales. The smaller
coneshouldbeblueshiftedonaveragewithrespecttothe
equivalent width at the smaller scale can be explained
nucleus.
by the higher opacity of the dust continuum at the con-
Much closer to the nucleus, Evans et al. (2003) found
ApJ in press 9
radialdarklanesouttoabout3′′fromthenucleusintheir mass loss rate of the galaxy is smaller than the outflow
near-infrared data. The lanes have an average position rate from the nucleus.
angle of about 15◦ and are on both sides of the nucleus.
4. DISCUSSION
The fainter lane to the north is in the same direction
as the reddest part of the optical U-shaped feature as 4.1. The Luminosity Source at the Nucleus
well as the CO elongation. There might be therefore a We have obtained several parameters to constrain the
changeoftheoutflowdirection,firstatP.A.∼15◦ inthe nature of the luminosity source such as the L , M ,
innermost region (r < 3′′ ≈ 500 pc) and then along the size, and N of the luminous core, and Σb(oLl )dy∼n
minor axis of the galaxy (P.A. ∼−30◦) outward. 108.5 L pc−H2 and L /M ∼ 500 L M −1boclalcu-
Weestimatethegasmassintheoutflowtobe∼4×107 ⊙ bol dyn ⊙ ⊙
lated from them. While we derived these from our 860
M . This is from the reddening and assumes a bipo-
⊙ µm observations alone they qualitatively agree with the
lar structure of the outflow. The excess color index in
observationsofRoche et al.(1986)andmanyothersthat
g′−r′ is 0.8 mag at the central peak, about 0.5 mag in
most of the luminosity of the galaxy must come from a
the reddest 1 kpc2 near the base of the U-shape, and 1
compact and deeply enshrouded nucleus. The param-
mag kpc2 when integrated over 2×2 kpc2 encompass-
eter values are comparable to or higher than those in
ing the entireU-shape. Adopting the Milky-Wayextinc-
the ultraluminous infrared galaxy Arp 220, where S08
tion law of Ag′ −Az′ = 0.71AV and Σ(H+H2)/AV = obtained L /M & 400 L M −1 and Σ(L ) &
2×1021 Hcm−2 = 16 M pc−2, we estimate the gas bol dyn ⊙ ⊙ bol
⊙ 107.6L pc−2 inthecentral80pcofthewesternnucleus.
column density to be N = 2×1021 cm−2 at the peak, ⊙
H
the mean mass surface density in the reddest 1 kpc2 to 4.1.1. Constraint from L/M
be about 10 M pc−2, and the gas mass in the one side
⊙ Ayoungstarburstcanhaveabolometricluminosityto
of the outflow (i.e., the U-shape) to be about 2× 107
mass ratio at or above 1000 L M −1 for 3–10 Myr de-
M . The mass in the entire outflow is twice the last ⊙ ⊙
⊙
pendingontheinitialmassfunctionwhileabareAGNat
value on the assumption of a symmetric bipolar struc-
ture. This mass estimate assumes foreground-screen ex- theEddingtonluminosityhasL/Mbh ≈104.5 L⊙ M⊙−1.
tinction. The 860 µm dust core mixed with stars in the Fig. 17 shows this for starbursts simulated using
galactic center does not contribute much to the redden- Starburst99 (Leitherer et al. 1999; Va´zquez & Leitherer
ingand,appropriately,tothemassestimatedhere. Note 2005; Leitherer et al. 2010) and the initial mass func-
thatthereddeningdata(andhencetheoutflowmassesti- tion (IMF) of Kroupa (2002). The duration of L/M &
matefromit)ismoresensitivethanourCOdatainwhich 1×103L⊙ M⊙−1 isabout5Myrforthenormal(i.e.,un-
the U-shapedconeisnotdetected. Theconewouldhave truncated) Kroupa IMF in both instantaneous and con-
aCOlinebrightnesstemperatureontheorderofonly0.1 tinuous starbursts. In a continuous starburst, the com-
Kwhentheoutflowgasis100%molecular,its CO-to-H bination of L/M ≈ 103 L M −1 and L ≈ 1011 L is
2 ⊙ ⊙ ⊙
conversionfactoris2×1020cm−2 (K km s−1)−1,andthe obtainedwiththe star formationrates(SFRs) of30–100
line has a velocity width of 30 km s−1 at each position. M yr−1 at the ages of 3–1 Myr. The starburst can be
⊙
This is about 0.5σ in our 0′.′6 resolution CO(3–2) data older and SFR lower if the IMF is biased toward high
cube. massescomparedto the standardone (e.g., age∼10Myr
Wefurtherestimatethekinematicalageoftheoutflow and SFR∼10 M yr−1 for the IMF mass range of 1–
⊙
to be on the order 10 v−1 Myr and the gas outflow rate 100 M ). The mass used for L/M in Fig. 17 is that of
200 ⊙
∼4 v M yr−1. The parameterv is the outflowve- newbornstars(orthe blackhole massforthe Eddington
200 ⊙ 200
locity normalized by 200 km s−1 and is about unity for limit). The dynamical mass in our observed Lbol/Mdyn
the projected minor-axis velocity shear of 200 km s−1 includes the gas covering the nucleus, the central black
measured in [N II]. The age (i.e., crossing time) of the hole, and the old stellar population preexisting the star-
outflow is the ratio between the de-projected size of the burst or nuclear activity. The gas mass is not negligible
outflow cone, 2 kpc, and the outflow velocity. The mass since it was estimated to be about half of the dynami-
outflowratefromthenucleusisfromthegasmassabove cal mass (§3.3.2). Removing these from the denomina-
and the crossing time. We caution about the uncertain- tor of the observed L/M ∼ 500 L M −1 will increase
⊙ ⊙
tiesintheseparametersduetotheunknownvelocitiesof the ratio to &1000L M −1. Fig. 17 therefore suggests
⊙ ⊙
various outflow media, the poorly constrained extent of that if a starburst is responsible for the core luminos-
the flow, and the uncertain projection effect. For exam- ity of NGC 4418 its L/M is near the maximum that a
ple,the ionizedgastracedby[NII]andthedustcausing starburst can have only when it is young, assuming the
the reddening may have different bulk velocities. Also standard IMF. The L/M data from our submillimeter
the opening angle of the outflow and the possible bend observations therefore provide a new way to constrain
of the outflow direction add uncertainty to our outflow the age in the starburst model.
velocity and to the parameters dependent on it. For ex- Regarding whether a compact and massive star clus-
ample, if the outflow velocity is as large as 1000 km s−1 ter mentioned above is able to form, it has been ar-
seeninsomemolecularoutflows(e.g.,Chung et al.2011; gued that the maximum L/M for a cluster-forming
Sakamoto 2012), the crossing time will be reduced to 2 gas cloud or disk is 500–1000 L M −1 (Scoville 2003;
Myrandtheoutflowratewillbe∼20M yr−1. Another ⊙ ⊙
⊙ Thompson et al. 2005). This limit is because too much
cautionisthat,withthevelocityinformationcurrentlyat
luminosity will blow out the star forming gas with radi-
hand,wecannottellwhetherthegasanddustblownout
ation pressure on dust and halt the star formation. The
fromthenucleusto&1kpcabovethegalacticplanewill
observed L/M is close to this limit but does not clearly
leave the galaxy or fall back to it. In the latter case, the
exceed it. Murray (2009) predicted a mass-size relation
10 SAKAMOTO et al.
for massive clusters taking the radiation pressure into Table3), their conversionformulaeto radialmassdistri-
account. The diameters of 107.5 and 108 M clusters, bution(Terzi´c & Graham2005),andtheK-bandmassto
⊙
which can have ∼1011 L when young, are 5 and 10 lightratio1.06M /L fora12Gyroldstellarpopula-
⊙ ⊙ ⊙K
pc respectively in his relation. Thus at least these con- tionwith[Fe/H]=0(Worthey1994). Fromtheviewpoint
siderations do not reject the starburst model with the of mass budget, AGN dominance in luminosity would
parameters we currently have. eliminate the need for young stars and help fit the sum
If the energy source of the 1011L core is an ac- of the component masses to the dynamical mass. The
⊙
creting black hole then its Eddington ratio must be information currently at hand, however, does not let us
log(L/L )=−0.5±0.3 assuming the black hole mass conclude the AGN dominance fromthe mass budget be-
Edd
of 107.0±0.3M (hence L = 1011.5±0.3 L ) estimated causethemassesusedheretypicallyhaveanuncertainty
⊙ Edd ⊙
fromthebulgeluminosity(C12,§3.4). ThusanAGNhas of a factor of a few. The dynamical mass will be better
no problem to be the dominant energy source as long as constrained with higher resolution observations.
theL/M isconcerned. Ithasnoproblemtobesmaller
bh
than the 20 pc core either. 4.1.4. Constraint from Time Scale
Thecurrenttentativeestimateoftheoutflowkinemat-
4.1.2. Constraint from Σ(L )
bol ical age, 10 Myr, is marginally longer than the age that
The luminosity surface density that we obtained, anenergetically-dominantstarburstcanhave(§4.1.1). If
Σ(Lbol) = 108.5±0.5 L⊙ pc−2 in the central 20 pc (Ta- this is firmly confirmed, it would be against the nascent
ble 6), is among the highest compared to the mid- starburst model.
infraredsurveyof(ultra)luminousinfraredgalaxies,star-
bursts, and Seyfert nuclei by Soifer et al. (2000, 2001, 4.1.5. Constraint from NH
2003, 2004). In their 0′.′3–0′.′5 resolution survey, only
The gas shroud around the nucleus can shield the X-
three out of twenty-two nuclei were found to have a sur-
rays from the putative AGN because it is highly Comp-
face brightness (or its lower limit) at 108 L⊙ pc−2 or ton thick with N & 1025 cm−2 to the galactic center
larger. All the three galaxies, Mrk 231, NGC 1275, and H
(§3.3.2).
NGC 7469,have an AGN. Their mid-IR cores have (up-
per limit) sizes of about 20 pc in two and 100 pc in 4.1.6. Provisional Verdict
one. For comparison, the luminosity surface densities
To summarize on the luminosity source, the large lu-
of star formation-dominated nuclei are generally at or
below 107 L pc−2. Between 107 and 108 L pc−2 minosity of the nucleus cancertainly be due to a hidden
⊙ ⊙
AGNasfarastheparametersL,L/M,Σ(L ),N ,and
are ultraluminous infrared galaxies where both AGN bol H
massbudgetofthe nucleusareconcerned. Itwouldhave
and starburst may be hidden behind a large amount
an Eddington ratio ∼0.3 with a factor of 2 uncertainty
of gas and dust (Soifer et al. 2001). More recent obser-
if the black hole and the bulge follow the normal scaling
vations by Imanishi et al. (2011) also support the mid-
relation. This bolometric Eddington ratio is high com-
IR compactness and higher luminosity surface densities
pared to the median ratio of 10−3 among local type 1
of AGNs compared to starbursts. There is a theoret-
Seyferts (Ho 2008). In this sense the AGN mass accre-
ical model explaining the apparent limit for starbursts
around 107 L pc−2. In it a nuclear gas disk opaque tion is rapid in NGC 4418 if the AGN is the dominant
⊙
luminosity source. The black hole should be growing at
at far-IR will have that luminosity surface density when
a higher rate than in most low luminosity AGNs. The
it supports itself by stellar radiation pressure and self-
mass accretion rate to the black hole would be 0.1 M
regulates to have Toomre’s Q ∼ 1 (Thompson et al. ⊙
yr−1 for a 10% radiative efficiency and the luminosity
2005). This is for dust temperatures less than about
200Korfor“warmstarbursts”inAndrews & Thompson of 1011L⊙ (i.e., Lbol ≈0.1M˙ c2). The characteristictime
(2011). For dust temperatures above 200 K, which the scalefortheblackholegrowthis108yratthisrate. This
0′.′1 core of NGC 4418 may have, and for a gas surface AGN gas consumption rate is much lower than the star
densityof105.8M pc−2thatcorrespondstoτ ∼1 formation rate required for the same luminosity because
⊙ 860µm
ofthe continuumcore,the diskatthe Eddingtonlimitis an AGN is more fuel-efficient.
in the “hot starburst” regime and will have a surface Acompactyoungstarburstthathasamass∼108 M⊙,
brightness of ∼108.5 L pc−2 (Andrews & Thompson size<20pc,age∼afewMyr,andL/M ∼103L M −1
⊙ ⊙ ⊙
2011, see their Equation 7). According to our data isanalternativethathasnotbeenexcludedwithourob-
and this model, the nucleus of NGC 4418 has either an servations. (The age could be up to 10 Myr as far as
energetically-dominantAGN or the hot starburst. L/M is concerned, but starbursts older than 3 Myr are
less favored because supernova explosions would be de-
4.1.3. Constraint from Mass Budget
structive to the dusty core as argued by Roussel et al.
Our estimates of the masses in the central 30 pc of (2003).) While such a starburst would be at an edge of
NGC 4418 are summarized in Table 8. The mass of theparameterspaceallowedforstarbursts,itisnotewor-
young stars there, 108.0±0.5 M⊙, is for a model young thy that the massive concentration of dense molecular
starburst having a luminosity of 1011 L⊙ and a plausi- gas is a favorable environment for star formation. This
ble luminosity-to-massratio of 103.0±0.5 L M −1. The is true even if there is a major AGN.
⊙ ⊙
mean star formation rate of this starburst would be 10– A combination of less luminous young starburstand a
100 M yr−1 for a starburst age of 3 Myr. The mass of less luminous AGN is also possible as long as their total
⊙
oldstarsinthecentral30pc,∼107.4M ,isestimatedfor luminosity is about 1011 L . It remains to be seen with
⊙ ⊙
anaverage(∼L∗) Sa-type galaxyusing the K-bandpho- more accurate mass, luminosity, and size measurements
tometric parameters of Graham & Worley (2008, their andtheconstraintsmentionedabovewhetherthenucleus