image/svg+xml
3D RT modelling of submm emission
❄ to get a model for the cloud structure
and to estimate the submm τ
❄ residuals close to the noise level
LDN 1642 has already been studied in several
papers (e.g. Malinen et al. 2014,2016)
❄
dust opacity and spectral index
❄
density vs. magnetic field structure
demo...
Modelling dust and line emission
in interstellar clouds
Dust continuum:
- dust extinction, emission, and scattering in LDN 1642
Line emission:
- LOC program for line radiative transfer (RT)
Other tools:
- Extracting filaments
Filament extraction
❄
many methods already exist
❄
template matching was proposed as one
possible tool (also) for this (Juvela 2016)
❄
we have updated this, precising the
definition
of a filament
(1) measure the image standard deviation at the
selected scale
(2) for each pixel, calculate the local
probability
of a filament, for the best position angle
(3)
trace
the continous structures, updating
the probabilities based on the filament
continuing in either direction
-14.2°
-14.3°
Dec (J2000)
B2
B1
B3
a
J
0
50
100
150
200
250
b
H
0
50
100
150
200
250
300
350
I (kJy/sr)
68.8°
68.7°
-14.2°
-14.3°
RA (J2000)
Dec (J2000)
c
K
0
25
50
75
100
125
150
175
200
68.8°
68.7°
RA (J2000)
d
3.4m
0
5
10
15
20
25
I (kJy/sr)
69.0°
68.8°
68.6°
68.4°
-14.0°
-14.2°
-14.4°
-14.6°
RA (J2000)
Dec (J2000)
e
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
N(H
2
) (cm
2
)
1e21
4
h
36
m
00
s
35
m
12
s
34
m
24
s
-14°00'
10'
20'
30'
RA (J2000)
Dec (J2000)
B1
B2
B3
a
N
(H
2
)(10
21
cm
2
)
0.2pc
0
1
2
3
4
4
h
36
m
00
s
35
m
12
s
34
m
24
s
-14°00'
10'
20'
30'
RA (J2000)
B1
B2
B3
b
T
dust
(K)
0.2pc
13
14
15
16
17
18
Herschel ⇒ τ(250µm), NIR data ⇒ τ(J)
❄ τ(250µm)/τ(J) ~ 10
-3
,
❄ 2-3 times the diffuse-medium value,
as in Juvela et al. (2015)
NIR-MIR scattered light
❄ HAWK-I J-K, WISE 3.4µm
❄ VLT observations in ON-OFF mode
to measure surface brightness
❄ B1 too bright for observations:
three pointings around the star
❄ J is ok; in H and especially in K,
frames do not match perfectly
a
I
(160m)(MJysr
1
)
10
20
30
40
50
60
e
I
(160m)(MJysr
1
)
10
20
30
40
50
60
i
k
BG
=0.98
(0.58±12.04)%
I
(160m)(MJysr
1
)
30
20
10
0
10
20
30
b
I
(250m)(MJysr
1
)
20
40
60
80
f
I
(250m)(MJysr
1
)
20
40
60
80
j
(3.17±3.73)%
I
(250m)(MJysr
1
)
30
20
10
0
10
20
30
c
I
(350m)(MJysr
1
)
10
20
30
40
50
g
I
(350m)(MJysr
1
)
10
20
30
40
50
k
(0.01±0.27)%
I
(350m)(MJysr
1
)
30
20
10
0
10
20
30
d
I
(500m)(MJysr
1
)
5
10
15
20
25
h
I
(500m)(MJysr
1
)
5
10
15
20
25
l
(5.74±3.53)%
I
(500m)(MJysr
1
)
30
20
10
0
10
20
30
-36.2°
-36.4°
-36.6°
-36.8°
Galactic Latitude
I
(R)(kJysr
1
)
a
R
0
10
20
30
40
50
60
I
(J)(kJysr
1
)
b
J
0
20
40
60
80
100
-36.2°
-36.4°
-36.6°
-36.8°
Galactic Latitude
I
(H)(kJysr
1
)
c
H
0
20
40
60
80
I
(K)(kJysr
1
)
d
K
0
10
20
30
40
50
60
211.2°
210.9°
210.6°
-36.2°
-36.4°
-36.6°
-36.8°
Galactic Longitude
Galactic Latitude
I
(3.4m)(kJysr
1
)
e
3.4m
0.0
2.5
5.0
7.5
10.0
12.5
15.0
17.5
211.2°
210.9°
210.6°
Galactic Longitude
I
(4.6m)(kJysr
1
)
f
4.6m
0
1
2
3
4
Predicted scattered light
❄ from R band to 4.6 µm, here for
the Jones et al. (2013) dust model
❄ ...but scattering is only part of the
net surface surface brightness
-14.2°
-14.3°
Dec (J2000)
a
J(OBS)
0
50
100
150
200
250
b
H(OBS)
0
50
100
150
200
250
300
350
c
K(OBS)
0
25
50
75
100
125
150
175
200
d
3.4m(OBS)
I
(kJysr
1
)
0
5
10
15
20
25
68.8°
68.7°
-14.2°
-14.3°
RA (J2000)
Dec (J2000)
e
J(MOD)
0
10
20
30
40
50
68.8°
68.7°
RA (J2000)
f
H(MOD)
0
10
20
30
40
50
60
70
68.8°
68.7°
RA (J2000)
g
K(MOD)
0
10
20
30
40
50
68.8°
68.7°
RA (J2000)
h
3.4m(MOD)
I
(kJysr
1
)
0
2
4
6
8
10
12
10
2
10
1
10
0
10
1
10
2
10
3
I
(MJysr
1
)
a
B2:J
n
0
R
0
=3.25
e
+04cm
3
=0.0107pc
=0.83
k
J
=7.16
J
=1.84
b
B2:H
k
H
=1.75
H
=1.07
c
B2:K
k
K
=0.43
K
=0.57
0
20
40
60
r
(arcsec)
10
2
10
1
10
0
10
1
10
2
10
3
I
(MJysr
1
)
d
B3:J
n
0
R
0
=1.78
e
+04cm
3
=0.0089pc
=0.10
k
J
=1.57
J
=2.17
0
20
40
60
r
(arcsec)
e
B3:H
k
H
=1.53
H
=1.26
0
20
40
60
r
(arcsec)
f
B3:K
k
K
=0.54
K
=0.67
L1624 is a rare star-forming high-latitude cloud
Would the low scattering be solved
by using a different dust model?
Ysard et al. (2016) modelled scattering
in Taurus successfully using the THEMIS
model that includes evolution
CM -> CMM -> AMMk -> AMMI
* from bare amorphous carbon and
silicate grains to core-mantle structure
and eventually ice layers
The consistency of submm emission was
not tested
The problem seems to be the low
τ(250µm)/τ(J) in all dust models ⇒
submm models overestimate extinction
One arcmin environment of the B2 and B3
sources
can
be modelled as pure scattering
❄
no
significant
emission
in the NIR
❄ consistent with flat density and only the
r
-2
drop in flux density
❄ B3 obviously also part of the L1642 cloud
Observations:
❄ Herschel: 100-500 µm thermal dust emission
❄ VLT/Hawk-I: J, H, and K - extinction, scattering
LDN 1642:
2 million cells
MHD models:
up to ~800
million cells
Discrepancy in NIR scattering
can be fixed by
(1) reducing the cloud optical
depth by hand
or
(2) assuming dust properties
with
τ(250µ
m
)/τ(
J
)
similar to
the measured (~0.001)
Dust emission and scattering in L1642
❄
Combination of dust emission and scattering
gives strong constraints on dust models
❄
None of the current models can explain these
self-consistently
❄
NIR surface-brightness observations are still
challenging - especially mosaics of multi-array
observations
16.0
16.5
17.0
17.5
18.0
log
10
(
r
/cm)
0
5
10
15
20
25
T
ex
(10)(K)
a
Model2a,
J
=10
16.0
16.5
17.0
17.5
18.0
log
10
(
r
/cm)
0
2
4
6
8
10
T
ex
(43)(K)
b
Model2a,
J
=43
16.0
16.5
17.0
17.5
18.0
log
10
(
r
/cm)
0
5
10
15
20
25
T
ex
(10)(K)
c
Model2b,
J
=10
16.0
16.5
17.0
17.5
18.0
log
10
(
r
/cm)
2
4
6
8
10
12
T
ex
(43)(K)
d
Model2b,
J
=43
4
6
8
10
T
ex
(K)
a
Cppsimu
LOC/1D
LOC/3D
LOC/3D, 3 levels
0.00
0.02
0.04
0.06
0.08
0.10
T
(pc)
0.95
1.00
1.05
r
(
T
ex
)
0
1
2
3
4
T
A
(K)
b
0.25
0.00
3.9
4.0
4.1
20
10
0
10
20
v
(kms
1
)
0.2
0.0
0.2
(
T
A
)(K)
20
50
100
200
500
N
10
1
10
2
10
3
10
4
Time(s)
Cppsimu
LOC/CPU
LOC/GPU
LOC/GPU/Octree routine
10
6
10
7
Cells
10
0
10
1
10
2
10
3
Time(s)
a
Rootgrid64
3
N
L
=1
N
L
=2
N
L
=3
N
L
=4
10
6
10
7
Cells
10
0
10
1
10
2
10
3
Time(s)
b
Rootgrid128
3
10
2
10
3
10
4
4
6
8
10
12
14
T
ex
(
12
CO,J=10)(K)
a
N
L
=3
1
2
3
4
5
10
2
10
3
10
4
n
(cm
3
)
0.2
0.1
0.0
T
ex
(
12
CO,
J
=10)(
K
)
10
2
10
3
10
4
b
N
L
=4
10
2
10
3
10
4
n
(cm
3
)
LOC - Line transfer with OpenCL
The 1D version tested against
Monte Carlo programme Cppsimu
and the benchmark problems in
van Zaddelhof et al. (2002)
LOC is well within the results from
different RT codes.
Largest differences are caused by
interpretations of the model setp up,
not necessarily by the RT methods
3D LOC was compared to Cppsimu and to spherical models discretised to octree grids
Different tracers show different structure
Signifcant difference also between LTE and
non-LTE line maps
Left: power spectra calculated for
- true column density
- simulated maps of dust emission
- simulated molecular line maps (line area),
in LTE and non-LTE cases
GPUs provide a
speed-up
of up to
~10
over a single CPU core
(still far from the theoretical peak
performance)
Run times scale very nearly linearly
with the number of cells
❄
SOC for continuum RT calculations published last year
❄
LOC to enable the use of GPUs for line RT
- a specialised version for 1D models
- 3D version using octree grids and non-Monte-Carlo ray tracing
- ALI convergence acceleration, optional line overlap,
(or hyperfine LTE components), coupling to dust calculations
Tests with MHD simulation data for a (250 pc)
3
box of ISM
- a root grid of 256
3
+ three refinements = effective 2048
3
10
0
10
x
(
pc
)
15
10
5
0
5
10
15
y
(
pc
)
a
4
3
2
1
0
1
W
(LTE)(Kkms
1
)
10
0
10
x
(
pc
)
15
10
5
0
5
10
15
y
(
pc
)
b
4
3
2
1
0
1
W
(nonLTE)(Kkms
1
)
10
0
10
x
(
pc
)
15
10
5
0
5
10
15
y
(
pc
)
c
0.2
0.4
0.6
0.8
1.0
1.2
1.4
W
(nonLTE)/
W
(LTE)
3
2
1
0
1
log
10
W
(Kkms
1
)
0
5000
10000
15000
20000
25000
30000
Pixels
d
LTE
non-LTE
NIR-MIR surface brightness:
observations
vs.
models
❄
NIR
optical depth
of submm models > NIR measurement
❄
NIR
surface brightness
observed > modelled
LOC for line radiative transfer
❄ SOC and LOC enable RT calculations with
models with hundreds of millions of cells
- SOC has been used with up to 5×10
8
cells,
LOC > 10
8
cells (laptop with 16GB memory!)
❄ with GPU-acceleration and octree grids, LOC
is ready for very high-resolution line modelling
SOC RT models
1
Title frame
L1642 allsky
L1642 allsky
L1642 allsky
L1642 allsky
L1642 allsky
L1642 old Herschel
L1642 tau ratio
L1642 NIR observations
L1642 submm model
L1642 scattering model
L1642 scattering obs + model
L1642 fixed model
L1642 fixed model
SOC cell numbers
SOC cell numbers
SOC conclusions
Full
LOC intro
LOC 1d
LOC 1d
LOC 3d
LOC 3d
LOC timings
SOC+LOC conclusions
Filaments
Full screen