Can galaxy evolution mimic cosmic reionization?
DDraft version October 2, 2020
Typeset using L A TEX twocolumn style in AASTeX63
Can galaxy evolution mimic cosmic reionization?
Sultan Hassan
1, 2, 3, ∗ and Max Gronke † Center for Computational Astrophysics, Flatiron Institute, 162 5th Ave, New York, NY 10010, USA Department of Astronomy, New Mexico State University, Las Cruces, NM 88003, USA Department of Physics & Astronomy, University of the Western Cape, Cape Town 7535, South Africa Department of Physics & Astronomy, Johns Hopkins University, Baltimore, MD 21218, USA
Submitted to ApJABSTRACTLyman- α (Ly α ) emitting galaxies are powerful tools to probe the late stages of cosmic reionization.The observed sudden drop in Ly α fraction at z > α photons. Crucially, this interpretationof the observations is only valid under the assumption that galaxies themselves experience a minimalevolution at these epochs. By modelling Ly α radiative transfer effects in and around galaxies, weexamine whether a change in the galactic properties can reproduce the observed drop in the Ly α fraction. We find that an increase in the galactic neutral hydrogen content or a reduction in theoutflow velocity toward higher redshift both lead to a lower Ly α escape fraction, and can thus mimican increasing neutral fraction of the IGM. We furthermore find that this change in galactic propertiesleads to systematically different Ly α spectra which can be used to differentiate the two competingeffects. Using the CANDELSz7 survey measurements which indicate slightly broader lines at z ∼ z . We alsoshow that a decrease in outflow velocity is not ruled out by existing data but leads to more prominentblue peaks at z >
6. Our results caution the use of Ly α observations to estimate the IGM neutralfraction without accounting for the potential change in the galactic properties, e.g., by mapping outthe evolution of Ly α spectral characteristics. Keywords:
Reionization – Lyman-alpha galaxies – Galaxy evolution INTRODUCTIONLyman- α (Ly α ) line is a promising tool to probe cos-mic reionization as the increasingly neutral intergalacticmedium (IGM) becomes more opaque to Ly α photonstowards higher redshifts (e.g. for extensive review seeDijkstra 2014). This increased optical depth is expectedto give rise to a decrease in the observed number of Ly α emitting galaxies at z (cid:38)
6. Specifically, the numberof Ly α selected galaxies (or Lyman- α emitters, LAEs)decreases dramatically at this redshift; similarly, contin-uum selected, or Lyman break galaxies (LBGs) show a Corresponding author: Sultan Hassanshassan@flatironinstitute.org ∗ Flatiron fellow † Hubble fellow modest increase of Ly α emission from z ∼ z ∼ z ≥ z galaxies – andis commonly parametrized by the ‘Ly α fraction’ whichdescribes the fraction of LBGs possessing a Ly α equiv-alent width W > W c where W c is an observationallydetermined cutoff, usually 20 ˚A.These different Ly α based observations are being usedto constrain the evolution of the cosmic neutral fraction(Furlanetto et al. 2006; McQuinn et al. 2007; Kakiichiet al. 2016; Mason et al. 2018, 2019; Whitler et al. 2020). a r X i v : . [ a s t r o - ph . C O ] S e p Hassan & Gronke
In fact, at z ∼ (cid:104) x HI (cid:105) ≈ . +0 . − . (1 σ error; Mason et al.2018) and are, thus, currently more constraining thanmeasures of the cosmic microwave background (PlanckCollaboration et al. 2016) or quasar proximity zones(e.g., Greig et al. 2017; Davies et al. 2018).However, these constraints are crucially dependent onthe assumption that the average galactic Ly α escapefraction does not change over this redshift interval asthe observed Ly α flux is proportional to this times theintergalactic transmission. Therefore, an evolution inthe cosmic neutral fraction is fully degenerate with theevolution of the Ly α escape fraction. While on the onehand, the duration from z = 7 to z = 6 is merely ∼ α is a resonant line with a large crosssection which implies that Ly α escape through the inter-stellar and circumgalactic medium is a highly non-linearprocess. Several theoretical studies have shown that Ly α escape is dependent not only on the dust and neutral hy-drogen abundance (Neufeld 1990; Dijkstra et al. 2006)but also on its kinematics (Bonilha et al. 1979; Zheng& Wallace 2014), and structure (Neufeld 1991; Gronkeet al. 2017), and that even small changes in these prop-erties can have large effects on the Ly α observables –and, in particular, the escape fraction.Independently of the question whether the currentlyemployed assumption of a constant Ly α escape fractionwith redshift is justified, it is important to incorporateour ignorance regarding the evolution of the interstellarand circumgalactic medium into the models constrainingcosmic evolution. Sadoun et al. (2017) took a first striveat this goal by demonstrating that the observed drop inLy α fraction can be entirely due to the increased neutralhydrogen content in the infalling region surrounding thedark matter halo hosting the galaxy. While in theirinterpretation this increased neutral fraction is due to achange in the ionzing background – and, thus, arguablyalso a sign of cosmic reionization – this result is veryimportant as it shows the potential impact of this changeof Ly α transmission not stemming from an evolution ofthe intergalactic medium.In this paper, we want to systematically explore whatchanges in galactic properties can mimic the observedevolution of Ly α visibility usually attributed to theEpoch of Reionization. We will, furthermore, study howsuch changes will impact the Ly α spectra. This will al-low future studies to use this additional constraints, andthus allow them to fold in the uncertainty regarding thegalactic evolution into the models. This paper is organized as follows: in Sec. 2, we de-scribe the quantities and the radiative transfer codeused, in Sec. 3 we present our results, and we discussthem in Sec. 4. METHODS2.1. Ly α fraction As stated above, the Ly α fraction, X Ly α , is commonlydefined as the fractional abundance of galaxies with Ly α equivalent width ( W ) above certain cut-off ( W c ), whichcan be written as: X Ly α, W c = (cid:90) ∞ W c p ( W ) d W, (1)where p ( W ) is the equivalent width distribution func-tion. As commonly used in the literature (Dijkstra &Wyithe 2012; Gronke et al. 2015; Sadoun et al. 2017),we adopt an exponential form for p ( W ): p ( W ) = exp ( − W/W ) W + W for W > , (2)where W and W are free parameters, which can befound by matching to observations. It has been foundthat this parametrization reproduces observations rea-sonably well (Schenker et al. 2014b) .By integrating Equation 1 using two differentobservationally-motivated thresholds W c = 25 ,
55 ˚A, theexponential scale W becomes: W = 30 ˚Alog( X Ly α, /X Ly α, ) . (3)The W is M UV magnitude dependent as observationsindicate. Using the measured Ly α fractions at z = 6by Stark et al. (2011), we find W ,z =6 = 43 . , . M UV > − .
25) and bright ( M UV < − . W ,z =6 valuefor the faint population throughout. By doing so, wematch the observed Ly α fraction measurement at z =6. We then attempt to reproduce the observed dropin Ly α fraction at z > α equivalent width W isgiven by W = f esc , Ly α f esc , UV T IGM W i , (4) Note while we focus on the evolution in the Ly α fraction, a changein the observed equivalent widths, e.g., due to a change in galacticproperties, W → aW leads to a change in the scale height W → aW . This means that our results can be understood as a changein the equivalent width distribution. an galaxy evolution mimic cosmic reionization? T IGM is the IGM transmission, W i is the intrin-sic equivalent width, and f esc , Ly α and f esc , UV are thephoton escape fractions for Ly α and UV photons re-spectively. We assume that the IGM doesn’t evolve (i.e. T IGM = const . ), and the dust optical depth τ d is thesame at z ≥
6, which also keeps the f esc , UV constant( f esc , UV = exp( − τ d )). With these assumptions, we cantranslate the change in the equivalent width distributionto change in the Ly α photon escape fraction. By simul-taneously solving the Ly α fraction equations for z = 6and z >
6, we obtain f esc , Ly α ( z > f esc , Ly α ( z = 6) = (cid:20) − W ,z =6 W c log (cid:18) X Ly α,c ( z > X Ly α,c ( z = 6) (cid:19)(cid:21) − . (5)This equation relates the change in the photon escapefraction to the change in the Ly α fraction between tworedshifts. This relation depends on the equivalent widthcut-off W c and the exponential scale W for the equiva-lent width distribution, which is a magnitude dependent.Here we focus on the measurements for the faint pop-ulations ( M UV > − .
25) with cut-off W c > α frac-tion at z ≥
6. For simplicity, we take the averagevalues from Stark et al. (2011) and De Barros et al.(2017) at z ∼
6, Pentericci et al. (2014), Schenker et al.(2014a), Caruana et al. (2014), Mason et al. (2018) at z ∼
7, Tilvi et al. (2014), Treu et al. (2013), Schenkeret al. (2014a) and Mason et al. (2019) at z ∼
8. Theseaverage values of the Ly α fraction are 0.46, 0.24 0.14at z ∼ , ,
8, respectively. Using Equation 5, these av-erage values indicates that the photon escape fractionmay equivalently drop by f esc , Ly α ( z = 7) /f esc , Ly α ( z =6) = 0 .
47, and f esc , Ly α ( z = 8) /f esc , Ly α ( z = 6) = 0 . α fraction from z = 6to z = 7 and z = 6 to z = 8, respectively. Our aim is tostudy the conditions with which a change in the galacticproperties leads to these differences in the photon escapefraction, and hence mimicking reionization.2.2. Monte-Carlo Ly α Radiative Transfer
We model Ly α emission from and around galaxies as-suming shell models as implemented within a Monte-Carlo radiative transfer (MCRT) code tlac (Gronke &Dijkstra 2014). The MCRT methods tracks the evolu-tion and properties of injected photons including direc-tion and frequency as they travel through the simula- tion domain. The ‘shell model’ is commonly adopted asit has been shown to reproduce observed Ly α spectrawell (Ahn et al. 2003; Verhamme et al. 2006; Gronke2017) . It is defined by the neutral hydrogen columndensity N HI , the dust optical depth τ d , the expanding/ outflowing velocity v exp , and the effective tempera-ture T . In all our runs, we consider an initial numberof photon packages of N = 10 which we inject at linecenter, unless otherwise stated. We here consider twoscenarios to the change in the galactic properties thatcan lead to a change in the photon escape fraction re-quired to mimic reionization. While keeping all otherproperties fixed, these scenarios are changing only ei-ther the column density N HI or the outflows v exp . Weleave T = 10 K fixed for all the runs.To this end, we have shown how the drop in Ly α frac-tion is equivalent to a drop in the photon escape fractionwhile keeping the IGM fixed. In the next section, wepresent the our results relating the change in the galac-tic properties to the photon escape fraction and spec-tral properties, as well as comparing with observationsto discriminate between these two scenarios. RESULTS3.1.
Impact of galactic properties on f esc and Ly α spectra Figure 1 is a visual summary of how the Ly α photonescape fraction f esc and spectral properties change asthe galactic properties evolve. We show the dust impacton f esc with variation in the optical depth τ d as quotedin the legend and represented by different linestyles. Wecolor-code f esc dependence on the galactic properties( N HI & v exp ) with the width (square root of second mo-ment) of the red side of Ly α emission, and the pointsizes reflect its offset.The left panel of Fig. 1 shows the f esc dependenceon the column density N HI at a fixed outflow velocity v exp = 130 km / s, whereas the right panel depicts the de-pendence on the outflow velocity v exp at a fixed columndensity N HI = 10 cm − . In general, f esc decreases asthe N HI increases and v exp decreases. Both a higher HIcolumn density and a lower outflow velocity imply thatthe optical depth at line-center increases, and thus, sodoes the path length of Ly α photons through the scat-tering medium. This in turn means that the effectivedust optical depth increases, lowering the escape frac-tion. There is an ongoing discussion in the literature regarding thephysical meaning behind the ‘shell model’ (e.g. Orlitov´a et al.2018; Gronke et al. 2017). We will comment on the interpretationof our results in light of the adopted model in § Hassan & Gronke
Figure 1.
The Ly α photon escape fraction f esc dependence on the galactic properties. Left : f esc as a function of columndensity N HI at fixed outflow velocity v exp = 130 km / s. Right : same as left but as a function of v exp at a fixed N HI = 10 cm − .Different linestyles correspond to different dust amounts τ d as shown in the legend. Different colors and point sizes representthe spectral width (square root of second moment) of the red peak and offset of the red side of Ly α emission, respectively. f esc decreases in denser and dustier media and increases with higher outflows. The width and offset both increase with increasingcolumn density and outflow velocity. An increase in the column density N HI or decrease in outflows v exp towards high redshift( z >
6) by a factor of ∼ × − α fraction X Ly α and mimics the increase in neutral fraction(i.e. reionization) without an evolving IGM. Note that the f esc dependence on N HI is steeper thanthat on v exp . At small dust amounts (solid lines), f esc isalmost unity as most of photons easily escape. In thisregime, the galactic properties are required to changedramatically in order to observe a factor of 2 differ-ence in f esc . With a dustier medium (dashed and dot-ted lines), it is easier to find such a difference withsmaller change in the galactic properties. For instance,at τ d ≥ .
3, a change by ≤ N HI or by ≤
200 km / s in v exp is needed to reduce f esc by factor 2. We also see that the spectral propertiessuch as the width and offset change accordingly. Thesechanges can potentially be tested against observations(cf. § N HI or/and v exp increase. Similarly, the depen-dence on N HI is steeper since we see the width changefrom about ∼
50 km / s at log N HI / cm − =17 up tomore than 500 km / s by log N HI / cm − =21. Over-all we find a tight correlation between offset and width,with the offset being roughly about ∼ v exp somewhat modest as the it changes from 300 km / s atvery low v exp ∼ / s to about less than 500 km / s by v exp ∼
500 km / s. Likewise, the offset dependence on N HI is more significant since point sizes change signifi-cantly towards high N HI values as opposed to the slowchange as the v exp increases.In summary, Fig. 1 illustrates nicely the facts that (i) it is possible to find examples where the change inthe galactic properties can reduce f esc significantly, and (ii) that such a change is accompanied by a change inLy α spectral properties. The questions now are if thesechanges in f esc are sufficient to reproduce the observedsudden drop in Ly α fraction X Ly α without an evolvingIGM, and if – or rather – in which parameter range thesechanges in galactic properties are realistic.3.2. Escape fraction variation consistent with thechange in X Ly α As can already be seen from the previous section andFig. 1, the measured drop in X Ly α of ∼
50% ( ∼ z ∼ → z ∼ →
8) can be reproduced entirely by achange in galactic properties. To explore systematicallyfor which parameters this is the case, we ran a grid ofmodels and show explicitly in Figure 2 the change in an galaxy evolution mimic cosmic reionization? Figure 2.
Grid of models with different column densities N HI and fixed outflow velocity v exp = 50 km / s (left) and with differentoutflows v exp at fixed column density N HI = 10 cm − (right), both at the same amount of dust τ d = 0 .
35 at z ≥
6. Thehorizontal axis represents the column density and outflows at z = 6 whereas the vertical axis shows the same quantities at higherredshifts, z = 7 ,
8. Both panels are color-coded with photon escape fraction f esc ratio between z > z = 6. Contours showthe possible column densities/outflows between z = 6 → z = 7 and z = 6 → z = 8 with f esc difference that mimics the sign ofreionization as often inferred from the observed drop in Ly α fraction X Ly α at these epochs. escape fraction that is produced by a change in columndensity N HI and outflow velocity v exp . We furthermoredraw black contour lines representing the change neededto reproduce the observed X Ly α drop for a fixed choiceof dust ( τ d = 0 . / s. Starting at lowcolumn density e.g. N HI = 10 − cm − at z = 6, arather large increase by ∼ − ∼ halving of f esc by z = 7. On the other hand, for larger column densitieslog N HI / cm − >
19 a significantly smaller increase of < (cid:46) . f esc a change in outflow velocity produceswhile fixing N HI = 10 cm − . We see here contours rep-resenting the observed drop in X Ly α are located in thelower right part, which is opposite to the left panel. Thisshows that in order to mimic reionization with only theoutflow velocity, higher outflows are required at low red-shift ( z = 6). The required change in outflows is about ≤ / s between these redshifts, which is somewhatmoderate. While these models have fixed dust, outflow(left) and column density (right), we can easily predictthe corresponding change in these contours for differ-ent choice of parameters. For instance, f esc increases atlower dust values, which means that the contours in theleft and right panel would be shifted accordingly to up- per and lower part of the grid. Higher outflows increase f esc and hence the contours would shift to the lower partin the left panel, whereas higher column density wouldshift the contours in the right panel to the upper part.In this section, we showed that both a change in col-umn density and a change in outflow velocity can re-produce the observed drop in Ly α detections at z (cid:38) f esc . A change consistent with the measured drop ofthe Ly α fraction towards at z > ∼ . N HI (cid:38) or suppressing outflows towards high red-shift. We next discriminate between these scenarios us-ing the change in the spectral properties.3.3. Ly α spectral line properties variation as afunction of the galactic properties As already visible from Fig. 1, the change in the galac-tic properties that mimics reionization doesn’t only pro-duce a different f esc , but also changes significantly theLy α spectral line properties which can be tested againstobservations.Figure 3 shows several examples for the variation inLy α spectral line properties between redshift z = 6 and z = 7 at same amount of dust, τ d = 0 .
2. Top rowshows the resulting spectral changes due to changingthe column density N HI at fixed outflows v exp whereas Hassan & Gronke
Figure 3.
Several examples for the change in spectral properties between z = 6 (red) and z = 7 (blue) as the galactic propertieschange at a fixed amount of dust τ d = 0 . / s. Top row shows the change in column density N HI at fixed outflows v exp whereas the bottom row shows the opposite as quoted in the legend and subtitles. In all panels, thedifference in f esc is equivalent to the observed drop in Ly α fraction X Ly α between z = 6 and z = 7. In the top row, the linewidth and offset both increase as the column density increases towards high-z, whereas the bottom row indicates that the linewidth and offset both decrease as the velocity decreases towards high-z. the bottom row depicts the opposite scenario. All theseexamples possess the required ∼ × / f esc differencebetween z = 6 and z = 7 to mimic the drop in theobserved Ly α fraction X Ly α as explained in the previoussection.In the top panels of Fig. 3, we show that the scenarioof changing the column density to mimic reionizationsuggests that the line width and offset both increasetowards high redshift. It is also noted that the lineis broader at higher column densities. Interestingly,the second scenario of changing the outflows shown inthe lower row of Fig. 3 indicates exactly the oppositethat broader lines exist at higher outflow velocities,i.e., at lower redshifts. This is due to the fact thatincreasing the column density, increases also the escapefrequency offset at which Ly α can escape through excur-sion (Neufeld 1990). On the other hand, for outflowingmaterial a fraction of photons are ‘backscattered’ andobtain ∼ N HI or v exp ) tomimic reionization can be distinguished with observa- tions of Ly α spectra. We explore this in detail in thenext section.3.4. Comparison to observations
We now use the observations to discriminate betweenour two scenarios. Using the CANDELSz7 survey, Pen-tericci et al. (2018) have measured the line widths oftwo stacks of 52 sources with (cid:104) z (cid:105) ∼ (cid:104) z (cid:105) ∼
7, and found that their full widthat half maximum (FWHM) are equal to 300 ±
30 km / sand 220 ±
25 km / s, respectively. This shows that theobservations indicate that line width is approximatelyconstant or slightly decreasing with increasing redshift.Recalling the results of the previous section, this auto-matically rules out the scenario of changing the columndensity N HI to mimic reionization since it predicts thatthe line width increases towards high redshift.We now explore this more quantitatively as well ascheck whether a change in outflow velocity or the col-umn density and an associated drop in escape fractionis consistent with the line width measurements of Pen-tericci et al. (2018).To do so, we attempt to follow the recipe presentedin Pentericci et al. (2018) to produce stacks using ourmodel at z = 6 and z = 7. Using an initial number ofphotons of N photon = 10 , we run grid of 1,125 mod- an galaxy evolution mimic cosmic reionization? Figure 4.
Example for stacks of randomly selected 52 spectra at z = 6 (red) and 19 spectra at z = 7 (blue) with f esc ( z =7) ≈ . f esc ( z = 6) in agreement with the measured drop of X Ly α . We assembled the stacks following the method outlinedin Pentericci et al. (2018) with spectral resolution R = 1390 (corresponding to ∆ = 216 km / s). All these spectra are eitherobtained by changing only outflows v exp while keeping the column density N HI fixed (left) or changing only the column density N HI while keeping outflows v exp fixed. In all cases, dust is kept fixed. In either scenario, the FWHM of stacks at z=6 is consistentwith values reported in Pentericci et al. (2018, FWHM( z ∼
6) = 300 ±
30) . The FWHM of stacks obtained by changing outflows v exp at z=7 is also consistent with Pentericci et al. (2018) measurements (FWHM( z ∼
7) = 220 ±
25) while those produced bychanging the column density N HI are not, and hence indicating that changing the outflow v exp can naturally lead to a higherLy α escape as well as broader lines at lower redshifts as observations indicate. els over 5 different amount of dust ( τ d = 0 . − . v exp = 5 −
500 km / s) and 15 column den-sity (log N HI = 17 −
21 cm − ); all equidistantly spaced.By considering all possible combinations at fixed columndensity and dust content, we then select those whose dif-ference in f esc is equivalent to the observed drop in X Ly α between z = 6 and z = 7. We find 650 or 1036 com-bination of models at z = 6 and z = 7 satisfying theserequirements in changing the outflows or column densityscenarios, respectively. Out of these models, we ignorecombinations that are inconsistent (i.e. with FWHM > ± σ ) with Pentericci et al. (2018) measurements at z ∼
6. This reduces the number of models combina-tion to 143 and 201 in the case of changing the outflowsand column density, respectively. This means that thepresented results here by construction consistent with z = 6 observations, and we aim to explore the differentscenarios predictions at high-z as compared with mea-surements. Similar to Pentericci et al. (2018), we bin allspectra using a spectral resolution of R = 1390, corre-sponding to velocity resolution of ∆v = 216 km / s, andproduce median stacks of randomly selected 52 spectraat z = 6 and only 19 spectra at z = 7. We also addnoise to each individual spectrum drawn from Gaussiandistribution of zero mean and standard deviation setby the signal-to-noise ratio of S/N = 5. We show tworandom examples for stacks of this procedure at z = 6(red) and z = 7 (blue) in Figure 4 as obtained by chang-ing only the outflows (left) or changing only the columndensity (right). While both scenarios produce stacks that have consistent FWHM values with observationsat z = 6, the changing outflows scenario yields also aconsistent FWHM with z = 7 measurements. This alsoconfirms that changing the column density predict veryhigh FWHM > km/s at z = 7. As quoted in thelegend for changing the outflows scenario (left), z = 6lines are broader than those at z = 7, and consistentwith Pentericci et al. (2018). The lines are highly asym-metric with extended red wings. The blue peaks disap-pear due to the poor resolution.To quantify the width using the above recipe, we gen-erate randomly 10,000 combination of stacks at z = 6and z = 7 from the total number of model combina-tions (i.e. 143 and 201) in each scenario and computetheir widths. We show the resulting width distribu-tion at these redshifts in Figure 5. Results from chang-ing outflows and changing column density scenarios arepresented by dashed and solid lines. Shaded red andblue areas show Pentericci et al. (2018) measurements at z ∼ z ∼
7, respectively. From this exercises alone,we constrain the FWHM, using changing column densityscenario, to 261 . ± . z = 6 and 434 . ± . z = 7, and using changing outflows scenario,to 291 . ± . z = 6 and 224 . ± . z = 7. This evident that, over all possible combinationsand the prior range assumed, the changing outflows sce-nario produces consistent width distribution within the1- σ level of Pentericci et al. (2018) measurements. Thisconfirms that changing the column density predicts in-consistent width distributions. Hassan & Gronke
Figure 5.
Line width distribution using randomly generated10,000 stacks at z = 6 (red) and z = 7 (blue) obtained bychanging only the outflows (solid) or changing only the col-umn density (dashed). Shaded red and blue areas show Pen-tericci et al. (2018) measurements at z ∼ z ∼
7, re-spectively. It is evident that the scenario of changing theoutflows is consistent with the observations, while changingthe column density scenario predict very high FWHM valuesat z ∼ Figure 6.
Predictions for the blue/red flux ratio at z = 6(red) and z = 7 (blue) for changing the column density sce-nario (dashed) or changing the outflows (solid). Both sce-narios produce similar flux ratio at z = 6. Changing theoutflows scenario predicts higher flux ratio at z = 7, indicat-ing the presence of more blue peaks at high redshift, whichis opposite to the change in column density scenario. We now use these scenarios to make predictions forthe blue/red flux ratio, which is defined as the totalblue line flux divided by the total red flux. We performthese predictions at the level of individual spectra notwith stacks, since the blue peaks disappear due to thepoor resolution. We show the blue/red flux ratio dis-tribution from total number of model combinations in Figure 6. Both scenarios produce similar flux ratio at z = 6. However, changing the outflows scenario predictshigher flux ratio at z = 7, indicating the presence ofmore blue peaks at high redshift, which is opposite tothe change in column density scenario. CONCLUDING REMARKSThe decrease in Ly α visibility for z > α fraction whilekeeping the IGM fixed at z ≥
6. We have considered twoscenarios: changing the column density N HI or the out-flow velocity v exp . We found that decreasing the columndensity by only (cid:46) (cid:46) . N HI (cid:38) cm − )or increasing the outflow velocity by ∼
100 km s − withdecreasing redshift can both successfully reproduce theobserved drop in the Ly α fraction, and thus, ‘mimic’ anincreasing IGM neutral fraction. Note that these exactvalues depend on the observed drop in X Ly α . We haveadopted values consistent with most studies (cf. §
2) butnote that recent work by Kusakabe et al. (2020) havereported much lower value of Ly α fraction at z ∼ X Lyα = 0 .
13, which is less than Stark et al. (2011) bya factor of 4. Naturally this would lead to a smallerevolution in the explored galactic properties.To differentiate between the evolution in galacticand intergalactic properties, we analyze the associatedchange in Ly α spectral properties. The line width andoffset both increase as the column density and outflowsincrease. The observed spectral properties can poten-tially discriminate between these scenarios, which indi-cate that the broader lines exists in low redshift (Pen-tericci et al. 2018). This automatically rules out thechanging column density scenario (cf. § . ± . / s at z = 6 and 224 . ± . / sat z = 7, which is consistent with the 1- σ level of thePentericci et al. (2018) measurements. We predict thatsuch a scenario of a change in outflow properties impliesa larger flux on the blue side of Ly α towards higherredshifts. While in principle this could be directly de-tectable – and there has been an increasing number ofblue peaks at high- z has been detected (Hu et al. 2016;Songaila et al. 2018; Matthee et al. 2018; Bosman et al.2020) – the IGM already at z (cid:46) an galaxy evolution mimic cosmic reionization? α photons on the blue side altering sys-tematically the observed spectra (Laursen et al. 2011;Hayes et al. 2020; Byrohl & Gronke 2020). Such an evo-lution would thus have to be indirect, i.e., through the(change in) Ly α halo properties.Throughout this study, we model the complex ra-diative transfer through the galactic and circumgalac-tic medium by the simple concentric, outflowing shell.While this ‘shell-model’ has been shown to reproduce ob-served Ly α spectra well , there is an ongoing discussionin the literature why this is, and what the shell-modelparameters represent (e.g., Gronke et al. 2017; Orlitov´aet al. 2018; Li et al. 2020). While a full discussion of theproblematic is beyond the scope of this work, we wanthighlight some points most relevant for this study. Inparticular, both theoretical (Dijkstra et al. 2016; Eideet al. 2018; Kakiichi & Gronke 2019) as well as obser-vational work (Vielfaure et al., in prep.) points towardsthe fact that Ly α spectra are in a way an extremumstatistics, i.e., they are heavily weighted towards lowopacity channels. This is maybe unsurprising as Ly α photons most easily escape through these ‘pathways ofleast resistance’. Specifically, from this theoretical workit became clear that Ly α spectra can be shaped by thelowest column densities channels – even when this is notalong the line-of-sight towards the observer (e.g., Eideet al. 2018) . This implies for our study that not theaverage galactic properties but instead the ‘extreme’ (interms of opacity) has to evolve in order to mimic reion-ization. As such lower density or higher velocity chan-nels can occur on relatively short timescales ( (cid:46) tens ofMyr), for instance, due to a burst of star formation (Nor-man & Ikeuchi 1989; Sparre et al. 2017; Faisst et al.2019), the required change in the ‘shell model’ param-eters from z ∼ ∼ , >
200 Myr) is in fact notunlikely. We expect future studies targeting the connec-tion of the ‘shell-model’ to more realistic gas geometriesto allow improved estimates on their variability.A limitation of this study is the homogeneous treat-ment of the galaxy population. A more realistic ap-proach is to develop a novel approach, such abundancematching, to link the equivalent width distribution tothe galaxy mass or stellar mass function at these epoch,and then study what change is required in the whole Something that cannot be claimed when using more complexinput geometry, e.g., from galactic simulations (see discussionin Gronke et al. 2018; Mitchell et al. 2020) which further justifiesthe usage of the ‘shell-model’ in this study. While observational confirmation is difficult as it requires an in-dependent tracer of the HI column density, for instance, GRBafterglow spectra offer an attractive opportunity (Vielfaure et al.2020). galaxy population to mimic the whole equivalent widthdistribution. We leave exploring possibilities to under-take such approach in future works.Additionally, the dust has been also assumed fixed inour analysis. To first order expected evolution of dustoptical depth for Ly α photons is to follow the metallic-ity, i.e., to decrease towards high redshift. This in turnwould increases the f esc , Ly α , and goes in the oppositedirection to the observed drop in X Ly α . In our bothscenarios, it is still possible to combine the increase ofdust towards high redshift with the increase of columndensity of decrease in outflows, but these changes inthe galactic properties would be larger to offset theevolution in dust. Given the uncertainty in dust andmetalicity at these high redshift epochs, and generaluncertainty of how does interacts with Ly α photons ,we have compared models at the same level of dustoptical depth.In summary, we have shown that the change in Ly α visibility towards higher redshift can be attributed to achange in galactic properties and do not require a changein IGM properties. Specifically, we find that both amodest decrease in column density (in agreement withSadoun et al. 2017) or an increase in outflow velocityin the galaxies’ evolution can lead to a increase in Ly α escape fraction, and thus, to the observed drop in de-tected Ly α emission towards at z >
6. We furthermorefound that this degeneracy between IGM transmissionand galactic Ly α radiative transfer can be broken usingthe emergent Ly α spectral properties. In particular, wefound that the scenario of a change in column densityleads to unnatural wide Ly α profiles and is ruled out byexisting data – but that the change is outflow proper-ties is not. We predict that in such a case there will bemore blue flux emergent from galaxies towards redshiftswhich can be detectable either directly or through anevolution in Ly α halo properties.Naturally, a fast evolution of the IGM neutral fractionat z > α radiative transfer should be taken into accountwhen constraining this evolution using the observed Ly α equivalent width distribution or luminosity function.ACKNOWLEDGEMENTSThe authors acknowledge helpful discussions withCharlotte Mason, Daniel Stark and Steven Finkelstein. How susceptible Ly α photons are to dust depends heavily itsdistribution and is focus of a large body of literature (e.g., Neufeld1991; Hansen & Oh 2006). Hassan & Gronke
Simulations and analysis were performed at NMSU’sDISCOVERY supercomputers. This work also used theExtreme Science and Engineering Discovery Environ-ment (XSEDE), which is supported by National ScienceFoundation grant number ACI-1548562, and computa- tional resources (Bridges) provided through the alloca-tion AST190003P. MG was supported by NASA throughthe NASA Hubble Fellowship grant HST-HF2-51409and acknowledges support from HST grants HST-GO-15643.017-A, HST-AR-15039.003-A, and XSEDE grantTG-AST180036.REFERENCES
Ahn, S.-H., Lee, H.-W., & Lee, H. M. 2003, MonthlyNotices of the Royal Astronomical Society, 340, 863869,doi: 10.1046/j.1365-8711.2003.06353.xBonilha, J. R. M., Ferch, R., Salpeter, E. E., Slater, G., &Noerdlinger, P. D. 1979, The Astrophysical Journal, 233,649, doi: 10.1086/157426Bosman, S. E. I., Kakiichi, K., Meyer, R. A., et al. 2020,ApJ, 896, 49, doi: 10.3847/1538-4357/ab85cdByrohl, C., & Gronke, M. 2020, arXiv e-prints,arXiv:2006.10041. https://arxiv.org/abs/2006.10041Caruana, J., Bunker, A. J., Wilkins, S. M., et al. 2014,MNRAS, 443, 2831, doi: 10.1093/mnras/stu1341Davies, F. B., Hennawi, J. F., Ba˜nados, E., et al. 2018,ApJ, 864, 142, doi: 10.3847/1538-4357/aad6dcDe Barros, S., Pentericci, L., Vanzella, E., et al. 2017,A&A, 608, A123, doi: 10.1051/0004-6361/201731476Dijkstra, M. 2014, PASA, 31, e040,doi: 10.1017/pasa.2014.33Dijkstra, M., Gronke, M., & Sobral, D. 2016, ApJ, 823, 74,doi: 10.3847/0004-637X/823/2/74Dijkstra, M., Haiman, Z., & Spaans, M. 2006, TheAstrophysical Journal, 649, 14, doi: 10.1086/506243Dijkstra, M., & Wyithe, J. S. B. 2012, MNRAS, 419, 3181,doi: 10.1111/j.1365-2966.2011.19958.xEide, M. B., Gronke, M., Dijkstra, M., & Hayes, M. 2018,ApJ, 856, 156, doi: 10.3847/1538-4357/aab5b7Faisst, A. L., Capak, P. L., Emami, N., Tacchella, S., &Larson, K. L. 2019, ApJ, 884, 133,doi: 10.3847/1538-4357/ab425bFurlanetto, S. R., Zaldarriaga, M., & Hernquist, L. 2006,MNRAS, 365, 1012,doi: 10.1111/j.1365-2966.2005.09785.xGreig, B., Mesinger, A., Haiman, Z., & Simcoe, R. A. 2017,MNRAS, 466, 4239, doi: 10.1093/mnras/stw3351Gronke, M. 2017, A&A, 608, A139,doi: 10.1051/0004-6361/201731791Gronke, M., & Dijkstra, M. 2014, MNRAS, 444, 1095,doi: 10.1093/mnras/stu1513Gronke, M., Dijkstra, M., McCourt, M., & Peng Oh, S.2017, Astronomy & Astrophysics, 607, A71,doi: 10.1051/0004-6361/201731013 Gronke, M., Dijkstra, M., Trenti, M., & Wyithe, S. 2015,MNRAS, 449, 1284, doi: 10.1093/mnras/stv329Gronke, M., Girichidis, P., Naab, T., & Walch, S. 2018,ApJL, 862, L7, doi: 10.3847/2041-8213/aad286Hansen, M., & Oh, S. P. 2006, MNRAS, 367, 979,doi: 10.1111/j.1365-2966.2005.09870.xHayes, M., Schaerer, D., ¨Ostlin, G., et al. 2011, ApJ, 730, 8,doi: 10.1088/0004-637X/730/1/8Hayes, M. J., Runnholm, A., Gronke, M., & Scarlata, C.2020, arXiv e-prints, arXiv:2006.03232.https://arxiv.org/abs/2006.03232Hoag, A., Bradaˇc, M., Huang, K., et al. 2019, ApJ, 878, 12,doi: 10.3847/1538-4357/ab1de7Hu, E. M., Cowie, L. L., Songaila, A., et al. 2016, ApJL,825, L7, doi: 10.3847/2041-8205/825/1/L7Jung, I., Finkelstein, S. L., Dickinson, M., et al. 2020, arXive-prints, arXiv:2009.10092.https://arxiv.org/abs/2009.10092Kakiichi, K., Dijkstra, M., Ciardi, B., & Graziani, L. 2016,MNRAS, 463, 4019, doi: 10.1093/mnras/stw2193Kakiichi, K., & Gronke, M. 2019, arXiv e-prints,arXiv:1905.02480. https://arxiv.org/abs/1905.02480Kusakabe, H., Blaizot, J., Garel, T., et al. 2020, A&A, 638,A12, doi: 10.1051/0004-6361/201937340Laursen, P., Sommer-Larsen, J., & Razoumov, A. O. 2011,ApJ, 728, 52, doi: 10.1088/0004-637X/728/1/52Li, Z., Steidel, C. C., Gronke, M., & Chen, Y. 2020, arXive-prints, arXiv:2008.09130.https://arxiv.org/abs/2008.09130Mason, C. A., Treu, T., Dijkstra, M., et al. 2018, ApJ, 856,2, doi: 10.3847/1538-4357/aab0a7Mason, C. A., Fontana, A., Treu, T., et al. 2019, MNRAS,485, 3947, doi: 10.1093/mnras/stz632Matthee, J., Sobral, D., Gronke, M., et al. 2018, A&A, 619,A136, doi: 10.1051/0004-6361/201833528McQuinn, M., Hernquist, L., Zaldarriaga, M., & Dutta, S.2007, MNRAS, 381, 75,doi: 10.1111/j.1365-2966.2007.12085.xMitchell, P., Blaizot, J., Cadiou, C., & Dubois, Y. 2020,arXiv e-prints, arXiv:2008.12790.https://arxiv.org/abs/2008.12790 an galaxy evolution mimic cosmic reionization?11