traphic( transportof photons(in cones
TRANSCRIPT
![Page 1: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/1.jpg)
TRAPHIC TRAnsport of PHotons In Cones
Current TRAPHIC team:
Andreas Pawlik (MPA) Joop Schaye (Leiden) Ali RahmaC (Leiden) Milan Raicevic (Leiden) Myoungwon Jeon (AusCn)
Collaborators:
Volker Bromm (AusCn) Milos Milosavljevic (AusCn) Claudio Dalla Vecchia (MPE)
Cosmological RadiaCve Transfer Code Comparison Project IV AusCn, Dec. 2012
![Page 2: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/2.jpg)
Outline
• Cosmological RadiaCve Transfer • TRAPHIC • Tests • ApplicaCons • Summary
Andreas Pawlik (MPA)
![Page 3: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/3.jpg)
Cosmological Ionizing RT
• Large dynamic range – Resolve dwarf galaxies with size ~1 kpc in volumes of size ~100 comoving Mpc
• Many ionizing sources – tens of thousands to millions
• Accurate knowledge of gas distribuCon (clumping factor) – hydrodynamical simulaCons including feedback
Andreas Pawlik (MPA)
![Page 4: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/4.jpg)
TRAPHIC – TRAnsport of PHotons In Cones
(AP & J. Schaye 2008, 2010)
• SpaCally adapCve transport in SPH • ComputaCon Cme independent of the
number of ionizing sources • Hydrodynamically coupled to Gadget Andreas Pawlik
(MPA)
![Page 5: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/5.jpg)
Andreas Pawlik (MPA)
SPH
![Page 6: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/6.jpg)
Andreas Pawlik (MPA)
SPH -‐ AdapCvity
Low density High density
AdapCve resoluCon element
![Page 7: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/7.jpg)
Andreas Pawlik (MPA)
TRAPHIC in SPH -‐ AdapCvity
![Page 8: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/8.jpg)
ParCcle-‐to-‐neighbor transport is biased towards center of mass
center of mass
net transport
α
Andreas Pawlik (MPA)
![Page 9: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/9.jpg)
A weighted transport does not remove the intrinsic bias (AP & J. Schaye 2008)
αcos
Random distribuCon of 64 neighbours
α
- - Equal weights __ Solid angle weights
Andreas Pawlik (MPA)
center of mass
net transport
![Page 10: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/10.jpg)
Andreas Pawlik (MPA)
TRAPHIC – Emission
Virtual ParCcle: create new direcCons as needed
Randomly oriented emission cones
here: Nngb = 4 here: NEC = 4
![Page 11: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/11.jpg)
Andreas Pawlik (MPA)
TRAPHIC – Transmission
here: Nngb = 4 here: NEC = 4
![Page 12: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/12.jpg)
Andreas Pawlik (MPA)
TRAPHIC – Transmission
here: Nngb = 4 here: NEC = 4
![Page 13: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/13.jpg)
Andreas Pawlik (MPA)
TRAPHIC – Transmission
here: Nngb = 4 here: NEC = 4
![Page 14: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/14.jpg)
Andreas Pawlik (MPA)
TRAPHIC – Merging
Randomly oriented recepCon cones (here: NRC = 4) -‐> ComputaCon Cme scaling: independent of # sources
![Page 15: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/15.jpg)
Flow chart -‐ single RT Cme step
≥ c Δt?
Create ViPs
Remove ViPs
Update parCcle properCes
Propagate photons to neighbors
No
Yes
Full control of:
• DirecConal sampling • Angular resoluCon of the
transport • Angular resoluCon of the
merging • Speed of light ValidaCon by convergence: in the limit of high angular resoluCon, TRAPHIC becomes equivalent to a classical Monte Carlo code
Andreas Pawlik (MPA)
![Page 16: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/16.jpg)
TRAPHIC in GADGET 2008/2012 • SPH -‐ no addiConal grid (regular/unstructured) • Photon packet merging – computaConal cost independent
of number of sources • MPI parallel/Gadget2 • MulCfrequency transport • Helium • PhotoheaCng/cooling (including H2/HD) • RHD/Gadget3 (+ new dynamical Cme
stepping, Durier & Dalla Vecchia 2011) • X-‐ray secondary ionizaCon (Jeon et al.) • RecombinaCon radiaCon (Raicevic et al.) • UV background (RahmaC et al. 2012)
new
Andreas Pawlik (MPA)
![Page 17: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/17.jpg)
Tests
• RT code comparison tests (Iliev et al.) • Comparison with 1d code (“tesmraphic1d”) • Convergence tests
Andreas Pawlik (MPA)
TRAPHIC - multi-frequency and thermal coupling 19
Figure 10. Test 3. Neutral fraction in a slice through z = Lbox/2 at time t = 0.2 Myr. From left to right: traphic thin (assuming greyoptically thin photoheating rates), traphic thick (assuming grey optically thick photoheating rates), traphic (using three frequencybins), c2-ray, crash and ftte. Contours show neutral hydrogen fractions ηHI = 0.9, 0.5, log ηHI = −1,−3 and −5, from the outside in.The results obtained with traphic thick are in excellent agreement with those obtained with ftte. They are also in excellent agreementwith the results obtained with c2-ray in highly ionised regions, where the neutral fraction is unaffected by spectral hardening. Thesmall differences in the neutral fractions obtained with traphic thick, traphic thin and traphic are mostly due to differences in therecombination rate, caused by differences in the gas temperatures (see Fig. 11).
Figure 11. Test 3. Temperature in a slice through z = Lbox/2 at time t = 0.2 Myr. From left to right: traphic thin (assumingoptically thin photoheating rates), traphic thick (assuming optically thick photoheating rates), traphic (using three frequency bins),c2-ray, crash and ftte. Contours show temperatures log10(T/K) = 3, 4, 4.2, 4.4 and 4.6, from the outside in. Most of the morphologicaldifferences may be attributed to differences in the spectral hardening of the ionising radiation, with the multi-frequency codes traphic,c2-ray and crash yielding a substantial amount of pre-heating and the monochromatic (grey) codes traphic thin, traphic thick andftte yielding sharp transitions between the hot ionised and the cold neutral phases. The differences in the maximum gas temperaturesare mainly due to photoheating being computed in the optically thick limit (traphic thick, c2-ray, ftte), the optically thin limit(traphic thin) or using multiple frequency bins (traphic, crash).
number of frequency bins as small as possible. Using a smallnumber of frequency bins in the present test should thusgive results that more closely resemble the results of future,larger simulations.
Figs. 10-12 show our results. Fig. 10 shows images of theneutral fraction in slices through the centre of the simula-tion box at time t = 0.2 Myr (our conclusions also hold forother times). The individual panels show results obtainedwith traphic thin, traphic thick and traphic. For ref-erence, we also show the results obtained with other RTcodes for the same test problem as published in Iliev et al.(2006a). Neutral fraction contours are shown to facilitate thecomparison. While the simulation with crash treated thepresent problem by performing a multi-frequency computa-tion, the simulation with ftte, as our simulation traphic
thick, solved it in the grey approximation using opticallythick photoheating rates. Finally, c2
-ray employed a hybridmethod that treats the transport of radiation with multiplefrequency bins but computes photoheating rates in the grey(optically thick) approximation.
The differences in the neutral fractions are generallysmall. The simulations that employ photoheating rates inthe optically thick limit (ftte, c
2-ray, traphic thick)
yield smaller minimum neutral fractions than the simu-lations that compute photoheating rates in the opticallythin limit (traphic thin) or using multiple frequency bins(crash, traphic). This is the result of lower recombina-tion rates caused by the higher temperatures these simula-tions yield16 (see Fig. 11). The regions with low ionisation(ηHI > 0.5) found with traphic are slightly smaller thanthose found with c
2-ray and crash, which indicates that
three frequency bins are not sufficient for obtaining highlyaccurate multi-frequency solution. Still, the simulations withtraphic seem to capture the main effects (see the discussionon pre-heating below).
Fig. 11 shows images of the gas temperature in slices
16 As noted earlier, the main reason why crash finds significantlylarger neutral fractions may be an insufficient sampling of thephoton field, see, e.g., Fig. 2 in Maselli, Ferrara, & Ciardi 2003.
c© RAS, MNRAS 000, 1–23
![Page 18: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/18.jpg)
Test 4 – opCcally thin vs. opCcally thick heaCng rates (AP & Schaye 2010)
TRAPHIC - multi-frequency and thermal coupling 19
Figure 10. Test 3. Neutral fraction in a slice through z = Lbox/2 at time t = 0.2 Myr. From left to right: traphic thin (assuming greyoptically thin photoheating rates), traphic thick (assuming grey optically thick photoheating rates), traphic (using three frequencybins), c2-ray, crash and ftte. Contours show neutral hydrogen fractions ηHI = 0.9, 0.5, log ηHI = −1,−3 and −5, from the outside in.The results obtained with traphic thick are in excellent agreement with those obtained with ftte. They are also in excellent agreementwith the results obtained with c2-ray in highly ionised regions, where the neutral fraction is unaffected by spectral hardening. Thesmall differences in the neutral fractions obtained with traphic thick, traphic thin and traphic are mostly due to differences in therecombination rate, caused by differences in the gas temperatures (see Fig. 11).
Figure 11. Test 3. Temperature in a slice through z = Lbox/2 at time t = 0.2 Myr. From left to right: traphic thin (assumingoptically thin photoheating rates), traphic thick (assuming optically thick photoheating rates), traphic (using three frequency bins),c2-ray, crash and ftte. Contours show temperatures log10(T/K) = 3, 4, 4.2, 4.4 and 4.6, from the outside in. Most of the morphologicaldifferences may be attributed to differences in the spectral hardening of the ionising radiation, with the multi-frequency codes traphic,c2-ray and crash yielding a substantial amount of pre-heating and the monochromatic (grey) codes traphic thin, traphic thick andftte yielding sharp transitions between the hot ionised and the cold neutral phases. The differences in the maximum gas temperaturesare mainly due to photoheating being computed in the optically thick limit (traphic thick, c2-ray, ftte), the optically thin limit(traphic thin) or using multiple frequency bins (traphic, crash).
number of frequency bins as small as possible. Using a smallnumber of frequency bins in the present test should thusgive results that more closely resemble the results of future,larger simulations.
Figs. 10-12 show our results. Fig. 10 shows images of theneutral fraction in slices through the centre of the simula-tion box at time t = 0.2 Myr (our conclusions also hold forother times). The individual panels show results obtainedwith traphic thin, traphic thick and traphic. For ref-erence, we also show the results obtained with other RTcodes for the same test problem as published in Iliev et al.(2006a). Neutral fraction contours are shown to facilitate thecomparison. While the simulation with crash treated thepresent problem by performing a multi-frequency computa-tion, the simulation with ftte, as our simulation traphic
thick, solved it in the grey approximation using opticallythick photoheating rates. Finally, c2
-ray employed a hybridmethod that treats the transport of radiation with multiplefrequency bins but computes photoheating rates in the grey(optically thick) approximation.
The differences in the neutral fractions are generallysmall. The simulations that employ photoheating rates inthe optically thick limit (ftte, c
2-ray, traphic thick)
yield smaller minimum neutral fractions than the simu-lations that compute photoheating rates in the opticallythin limit (traphic thin) or using multiple frequency bins(crash, traphic). This is the result of lower recombina-tion rates caused by the higher temperatures these simula-tions yield16 (see Fig. 11). The regions with low ionisation(ηHI > 0.5) found with traphic are slightly smaller thanthose found with c
2-ray and crash, which indicates that
three frequency bins are not sufficient for obtaining highlyaccurate multi-frequency solution. Still, the simulations withtraphic seem to capture the main effects (see the discussionon pre-heating below).
Fig. 11 shows images of the gas temperature in slices
16 As noted earlier, the main reason why crash finds significantlylarger neutral fractions may be an insufficient sampling of thephoton field, see, e.g., Fig. 2 in Maselli, Ferrara, & Ciardi 2003.
c© RAS, MNRAS 000, 1–23
Andreas Pawlik (MPA)
TRAPHIC - multi-frequency and thermal coupling 19
Figure 10. Test 3. Neutral fraction in a slice through z = Lbox/2 at time t = 0.2 Myr. From left to right: traphic thin (assuming greyoptically thin photoheating rates), traphic thick (assuming grey optically thick photoheating rates), traphic (using three frequencybins), c2-ray, crash and ftte. Contours show neutral hydrogen fractions ηHI = 0.9, 0.5, log ηHI = −1,−3 and −5, from the outside in.The results obtained with traphic thick are in excellent agreement with those obtained with ftte. They are also in excellent agreementwith the results obtained with c2-ray in highly ionised regions, where the neutral fraction is unaffected by spectral hardening. Thesmall differences in the neutral fractions obtained with traphic thick, traphic thin and traphic are mostly due to differences in therecombination rate, caused by differences in the gas temperatures (see Fig. 11).
Figure 11. Test 3. Temperature in a slice through z = Lbox/2 at time t = 0.2 Myr. From left to right: traphic thin (assumingoptically thin photoheating rates), traphic thick (assuming optically thick photoheating rates), traphic (using three frequency bins),c2-ray, crash and ftte. Contours show temperatures log10(T/K) = 3, 4, 4.2, 4.4 and 4.6, from the outside in. Most of the morphologicaldifferences may be attributed to differences in the spectral hardening of the ionising radiation, with the multi-frequency codes traphic,c2-ray and crash yielding a substantial amount of pre-heating and the monochromatic (grey) codes traphic thin, traphic thick andftte yielding sharp transitions between the hot ionised and the cold neutral phases. The differences in the maximum gas temperaturesare mainly due to photoheating being computed in the optically thick limit (traphic thick, c2-ray, ftte), the optically thin limit(traphic thin) or using multiple frequency bins (traphic, crash).
number of frequency bins as small as possible. Using a smallnumber of frequency bins in the present test should thusgive results that more closely resemble the results of future,larger simulations.
Figs. 10-12 show our results. Fig. 10 shows images of theneutral fraction in slices through the centre of the simula-tion box at time t = 0.2 Myr (our conclusions also hold forother times). The individual panels show results obtainedwith traphic thin, traphic thick and traphic. For ref-erence, we also show the results obtained with other RTcodes for the same test problem as published in Iliev et al.(2006a). Neutral fraction contours are shown to facilitate thecomparison. While the simulation with crash treated thepresent problem by performing a multi-frequency computa-tion, the simulation with ftte, as our simulation traphic
thick, solved it in the grey approximation using opticallythick photoheating rates. Finally, c2
-ray employed a hybridmethod that treats the transport of radiation with multiplefrequency bins but computes photoheating rates in the grey(optically thick) approximation.
The differences in the neutral fractions are generallysmall. The simulations that employ photoheating rates inthe optically thick limit (ftte, c
2-ray, traphic thick)
yield smaller minimum neutral fractions than the simu-lations that compute photoheating rates in the opticallythin limit (traphic thin) or using multiple frequency bins(crash, traphic). This is the result of lower recombina-tion rates caused by the higher temperatures these simula-tions yield16 (see Fig. 11). The regions with low ionisation(ηHI > 0.5) found with traphic are slightly smaller thanthose found with c
2-ray and crash, which indicates that
three frequency bins are not sufficient for obtaining highlyaccurate multi-frequency solution. Still, the simulations withtraphic seem to capture the main effects (see the discussionon pre-heating below).
Fig. 11 shows images of the gas temperature in slices
16 As noted earlier, the main reason why crash finds significantlylarger neutral fractions may be an insufficient sampling of thephoton field, see, e.g., Fig. 2 in Maselli, Ferrara, & Ciardi 2003.
c© RAS, MNRAS 000, 1–23
log T [K]
2.0 2.6 3.2 3.8 4.4 5
TRAPHIC C2RAY CRASH FTTE
TRAPHIC thin TRAPHIC thick
![Page 19: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/19.jpg)
RecombinaCon radiaCon Milan Raicevic
Andreas Pawlik (MPA)
TRAPHIC exact, Case B exact, Case A
![Page 20: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/20.jpg)
RecombinaCon radiaCon Milan Raicevic
Andreas Pawlik (MPA)
Ritzerveld 2005 (analyCc)
![Page 21: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/21.jpg)
ApplicaCons
• The first stars (Jeon et al., in prep)
• The first galaxies (AP, Milosavljevic, Bromm 2012, arxiv:1208.3698)
• ReionizaCon (AP, Schaye, et al., LOFAR, in the future)
• HI absorpCon systems auer reionizaCon (RahmaC et al. 2012, arxiv:1210.7808+)
Andreas Pawlik (MPA)
![Page 22: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/22.jpg)
HI absorpCon systems in the UV background at z = 0-‐5
Ali RahmaC
L ~ 6-‐100 Mpc/h mgas ~ 105 – 108 solar (1283-‐5123) See also talks by Ken Nagamine,
Gabriel Altay, Mam McQuinn
![Page 23: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/23.jpg)
Andreas Pawlik (MPA)
(non-‐) EvoluCon of the HI column density distribuCon
Ali RahmaC, AP, Raicevic, & Schaye 2012a
![Page 24: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/24.jpg)
The first miniquasars Myoungwon Jeon
Andreas Pawlik (MPA)
• The first stars form in minihalos with masses 105-‐6 at z ~ 30
• Miniquasars; binaries evolve into HMXBs
• Strong emimers of X-‐rays • PreheaCng of the IGM
![Page 25: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/25.jpg)
Andreas Pawlik (MPA)
RadiaCve feedback No feedback
Virial mass
Maximum density
SFR
Atomic cooler
SF threshold
minihalo
RadiaCve feedback in the first galaxies AP, Milosavljevic, & Bromm 2012
Dwarf galaxy reaching a halo mass of 109 solar at z = 10
![Page 26: TRAPHIC( TRAnsportof PHotons(In Cones](https://reader031.vdocuments.net/reader031/viewer/2022020704/61fb6e5b2e268c58cd5e1611/html5/thumbnails/26.jpg)
Summary
• TRAPHIC – spaCally adapCve transfer at the speed of light and at a computaConal cost independent of the number of sources
• Many new features since first publicaCon, including RHD, recombinaCon radiaCon, etc.
• We apply TRAPHIC: feedback in the first galaxies, HI absorpCon systems, reionizaCon
Andreas Pawlik (MPA)