What can DLAs and the gas around galaxies tell us about
galaxy forma9on?
Simeon Bird, CMU arXiv:1407.7858, 1405.3994
with Mar9n Haehnelt, Marcel Neeleman, Kate Rubin, Mark Vogelsberger, Shy Genel, Lars Hernquist
What are DLAs?
DLAs: neutral hydrogen reservoirs at z=4-‐2
HI 6 Ellison et al.
Figure 3. Voigt profile models to the Lyα line for the absorbers in our sample are shown as red curves. The dashed lines indicate the uncertainties given in thepanel captions.
data were obtained in TIME-TAGmode with at least two exposuresper target, each in a different FP-POS position. Total exposure timesand central wavelength settings are given in Table 4.
The COS spectra were homogeneously reduced using CAL-COS v2.15.4 and custom routines for background estimationand co-addition of individual exposures in the Poisson limit(e.g. Worseck et al. 2011). The data were flat-fielded, deadtime-corrected, doppler-corrected and wavelength-calibrated using thedefault CALCOS calibration files as of September 2011. For 1-dextraction we adjusted the source extraction window to a rectangu-lar box (height 17 pixels, corresponding to 5× the spatial FWHM)to preserve integer extracted counts in the Poisson regime, andto maximise the signal-to-noise ratio (S/N) for our well-centredpoint sources. Likewise, the background extraction windows weremaximised (150 pixels each), and background smoothing by thepipeline turned off to investigate spectral variations of the back-ground. After visual inspection, the smoothed background was de-termined by a 51-pixel running average in wavelength. Individualexposures were co-added by adding the integer source counts andthe non-integer mean background per pixel on the homogeneousNUV wavelength grid, accounting for individual pixel exposure
times induced by offsets in dispersion direction. The counts werethen flux-calibrated with the NUV flux calibration curve and thepixel exposure times. The S/N ratios per pixel (in the continuum oneither side of the Lyα absorption) of the final combined spectra aregiven in Table 4.
The Starlink package DIPSO was used to model a Voigt-profile absorption to the normalised spectra. To normalise the spec-tra, a cubic spline was fit through unabsorbed regions of the QSOcontinuum and the data then divided by the polynomial fit. Despitethe relatively low S/N, the N(H i) could usually be determined towithin ± 0.1 dex, which is typical for moderate resolution spec-tra of low S/N (e.g. Russell, Ellison & Benn 2006). The dominantsource of error in N(H i) is from uncertainty in the continuum place-ment. This is a particular issue for the proximate DLA towardsB0105−008 where the Lyα absorption lies on top of the QSO’sLyα emission. Nonetheless, errors on N(H i) are less ! 0.15 dex forall the bona-fide DLAs (log N(H i) " 20.3) in our sample. Table 4lists the measured N(H i) column densities for both the COS andSTIS data and Figure 3 presents the Lyα fits to the absorbers in oursample.
From Table 4 it can be seen that 4/6 absorbers in our sample
c⃝ 0000 RAS, MNRAS 000, 000–000
NHI > 2x1020 cm-‐2
Prochaska+ 2008
What are DLAs?
Dense neutral self-‐shielded regions HI Density: 0.01 cm-‐3 or 10% star forma9on
threshold
Why are we interested in DLAs?
1. Several problems making them – Velocity width – Bias
2. Direct probe of nearly star-‐forming gas around galaxies
Why are we interested in gas around galaxies?
• We know star forma9on changes overall galaxy evolu9on
• People tune sub-‐grid models to stellar mass func9on
• Gas provides an independent model check
Why are we interested in gas around galaxies?
1. Gas provides an independent model check
– Few systema9cs: sensi9ve to one ion – Hard to get sta9s9cs: Quasars are rarer than galaxies
Basic Results
1. Column Density
2. Metallicity
Column Density at z=3 N CDD
F
Stellar Feedback
Ouclows with velocity propor9onal to halo circular velocity (high mass loading to suppress star forma9on in dwarfs)
Mass loading: Wind velocity:
Ouclows in small halos SLOWER but push out more mass
⌘ / v�2w
vw = 3.7�1D
Stellar Feedback
Four different models: 1. NOSN: No (effec9ve) feedback 2. ILLUS: Illustris model 3. 2xUV: Illustris model with 2xUVB amplitude 4. HVEL: Constant velocity winds: 600 km/s
Blue is my favourite Green is my least favourite
Column Density at z=3 N CDD
F
Basic Results
1. Column Density • 2xUV is ok at z=3!
2. Metallicity
Neutral Hydrogen Abundance
Data: Noterdaeme+ 2012
Neutral Hydrogen Abundance
• At z=2 larger halos are forming • Here supernova winds are not as effec9ve • Too much cold gas
Basic Results
1. Column Density • Ok at z>3! • Too much at z=2
2. Metallicity
Basic Results
1. Column Density • Ok at z>3! • Too much at z=2
2. Metallicity
DLA Metallicity: z=3
DLA Metallicity: z=4,2
Change in mean metallicity due to change in mean halo size
Basic Results
1. Column Density • Ok at z>3! • Too much at z=2
2. Metallicity • Ok!
Less Basic Results
1. Velocity Width
2. Edge-‐leading Spectra
3. Bias
Measuring DLA Velocity Structure
Use metal lines: Lyman-‐α saturated
Velocity width: 90% of total op9cal depth
Correlates with halo virial velocity
Stellar Feedback
New models: 1. NOSN: No (effec9ve) feedback 2. DEF: Illustris model and 2xUV 3. FAST: Same as DEF with 50% faster winds 4. HVEL: Constant velocity winds: 600 km/s 5. (SMALL: As DEF but 10 Mpc box)
Velocity Widths
Velocity Width (km/s)
Velocity W
idth pdf
Data: Neeleman+ 2013
Run cosmological hydro simula9on (box: 25 & 10 Mpc)
Sample snapshot randomly for 5000 DLA spectra
How did we do this?
Two Important Ingredients
1. Make metal (SiII) cover larger region of halo 2. Make halos (a bit) larger
How did we do this?
Shielding
Rahma9+ 2013
On the evolution of the HI CDDF 11
Figure 3. The hydrogen neutral fraction (left) and the photoionization rate (right) as a function of hydrogen number density do notchange by varying the simulation box size or mass resolution. This is shown for different simulations at z = 3 in the presence of the UVBand recombination radiation. Purple solid, blue dashed and red dot-dashed lines show, respectively, the results for L12N256, L06N128
and L06N256. The green dotted line indicates the results for the L06N128 simulation if the gas is assumed to be optically thin tothe UVB radiation (i.e., no RT calculation is performed). The deviation between the optically thin hydrogen neutral fractions and RTresults at n
H! 10−2 cm−3 shows the impact of self-shielding. The lines show the medians and the shaded areas indicate the 15%− 85%
percentiles. At the top of each panel we show HI column densities corresponding to each density.
ΓPhot is the total photoionization rate. Moreover, the self-shielding density threshold, nH,SSh, is given by equation (13)and is thus a function of ΓUVB and σνHI
which vary withredshift. As explained in more detail in Appendix A1, thenumerical parameters representing the shape of the pro-file are chosen to provide a redshift independent best fitto our RT results. In addition, the parametrization is basedon the main RT related quantities, namely the intensity ofUVB radiation and its spectral shape. It can therefore beused for UVB models similar to the Haardt & Madau (2001)model we used in this work (e.g., Faucher-Giguere et al.2009; Haardt & Madau 2012). For a given UVB model, oneonly needs to know ΓUVB and σνHI
in order to determine thecorresponding nH,SSh from (13) (see also Table 2). Then, af-ter using equation (14) to calculate the photoionization rateas a function of density, the equilibrium hydrogen neutralfraction for different densities, temperatures and redshiftscan be readily calculated as explained in Appendix A1.
We note that the parameters used in equation (14)are only accurate for photoionization dominated cases. Aswe show in §3.5, at z ∼ 0 the collisional ionization rateis greater than the total photoionization rate around theself-shielding density threshold. Consequently, equation (13)does not provide an accurate estimate of the self-shieldingdensity threshold at low redshifts. In Appendix A1 we there-fore report the parameters that best reproduce our RT re-sults at z = 0. Our tests show that simulations that useequation (14) reproduce the f(NHI, z) accurately to within10% for z ! 1 where photoionization is dominant (see Ap-pendix A1).
Although using the relation between the median pho-toionization rate and the gas density is a computationallyefficient way of calculating equilibrium neutral fractions inbig simulations, it comes at the expense of the informa-tion encoded in the scatter around the median photoioniza-tion rate at a given density. However, our experiments show
that the error in f(NHI, z) that results from neglecting thescatter in the photoionization rate profile is negligible forNHI ! 1018 cm−3 and less than " 0.1 dex at lower columndensities (see Appendix A1).
3.4 The roles of diffuse recombination radiation
and collisional ionization at z = 3
To study the interplay between different ionizing processesand their effects on the distribution of HI, we compare theirionization rates at different densities. We start the analysisby presenting the results at z = 3 and extend it to otherredshifts in §3.5.
The total photoionization rate profiles shown in theright panel of Figure 3 are almost flat at low densitiesand decrease with increasing density, starting at densitiesn
H∼ 10−4 cm−3. Just below n
H= 10−2 cm−3 self-shielding
causes a sharp drop, but the fall-off becomes shallower forn
H> 10−2 cm−3 and the photoionization rate starts to in-
crease at nH> 10 cm−3. As shown in Figure 4, the shallower
fall-off in the total photoionization rate with increasing den-sity is caused by RR. The increase in the photoionizationrate with density at the highest densities on the other hand,is an artifact of the imposed temperature for ISM particles(i.e., T = 104 K) which produces a rising collisional ioniza-tion rate with increasing density. As the comparison betweenthe UVB and RR photoionization profiles shows (see Figure4), RR only starts to dominate the total photoionizationrate at n
H> 10−2 cm−3, where the UVB photoionization
rate has dropped by more than one order of magnitude andthe gas is no longer highly ionized. RR reduces the totalHI content of high-density gas by ≈ 20%. Although ion-ization rates remain non-negligible at higher densities, theycannot keep the hydrogen highly ionized. For instance at
c⃝ 0000 RAS, MNRAS 000, 000–000
Photo-‐ioniza9on rate reduced by self-‐shielding
neutral
ionised
Shielding for Metals Same self-‐shielding for E > 1 Rydberg SiII at lower density than HI (16.3 eV > 13.6 eV)
Ionisa9on state
Blue: SiII traces HI n(SiII)/n(Si) = n(HI)/n(H)
Velocity width approx virial velocity
DEF: Illustris model NOSN: No feedback Disc: A semi-‐analy9c model of DLAs as discs HSR98: Haehnelt+98 simula9ons
Velocity width / virial velocity
Two Important Ingredients
1. Make metal (SiII) cover larger region of halo 2. Make halos (a bit) larger
How did we do this?
Host Virial Velocity
• Feedback suppresses halos < 40 km/s • Characteris9c velocity now 70 km/s
Viria
l velocity
Velocity Widths
Velocity Width (km/s)
Velocity W
idth pdf
Data: Neeleman+ 2013
Less Basic Results
1. Velocity Width
2. Edge-‐leading Spectra
3. Bias
Edge-‐Leading Spectra
• Edge-‐leading sta9s9c
• Difference between posi9on of peak and midpoint of total absorp9on in units of velocity width
vpk � vmean
�v/2
Edge-‐Leading Spectra
Absorp9on concentrated at edges: marginally significant
Edge-‐Leading Spectra
A guess: ouclows have a special velocity.
Less Basic Results
1. Velocity Width 2. Edge-‐leading Spectra 3. DLA Bias
DLA Bias
• Bias b = DLA power / maper power • Gives host halo mass
• Observed (Font-‐Ribera 2012): bias (b) = 2.4 => M = 1011 – 1012
• We have halos M = 1010 – 1010.5
• Evidence for larger halos?
DLA Bias
Data: Font-‐Ribera+2012
DLA Bias
• Two reasons we get higher bias than expected: – b = 2.2 => M = 1011 – 1012 assumes:
– More DLAs in large galaxies
�DLA = �0
✓M
1010
◆↵
↵ = 1 vs 1/2
DLA Bias
• DLAs are an integral over a wide range of halo masses
• Non-‐linear effects occur at the non-‐linear scale for the most biased objects, not objects at the mean bias
• Non-‐linear growth is faster, so larger objects contribute more
DLA Bias
Data: Font-‐Ribera+2012
Non-‐linear effects boost DLA power spectrum
H[b2P (k)]
H[P (k)]> b2
DLA Bias
Data: Font-‐Ribera+2012
Odd scale dependence
Power decrease
Non-‐linear growth faster, more power on small scales
Cross-‐correla9on?
Conclusions
• People have asked whether DLAs are large milky-‐way discs (1012) or small dwarfs (1010)
• They are both, but mostly in between (1011).
• Can reproduce velocity width and DLA bias
Conclusions
• Reducing star forming gas works
• A model tuned to the stellar mass func9on makes a decent showing on DLAs.
• DLAs sensi9ve to mass of gas removed and return 9me – not details of ouclows