The Circumnuclear *Multi-phase* Torus in the Circinus Galaxy Revealed by ALMA

1

June 6-8, 2018 Galaxy-Evolution Workshop 2018@Ehime Univ.

Takuma Izumi(NAOJ Fellow), 
K.Wada, R.Fukushige, S.Hamamura (Kagoshima Univ.),

K.Kohno (IoA/Univ. of Tokyo)

→ Izumi et al. 2018c, to be submitted

2

~a few to 10 pc?

Urry & Padovani 1995, PASP, 107, 803

(Strict-) Unified scheme

(e.g., Antonucci 1993, ARA&A, 31, 473)

Type-1

Type-2

Hα

λ (Å)

SMBH obscuration: Torus

Type-1 Type-2

Q. Physical origin of the torus??

3Big challenge to the paradigm

• Polar elongation in MIR continuum of the Circinus galaxy!?

• Statistical confirmation (e.g., Lopez-Gonzaga et al. 2016)

A&A

563,

A82 (

2014

)

direc

tlyest

imate

d from

thebo

otstra

p dist

ributi

onof

theres

pec-

tive p

arame

teran

d mark

the68.3%

(1σ) c

onfid

ence

interv

als.

Our m

odel

can

fitthe

data

onthe

short

estba

seline

s very

well,

which

mean

s tha

t it r

eprod

uces

thelow

spati

alfre

quen

-

cies o

f the

sourc

e ade

quate

ly.On

longe

r base

lines,

howe

ver, t

he

data

isno

t well

reprod

uced

byou

r mod

el.Th

isis

predo

mina

ntly

due t

o sma

ll sca

leva

riatio

nsof

theco

rrelat

edflu

xes a

nddiff

er-

entia

l pha

sesat

longe

r base

lines

(cf. F

ig.4),

which

cann

otbe

reprod

uced

byou

r smo

othmo

del. W

e inte

rpret

these

varia

tions

assig

natur

esfor

small

scale

struc

tures

that o

urmo

del o

bviou

sly

cann

otrep

licate

.

Finall

y,a f

ewrem

arks o

n deg

enera

cies:

severa

l para

meter

s

ofou

r mod

elare

not i

ndep

ende

nt.Th

e clea

rest e

xamp

leis

the

dege

nerac

y betw

een

thetem

perat

ureT i

and

thesu

rface

filling

factor

f i.Be

caus

ewe

arefitt

inga

narro

wwa

velen

gthran

ge

(8µm<λ<

13µm

),the

tempe

rature

s of o

urdu

stco

mpon

ents

areno

t well

cons

traine

d.A

small

chan

gein

tempe

rature

has a

direc

t infl

uenc

e on

thebri

ghtne

ssof

theso

urce,

which

can

be

comp

ensat

edby

chan

ging t

hesu

rface

filling

factor

. Sim

ilar d

e-

gene

racies

arepre

sent b

etwee

n the

size a

ndthe

axis

ratio

ofthe

sourc

e,wh

ichall

chan

gethe

emitte

d flux

dens

ity. D

epen

ding o

n

how

well t

hese

param

eters

areco

nstra

ined b

y the

interf

erome

tric

measu

remen

ts,the

sepa

ramete

rsca

n bec

ome d

egen

erate.

5.Di

scus

sion

5.1. M

orph

ology

The

direc

t ana

lysis

ofthe

data

(Sec

t.3)

and

our m

odell

ing

(Sec

t.4)

confi

rmtha

t the

mid-i

nfrare

d emi

ssion

inthe

nucle

us

ofthe

Circi

nus g

alaxy

come

s from

atlea

sttw

o dist

inct c

ompo

-

nents

: ahig

hlyelo

ngate

d,co

mpac

t “dis

k-like

” com

pone

ntan

d

a mod

eratel

y elon

gated

, exte

nded

comp

onen

t. To s

ome d

egree

,

thedis

tincti

onbe

twee

n the

two c

ompo

nents

issu

ggest

edby

the

two

differe

ntreg

imes

ofthe

corre

lated

fluxe

s as a

functi

onof

thepro

jected

basel

inelen

gth(se

eSe

ct.3.1

).Pr

imari

ly,ho

w-

ever,

thedis

tincti

onis

sugg

ested

bythe

differe

ntori

entat

ions o

f

thetw

o com

pone

nts: th

e two

comp

onen

tsare

elong

ated r

ough

ly

perpe

ndicu

larto

one a

nothe

r.Tw

ocle

arly

separa

tedem

ission

comp

onen

tsha

veals

o bee

n fou

ndin

NGC

1068

and N

GC37

83

(Rab

anet

al.20

09; H

önig

etal.

2013

),an

da t

wo-co

mpon

ent

morph

ology

inthe

infrar

edap

pears

tobe

comm

onto

alar

ge

numb

erof

AGN

(Kish

imoto

etal.

2011

b;Bu

rtsch

eret

al.20

13).

We

interp

retthe

mid-i

nfrare

dem

ission

asem

ission

from

warm

dust

inthe

conte

xtof

thehy

drod

ynam

icmo

dels

ofdu

sty

tori in

AGN

bySc

hartm

ann e

t al. (

2009

), Wad

a et a

l. (20

09) a

nd

Wad

a (20

12). T

hese

mode

lsfin

d arel

ative

lyco

ld,ge

ometr

ically

thin a

ndve

rytur

bulen

t disk

inthe

mid-p

lane o

f the

torus

, sur-

round

edby

a filam

entar

y stru

cture.

The l

atter

cons

ists o

f lon

g

radial

filame

ntswi

tha h

otten

uous

mediu

min

betw

een.

We a

s-

socia

tethe

centr

al,hig

hlyelo

ngate

d com

pone

ntin

theCi

rcinu

s

nucle

uswi

ththe

dens

e disk

inthe

sesim

ulatio

ns, a

ndwe

inter-

pret th

e exte

nded

mid-i

nfrare

d emi

ssion

inthe

conte

xtof

thefil-

amen

tary t

orus s

tructu

resee

n in t

hese

mode

ls.

Afal

se-co

lour i

mage

ofou

r best

fitting

mode

l (fit

3)is

show

nin

Fig. 7

, with

themo

del i

mage

s at 1

3.0µm

, 10.5µm

and 8.0µm

mapp

edto

thered

, gree

n and

blue c

hann

elsof

the

imag

e,res

pecti

vely.

Whe

n inte

rpreti

ngou

r obs

ervati

ons, w

e hav

e to t

ake i

ntoac

-

coun

t tha

t the

emiss

ionis

domi

nated

bythe

warm

estdu

stat

a

certa

inloc

ation

, whic

h norm

ally c

omes

from

thedu

stclo

uds d

i-

rectly

illumi

nated

bythe

centr

alUV

sourc

e.Th

ereare

proba

bly

also c

onsid

erable

amou

ntsof

coole

r dus

t.Ho

weve

r,the

coole

r

1 pc

−100

−50

0

50

100

offse

t δRA

[mas

]

−100

−50

0

50

100

offse

t δDE

C [m

as]

E

N

Fig.

7.Fa

lse-co

lour i

mage

ofthe

three

-comp

onen

t mod

elfor

themi

d-

infrar

edem

ission

ofthe

nucle

usof

theCi

rcinu

s gala

xy(fi

t 3).

The

colou

rsred

, gree

n and

blue c

orresp

ond t

o the

mode

l at 1

3.0µm

, 10.5µm

and 8.0µm

, resp

ectiv

ely. T

heco

lour s

calin

g is l

ogari

thmic

inord

erto

show

both

brigh

t and

faint

featur

es.Cl

early

theco

lour g

radien

t of t

he

exten

ded

comp

onen

t due

tothe

increa

sein

thesil

icate

depth

toward

s

theso

uth-ea

stis

visibl

e.Th

isco

lour g

radien

t lead

s to a

chrom

atic p

ho-

tocen

tresh

ifttow

ards t

heno

rth-w

est. D

espite

thelow

ersu

rface

brigh

t-

ness,

80%

ofthe

emiss

ionco

mes f

romthe

exten

ded c

ompo

nent.

Also

plotte

d is t

hetra

ceof

thewa

terma

serdis

k:the

blue a

ndred

parts

trace

theap

proac

hing a

ndrec

eding

sides

ofthe

maser

disk r

espec

tively

. Note

that t

herel

ative

offset

ofthe

mid-i

nfrare

d emi

ssion

with

respe

ctto

the

maser

disk i

s not

know

n (see

text f

orde

tails)

.

mater

ialon

lyco

ntribu

tesins

ignific

antly

tothe

infrar

edem

ission

(see a

lsoSe

ct.5.3

).

5.1.1.

The d

isk-lik

e com

pone

nt

The d

isk-li

keco

mpon

ent i

s high

lyelo

ngate

dan

dha

s ama

jor

axis

FWHM

of∆ 2∼

1.1pc

. Due

tothe

stron

g pos

ition

angle

depe

nden

cyof

theco

rrelat

edflu

xes f

orthe

longe

stba

seline

s,

thepo

sition

angle

ofthe

major

axis

isve

rywe

llco

nstra

ined:

ψ 2=

46± 3

°.Th

e stro

ngelo

ngati

onof

this c

ompo

nent

with

an

axis

ratio

ofmo

retha

n 6: 1

atfir

stsu

ggest

s an i

nterpr

etatio

n

asa h

ighly

inclin

eddis

k,as

inTr

istram

etal.

(2007

).Th

isin-

terpre

tation

issu

pport

edby

theclo

seag

reeme

ntin

orien

tation

and

size o

f this

comp

onen

t with

thewa

rped

maser

disk

from

Gree

nhill

etal.

(2003

). The

maser

s were

mode

lled b

y athi

n disk

exten

ding f

romr in∼ 0.1

pcto

r out∼ 0.4

pc. T

hema

serdis

k is

warpe

dwi

ththe

posit

ionan

glech

angin

gfro

m29

° ±3°

atr in

to56

° ±6°

atr ou

t.W

itha p

ositio

n ang

leofψ 2∼ 4

6°, o

urdis

k-

like c

ompo

nent

now

match

esthi

s orie

ntatio

n muc

h bett

ertha

n

previo

usly.

The l

arger

size o

f the

mid-i

nfrare

d disk

asco

mpare

d

tothe

maser

disk

could

beev

idenc

e of t

hedis

kex

tendin

g out

tolar

ger r

adii

than

ispro

bed

bythe

maser

emiss

ion. W

e em-

phasi

setha

t the a

greem

ent is

only

inori

entat

ionan

d size

, not

in

theab

solut

e pos

ition.

With

MID

I alon

e,no

abso

lute a

strom

etry

ispo

ssible

beca

use t

heab

solut

e pha

sesig

nal is

destr

oyed

bythe

atmos

phere

(see S

ect. 2

.2). B

y con

seque

nce,

therel

ative

posit

ion

betw

een t

hema

serdis

k and

our d

isk-li

keco

mpon

ent c

anno

t be

A82,

page

12of

30

A&A 563, A82 (2014)

100 50 0 -50 -100u [m]

-100

-50

0

50

100

v [m

]

N

E

U1-U2

U1-U3

U2-U3

U2-U4

U3-U4

E0-G0

H0-G0

D0-B2

0

2

4

6

8

Fco

r [Jy

]Fig. 4. Correlated fluxes of the Circinus galaxyat 12 µm for all uv points containing useful in-terferometric data. The points are colour-codedaccording to their correlated flux, Fcor(12 µm),using a square root colour scaling as indicatedin the colour bar on the right. The uv point atthe origin represents the averaged total flux ofthe source, which is outside the plotted range ofcolours: Ftot(12 µm) = 10.7 Jy.

3. Results

In total we obtained 152 useful measurements of the correlatedflux spectra and differential phases and 74 useful measurementsof the total flux spectra. This includes 20 correlated flux mea-surements already published in Tristram et al. (2007).

The new reduction of the previously published data in gen-eral increases the data quality. The positive bias of the corre-lated fluxes and visibilities in the ozone feature between 9.5and 10.0 µm and at the edges of the N-band is reduced, espe-cially for the data observed in 2004 and 2005. The more accurategroup delay estimation leads to a slight increase in the correlatedfluxes at the long wavelength end, but the overall spectral shapeand flux levels remain unchanged. With correlated flux levels ofmore than 0.4 Jy at 12.0 µm in most cases, the result is robustwith respect to the data reduction, and we obtain no contradic-tions to the values published in 2007. Due to the improved mask-ing and sky residual estimation, the scatter of the total flux spec-tra is reduced significantly. The wavelength calibration of theMIDI spectra in EWS was also corrected slightly, resulting in ashift of the spectra to shorter wavelengths by about 0.1 µm. Allspectra were corrected for the peculiar redshift of the Circinusgalaxy of z = 0.00145 (vsys = 434 ± 3 km s−1, Koribalski et al.2004).

All 74 useful measurements of the total flux spectra werecombined by a weighted average to obtain a single estimate forthe total flux spectrum of the Circinus nucleus: F tot. The spec-trum agrees with the one published in 2007. It is shown as partof Fig. 10. The spectrum rises from ∼6 Jy at 8 µm to ∼16 Jyat 13 µm, which is quite “red” (Ftot(8 µm) < F tot(13 µm)) and in-dicative of emission from warm (T ∼ 290 K) dust. The spectrumis dominated by a deep silicate absorption feature over almostthe entire N-band. For the following, we will consider the total

flux spectrum as a measurement with a projected baseline lengthof BL = 0 m.

We use all measurements of the correlated fluxes and phasesindividually and do not average measurements close in uv space.All useful correlated fluxes at 12 µm are listed in Table A.1and plotted in Fig. 4. With correlated fluxes at 12 µm rangingfrom ∼8 Jy (corresponding to V ∼ 0.8) on the shortest baselinesto ∼0.4 Jy (V ∼ 0.04) on certain long baselines, we clearly re-solve the mid-infrared emission in the nucleus of the Circinusgalaxy. The uv plane also shows that along certain position an-gles, the correlated flux is higher than along others. A veryprominent example is the increase at the end of the baselineUT2-UT4, leading to cyan and green colours (corresponding toFcor(12 µm) > 1.5 Jy) in Fig. 4. This increase will be discussedin Sect. 3.2.

Many of the correlated flux spectra (see Fig. A.1 in theappendix) have a shape similar to the total flux spectrum.Especially on short baselines, the correlated flux spectra are sim-ilar to the total flux spectrum when the spectral change due to theresolution effect at different wavelengths is taken into account.

On longer baselines, however, this is not always the case.Most noticeably, the short wavelength emission often either dis-appears completely or there is a downturn in the correlated fluxshortward of 8.7 µm without any significant signal in the differ-ential phases (see e.g. C6 in Fig. 2). Interestingly, a similar de-crease, albeit mainly in the visibilities and for λ < 9.0 µm, mightalso be present in certain uv points for NGC 424 (Hönig et al.2012) and NGC 3783 (Hönig et al. 2013). Contamination of thetotal flux by the wings of a spatially extended polycyclic aro-matic hydrocarbon (PAH) feature at 7.7 µm was discussed as apossible reason in Hönig et al. (2012). This would, however, onlyaffect the visibilities and not the correlated fluxes we are consid-ering here. Furthermore, the 8.6 and 11.3 µm PAH features are

A82, page 8 of 30

12μm/VLTI

Pola

r elo

ngat

ion!

H2O maser disk

Tristram et al. 2014, A&A, 563, A82

Circinus galaxy

4Multi-phase Dynamic Torus model

Wada 2012, ApJ, 758, 66

5Fountain scheme + Dust rad. transfer

Schartmann et al. 2014, MNRAS, 445, 3878

IR emission from central obscuring structures 3883

Figure 3. Wavelength dependence of model ER20 at a time 2.0 Myr after the central radiation has been switched on. The various columns correspond toinclination angles of 0◦, 30◦, 60◦ and 90◦ (from left to right). The rows correspond to wavelengths of 0.1 (upper row), 12 (middle row) and 500 µm (lowerrow). The dynamic range is chosen to scale from the maximum intensity of dust emission (excluding the central source) down to the 10−6th fraction of it.Labels are given in parsec.

Figure 4. Ratio of intensity maps at a wavelength of 10 and 20 µm – indicative of dust temperature – of model ER20 at a time 2.0 Myr after the centralradiation has been switched on, where brighter colours correspond to higher temperatures. The panels correspond to inclination angles of 30◦, 60◦ and 90◦

(from left to right). The dynamic range is chosen to scale from the maximum ratio down to the 10−3rd fraction of it. Labels are given in parsec.

appearance of the dust configuration. An example can be seen inthe middle panel.

4.3 Time-resolved spectral energy distributions

Fig. 6 shows a compilation of SEDs of all models and time stepsdiscussed in this paper. The different panels refer to the three Ed-dington ratios taken into account: from 1 per cent (upper panel,ER01) to 10 per cent (middle panel, ER10) and 20 per cent (lowerpanel, ER20). Inclinations are colour coded (black solid – i = 0◦,face-on; blue dotted – i = 30◦; green dashed – 60◦; red dash–dotted– 90◦, edge-on), where the thin lines in light colour refer to theindividual SEDs from the time series and thick lines show the SEDsaveraged over the available time steps. In general two bumps arevisible: (i) the ‘big blue bump’ – peaking at roughly 0.1 µm – and(ii) the ‘IR bump’ – peaking at around 10 µm. The first representsthe SED of the central source, which acts as the only source ofenergy in our simulations (see Section 3 and fig. 3 b in Schart-mann et al. 2005 for details). It is partially absorbed due to dust on

the line of sight and slightly altered by scattering off dust grains.Heating the surrounding dust to temperatures close to the subli-mation temperature results in re-emission of the grains in the IRwavelength regime, visible in the form of the ‘IR bump’. Overlaidover the two bumps are spectral features: most prominent and alsomost important for our analysis is the 10 µm feature (peaking moreaccurately at 9.7µm, see discussion in Section 5.3), arising dueto stretching modes in silicate tetrahedra (Si–O stretching modes).Only slightly visible is the second resonance of silicate dust, namelythe 18.5 µm feature (attributed to O–Si–O bending modes). The verypronounced absorption feature at around 0.08 µm has contributionsboth from the small graphite grains and small silicate grains. At0.2175 µm, graphite shows a significant feature. Given the temper-ature constraints at these wavelengths, the latter only show up asabsorption features against the background of the flux from the cen-tral source SED. As the dust extinction properties and the centralsource SED peak at roughly 0.1 µm, dust gets heated very efficientlyto temperatures up to the sublimation temperature (approximately1500 K for graphite grains and 1000 K for silicate grains). When

MNRAS 445, 3878–3891 (2014)

at University of Tokyo Library on M

arch 11, 2015http://m

nras.oxfordjournals.org/D

ownloaded from

• Indeed, MIR polar elongation was reproduced

• IR SED is well-explained, too.

• Consistent with observations

(morphological/photometric manner)

12μm

30 pc

6

Out Study: Multi-phase circumnuclear obscuring structures studied with ALMA

• Atomic vs molecular gas dynamics? 
→ geometrically thicker structure in an atomic disk than a molecular disk?

• Geometrical thickness due to outflows?

7ALMA Cycle 4 Observations (Band 7 + 8)

• Compton-thick AGN (Arevalo+2014)
- NH ~(6-10) × 1024 cm-2
- L2-10keV ~ (2-5) × 1042 erg/s

• Low SFR (i.e., weak SN-feedback) 
→ Test the fountain scheme

• High resolution CO(3-2) + [CI](1-0) in ALMA Cycle 4 (PI: T.Izumi)
→ molecular + atomic structures and their dynamical differences

• θ ~ 5 pc (CO) & ~15 pc (CI)

66 S. J. Curran, B. S. Koribalski and I. Bains

Figure 1. High-resolution H I moment maps of the Circinus galaxy. Left: the H I distribution; the contour levels are 0.4, 1, 2, 3, 4, 6, 8, 10, 12, 18 and20 Jy km s−1 beam−1. Right: the mean H I velocity field; the contour levels range from 300 (NE) to 570 km s−1 (SW) in steps of 10 km s−1. The synthesizedbeam (124 × 107 arcsec2) is shown in the bottom left-hand corner.

Figure 2. Low-resolution H I moment maps of the Circinus galaxy as obtained from the deep Parkes H I survey of the Zone of Avoidance (Juraszek et al.2000; Henning et al., in preparation). Left: the H I distribution; the contour levels are (0.5, 1, 2, 5, 10, 20, 30, 40, 50, 60 and 66) × 7 Jy km s−1 beam−1, where7 Jy km s−1 beam−1 correspond to an H I column density of 1019 cm−2. The contour levels were chosen to match those by Freeman et al. (1977, their fig. 6)above their detection limit of NH I = 5 × 1019 cm−2; our detection limit is a factor of ∼10 lower. Right: the mean H I velocity field (masked at NH I = 5 ×1018 cm−2); the contour levels range from 324 (NE) to 534 km s−1 (SW) in steps of 10 km s−1. The cross marks the centre of the Circinus galaxy as given inTable 1. The gridded beam of 15.5 arcmin is shown in the bottom left-hand corner.

C⃝ 2008 The Authors. Journal compilation C⃝ 2008 RAS, MNRAS 389, 63–74Downloaded from https://academic.oup.com/mnras/article-abstract/389/1/63/992361by gueston 20 February 2018

HI 21cm

Curran et al. 2008,

MNRAS, 389, 63

18 kpc

The Circinus Galaxy (4.2 Mpc)

8Molecular & Atomic gas distributions

• CND (D ~70 pc) + Spirals

• MH2 ~ 3×106 Msun (CND)


- c.f., MBH ~ 2×106 Msun

• NH2 ~ 5×1023 cm-2 (AGN)
- beam-averaged
- NH (Xray) ~ 6×1024 cm-2

• CND will be a significant nuclear obscurer!

12CO(3-2)

CND

+

Spiral arm

Izumi et al. 2018c, to be submitted

9

[CI](1-0)

+

Molecular & Atomic gas distributionsIzumi et al. 2018c, to be submitted

• Similar 2D distribution to the CO(3-2)

• Tracing the same volume???

• But we don’t know 3D structures!

10Decomposition of Cold Gas Dynamics

• Global motion is dominated by rotation

• We decomposed the dynamics with tilted-ring models 


- Vrot, σ, inclination, P.A.

16 T. Izumi et al.

(a) (b) (c)

(d) (e) (f)

Figure 12. (Top row) (a) Intensity-weighted mean velocity map of CO(3–2) in the central 10′′ × 10′′ (200 pc × 200 pc) boxof the Circinus galaxy. The central plus marks the AGN position, which applies to the other panels too. Contours indicate theline-of-sight velocity of 300–600 km s−1 in steps of 15 km s−1. (b) Same contours as in (a), but are superposed on the CO(3–2)integrated intensity map. (c) Intensity-weighted velocity dispersion (Gaussian sigma) map of CO(3–2) shown by the contours(10–60 km s−1 in steps of 10 km s−1), superposed on the CO(3–2) integrated intensity map. (Bottom row) (d) Intensity-weightedmean velocity map of [C I](1–0) in the same region and with the same contour levels as in (a). (e) Superposition of (d) in contourson the [C I](1–0) integrated intensity map. (f) Intensity-weighted velocity dispersion (Gaussian sigma) map of [C I](1–0) shownin contours (same levels as in (c)) overlaid on the [C I](1–0) integrated intensity map. The systemic velocity (445.8 km s−1)estimated from our single Gaussian fit to the CO(3–2) spectrum (Figure 10) is highlighted by the red line in both (a) and (d).The integrated intensities are shown in the unit of Jy beam−1 km s−1.

are formed in situ within the AGN outflow, as recently912indicated by the hydrodynamic plus chemical simula-913tion (Richings & Faucher-Giguère 2018). Further high914resolution multi-phase gas observations of AGNs with915molecular outflows are desired to reveal their physical916and chemical origin.917

5.2. Tilted-ring models918

In order to extract basic beam-deconvolved dynami-919cal information, particularly the rotation velocity (Vrot)920and dispersion (σdisp), we fitted titled rings to the921observed velocity structures with the 3DBarolo code922(Di Teodoro & Fraternali 2015). The main parame-923ters are dynamical center, systemic velocity (Vsys), Vrot,924σdisp, galactic inclination (i), and P.A., all of which can925be varied for each ring. Here we fix the dynamical center926to the AGN position and Vsys to 446 km s−1 based on927our Gaussian fits to the observed nuclear spectra (Table9285): Vrot, σdisp, i, and P.A. are thus the free parameters929in our study. Note that the code itself will not account930

non-circular motions, those can be revealed by subtract-931ing the modelled rotation map from the observed one932(e.g., Salak et al. 2016).933Through our experiments, however, we noticed that934

the results of the fitting are sensitive to the spatial reso-935lution of the data. Given that the spatial resolutions are936significantly different between the CO(3–2) cube and the937[C I](1–0) cube, we first present the results with using938the full-resolution CO(3–2) cube in § 5.2.1 to describe939the details of the gas dynamics as much as possible, then940do with using the beam-matched CO(3–2) and [C I](1–9410) cubes in § 5.2.2. The latter results are used to delin-942eate qualitative difference of molecular and atomic gas943dynamics and/or geometry.944

5.2.1. The CO(3–2) full-resolution data945

The full-resolution CO(3–2) cube (θ = 0′′.29× 0′′.24,946dV = 10 km s−1) is used here. We modelled 50 con-947centric ring with the separation ∆r = 0′′.15, which is948roughly the half size of the major axis of its synthe-949

CO(3-2) Izumi et al. 2018c, to be submitted

11CI vs CO: Multi-phase Torus Dynamics18 T. Izumi et al.

(a)

(b)

Figure 17. (a) Radial profiles of the rotation velocity (Vrot;circles) and the velocity dispersion (σdisp; diamonds) derivedwith the CO(3–2) (blue) and [C I](1–0) (red) emission lines,in the central r ! 100 pc of the Circunus galaxy. (b) Radialprofiles of the σdisp/Vrot ratios traced by the CO(3–2) andthe [C I](1–0) lines, at the same region as in (a). The uv-ranges and the beam sizes are matched between the CO(3–2)cube and the [C I](1–0) cube here.

to puff-up the disk) should be weak given the low SFR927(∼ 0.1 M⊙ yr−1; Esquej et al. 2014).928Meanwhile, the discrepancy of the σdisp/Vrot ratios at929

r ! 10 pc (∼ 0.15 in CO(3–2) and ∼ 0.4 in [C I](1–0) ,930respectively) is more evident, which indicates the exis-931tence of multi-phase dynamical structures in the torus.932Note that it is hard to distinguish coherent outflow mo-933tion that can widen the line profile from isotropic tur-934bulent motion due to failed winds, based simply on the935observed σdisp that mixes up these motions. However,936as the outflows are preferentially seen at ionized and/or937atomic gas at the very nuclear region in our model938(Wada et al. 2016), we expect that such outflows are939the prime source of the geometrically thicker nature of940the C I disk than the CO disk at this very central re-941gion around the AGN. Indeed, we found that adding942two more components (i.e., triple Gaussians) can fit the943observed nuclear [C I](1–0) spectrum significantly better944

Figure 18. Line profile of the [C I](1–0) measured atthe AGN position (with the single 0′′.71 × 0′′.66 beam) ofthe Circinus galaxy. Three-component Gaussian fit was per-formed: each component is indicated by the black-dashedlines, and the total model is shown by the red-solid line, re-spectively. The derived line properties are summarized inTable 6.

Table 6. Results of the three-component Gaussian fit to the[C I](1–0) spectrum

Peak Centroid FWHM

(mJy beam−1) (km s−1) (km s−1)

Central 1107.9 ± 1.6 453.7 ± 5.6 97.7 ± 0.6Blueshifted 466.4 ± 5.6 379.1 ± 0.2 56.7 ± 0.5Redshifted 365.9 ± 4.4 512.3 ± 0.1 28.3 ± 0.4

Note—These fits were performed for the spectra at theAGN position, measured with the single synthesized beam.

Results of the single Gaussian fit are listed in Table 5.

(reduced χ2 = 656.2 for the single Gaussian, but 64.3 for945the triple Gaussians) 8 as shown in Figure 18 (see also946Table 6). The central Gaussian component has almost947the same centroid velocity and FWHM as the CO(3–9482) spectrum (Table 5), suggesting that this traces the949same region as the CO(3–2) line does. We also found950that the blueshifted and redshifted components appear951almost symmetrically with respect to the central com-952ponent, with the velocity offset of ∼ 75 km s−1 (blue)953and ∼ 60 km s−1 (red), respectively: this is sugges-954tive of a coherent atomic gas motion around the AGN.955These results therefore support the view that the diffuse956atomic gas is distributed in a geometrically thicker vol-957ume than the dense molecular gas at around the AGN,958due to atomic outflows, as is expected in our multi-phase959dynamic torus model. As this argument is a rather qual-960itative, however, we will perform a more quantitative961

8 The goodness of the fit was evaluated at the velocity range of250 to 650 km s−1.

Velo

city

(km

/s)

40

80

120

160

Radius (pc)0 20 40 60 80 100

σ/V r

ot (~

H/R

)

Radius (pc)0

0.1

0.2

0.3

0.4

0.5

0 20 40 60 80

Blue = CO(3-2) Red = [CI](1-0)

Rotation

Dispersion

0

• Geometrically thin mol. & atomic disks at r > 20 pc (c.f., low-SFR in Circinus)

• Geometrically thicker atomic disk than the dense mol. disk at r < 10 pc! 
→ Multi-phase geometrical structures!

Izumi et al. 2018c, to be submitted

Circumnuclear Multi-phase Gas in the Circinus Galaxy 13

Table 4. The [C I](1–0)/CO(3–2) line ratio

AGN CND 10′′ box

Jy km s−1 unit 1.80 ± 0.25 1.03 ± 0.15 0.97 ± 0.14K km s−1 unit 0.89 ± 0.13 0.51 ± 0.07 0.48 ± 0.07

Note—The “AGN” indicates that the ratio is measured atthe exact AGN location with the single

[C I](1–0) synthesized beam (0′′.71× 0′′.66). The CNDrefers to the 3′′.62× 1′′.66 (P.A. = 212◦) region defined inthe CO(3–2) map (§ 4.1.2). The 10′′ box contains both the

CND and the spiral arms. The 10% absolute fluxuncertainties are included here.

CND + spiral arms), the [C I](1–0) emission clearly be-695comes brighter than the CO(3–2) emission at the AGN696position. The relevant RCI/CO measured at several spa-697tial scales are listed in Table 4: this high RCI/CO at698the AGN position is uncommon in star-forming galax-699ies, suggesting an efficient CO dissociation due to hard700X-ray irradiation from the AGN (XDR, Maloney et al.7011996; Meijerink & Spaans 2005).702Based on the one-zone XDR model of Maloney et al.703

(1996), we estimated the X-ray energy deposition rate704per particle (HX/n) with705

HX ∼ 7× 10−22L44r−22 N−122 erg s

−1, (2)

where L44 is the 1–100 keV X-ray luminosity in the unit706of 1044 erg s−1, r2 is the distance from the AGN to the707point of our interest in the unit of 100 pc, and N22 is708the X-ray attenuating column density in the unit of 1022709cm−2, respectively. Here we adopted L44 = 0.13 (with710the photon index = 1.31 and the cutoff energy = 160711keV) from Arévalo et al. (2014), as well as nH2 = 10

5712

cm−3 as a typical value in the molecular part of CNDs of713Seyfert galaxies (e.g., Izumi et al. 2013; Viti et al. 2014).714Then, we estimated the HX/n at r = 7 pc from the cen-715ter (i.e., half-major axis of the [C I](1–0) synthesized716beam) as log(HX/n) = −27.3. According to the mod-717els in Maloney et al. (1996), this rate is almost exactly718the one at which the fractional abundance of C I (XCI)719becomes equal to that of CO (XCO). At the regions720with log(HX/n) > −27.3, we can expect XCI/XCO > 1.721Thus, we argue that the molecular dissociation due to722X-ray irradiation is significant at the close vicinity of723this Circinus AGN, which can cause the high RCI/CO724there.725However, the actual 3-dimensional geometrical struc-726

ture of the XDR needs careful consideration: it does727not necessarily form a simple one-zone structure. In-728deed, single Gaussian fits to the observed CO(3–2) and729[C I](1–0) spectra measured at the AGN position (Figure73011b and Table 5) revealed that the [C I](1–0) spectrum731

(a)

(b)

Figure 11. Line profiles of the [C I](1–0) and the CO(3–2)measured at (a) the inner 10′′ × 10′′ box and (b) the AGNlocation (with the single 0′′.71×0′′.66 beam) of the Circinusgalaxy. The uv-range and the beam size of the CO(3–2) dataare matched to those of the [C I](1–0) data. The multiplehorn-line profile in (a) is due to the spiral arms. Results ofsingle Gaussian fits to the CO(3–2) and [C I](1–0) spectraat the AGN position (indicated by the dashed line and dot-dashed line, respectively) are summarized in Table 5.

Table 5. Results of the single Gaussian fit

Peak Centroid FWHM

(mJy beam−1) (km s−1) (km s−1)

CO(3–2) 807.9 ± 0.9 445.8 ± 0.1 100.3 ± 0.1[C I](1–0) 1115.1 ± 4.6 446.1 ± 0.1 134.3 ± 0.2

Note—These fits were performed for the spectra at theAGN position, measured with the single[C I](1–0) synthesized beam (Figure 11b).

has a wider FWHM than the CO(3–2) spectrum, as well732as deviated components from the single Gaussian pro-733file at about ±70 km s−1 to the systemic velocity (446734km s−1). Both of these imply that the [C I](1–0) emis-735sion also traces a different gas volume that the CO(3–2)736

Velocity (km/s)250 300 350 400 450 500 550 600 650

Flux

den

sity

(mJy

/bea

m)

0

200

400

600

800

1000

1200

[CI](1-0)

CO(3-2)

Nuclear-scale (15 pc) spectrum

12Then, what makes the geometrical thickness?

• Nuclear [CI](1-0) spectrum shows a larger FWHM than CO(3-2).

• Multi-Gaussians yielded a better fit
→ Almost symmetric emergence of the additional blue- and red-components

• Coherent motion like outflows??
- consistent with the fountain model
- will fall back to the disk (failed wind)

FWHM ~ 240 km/s

FWHM ~ 320 km/s

18 T. Izumi et al.

(a)

(b)

Figure 17. (a) Radial profiles of the rotation velocity (Vrot;circles) and the velocity dispersion (σdisp; diamonds) derivedwith the CO(3–2) (blue) and [C I](1–0) (red) emission lines,in the central r ! 100 pc of the Circunus galaxy. (b) Radialprofiles of the σdisp/Vrot ratios traced by the CO(3–2) andthe [C I](1–0) lines, at the same region as in (a). The uv-ranges and the beam sizes are matched between the CO(3–2)cube and the [C I](1–0) cube here.

to puff-up the disk) should be weak given the low SFR927(∼ 0.1 M⊙ yr−1; Esquej et al. 2014).928Meanwhile, the discrepancy of the σdisp/Vrot ratios at929

r ! 10 pc (∼ 0.15 in CO(3–2) and ∼ 0.4 in [C I](1–0) ,930respectively) is more evident, which indicates the exis-931tence of multi-phase dynamical structures in the torus.932Note that it is hard to distinguish coherent outflow mo-933tion that can widen the line profile from isotropic tur-934bulent motion due to failed winds, based simply on the935observed σdisp that mixes up these motions. However,936as the outflows are preferentially seen at ionized and/or937atomic gas at the very nuclear region in our model938(Wada et al. 2016), we expect that such outflows are939the prime source of the geometrically thicker nature of940the C I disk than the CO disk at this very central re-941gion around the AGN. Indeed, we found that adding942two more components (i.e., triple Gaussians) can fit the943observed nuclear [C I](1–0) spectrum significantly better944

Figure 18. Line profile of the [C I](1–0) measured atthe AGN position (with the single 0′′.71 × 0′′.66 beam) ofthe Circinus galaxy. Three-component Gaussian fit was per-formed: each component is indicated by the black-dashedlines, and the total model is shown by the red-solid line, re-spectively. The derived line properties are summarized inTable 6.

Table 6. Results of the three-component Gaussian fit to the[C I](1–0) spectrum

Peak Centroid FWHM

(mJy beam−1) (km s−1) (km s−1)

Central 1107.9 ± 1.6 453.7 ± 5.6 97.7 ± 0.6Blueshifted 466.4 ± 5.6 379.1 ± 0.2 56.7 ± 0.5Redshifted 365.9 ± 4.4 512.3 ± 0.1 28.3 ± 0.4

Note—These fits were performed for the spectra at theAGN position, measured with the single synthesized beam.

Results of the single Gaussian fit are listed in Table 5.

(reduced χ2 = 656.2 for the single Gaussian, but 64.3 for945the triple Gaussians) 8 as shown in Figure 18 (see also946Table 6). The central Gaussian component has almost947the same centroid velocity and FWHM as the CO(3–9482) spectrum (Table 5), suggesting that this traces the949same region as the CO(3–2) line does. We also found950that the blueshifted and redshifted components appear951almost symmetrically with respect to the central com-952ponent, with the velocity offset of ∼ 75 km s−1 (blue)953and ∼ 60 km s−1 (red), respectively: this is sugges-954tive of a coherent atomic gas motion around the AGN.955These results therefore support the view that the diffuse956atomic gas is distributed in a geometrically thicker vol-957ume than the dense molecular gas at around the AGN,958due to atomic outflows, as is expected in our multi-phase959dynamic torus model. As this argument is a rather qual-960itative, however, we will perform a more quantitative961

8 The goodness of the fit was evaluated at the velocity range of250 to 650 km s−1.

[CI](1-0)

13

• CO(3-2) → mid-plane of the CND/torus

• [CI](1-0) → mid-plane + puffed-up component due to outflows

Quantitative comparison with our modelCO(3-2): TB [CI](1-0): TB

Izumi et al. 2018c, to be submitted(inclination = 75°)

• MBH, Eddington ratio, CND-scale Mgas: Circinus-like values

• Hydrodynamic simulation + XDR chemistry (Wada+16) + rad. transfer

14

• Indeed, we found different line profiles for the simulated CO(3-2) and [CI](1-0)

• Triple-Gaussians can well fit the profile → outflow components stand out

• Good consistency between ALMA obs. and our simulation
→ Support the fountain scheme, where atomic outflows play the key role in determining the “torus” geometry.

Quantitative comparison with our model

Izumi et al. 2018c, to be submitted

Summary15

• Now we are testing the multi-phase dynamic torus model with ALMA: ~5-15 pc resolution CO(3-2) and [CI](1-0) data are in hand.

• Gas (and dust) distributions = CND/torus + spiral arms.

• The CND can provide a significant nuclear obscuration (as a torus).

• CI (atom) gas is in a geometrically thicker volume than the CO

(molecule) gas at r < 10 pc. 
→ multi-phase gas dynamics in the torus!

• The thickness of the atomic disk would be due to nuclear atomic outflows, as expected in our fountain model 
→ Physical origin of the torus!?