Redshift Evolution of the H2/HI Mass Ratio In Galaxies
Laura Morselli, Alvio Renzini, Andrea Enia, Giulia Rodighiero
MMNRAS , 1–6 (2021) Preprint 27 January 2021 Compiled using MNRAS L A TEX style file v3.0
Redshift Evolution of the H /HI Mass Ratio In Galaxies Laura Morselli , ★ , A. Renzini , A. Enia , , G. Rodighiero , Dipartimento di Fisica e Astronomia, Università di Padova, vicolo dell’Osservatorio 3, I-35122 Padova, Italy INAF − Osservatorio Astrofisico di Padova, vicolo dell’Osservatorio 5, I-35122 Padova, Italy Dipartimento di Fisica e Astronomia, Università di Bologna, Via Gobetti 93/2, I-40129, Bologna, Italy INAF - Osservatorio di Astrofisica e Scienza dello Spazio, Via Gobetti 93/3, I-40129, Bologna, Italy
Accepted 2021 January 18. Received 2021 January 12; in original form 2020 December 11
ABSTRACT
In this paper we present an attempt to estimate the redshift evolution of the molecular toneutral gas mass ratio within galaxies (at fixed stellar mass). For a sample of five nearby granddesign spirals located on the Main Sequence (MS) of star forming galaxies, we exploit mapsat 500 pc resolution of stellar mass and star formation rate ( 𝑀 ★ and SFR). For the same cells,we also have estimates of the neutral ( 𝑀 HI ) and molecular ( 𝑀 H ) gas masses. To computethe redshift evolution we exploit two relations: i) one between the molecular-to-neutral massratio and the total gas mass ( 𝑀 gas ), whose scatter shows a strong dependence with the distancefrom the spatially resolved MS, and ii) the one between log ( 𝑀 H / 𝑀 ★ ) and log ( 𝑀 HI / 𝑀 ★ ) . Forboth methods, we find that 𝑀 H / 𝑀 HI within the optical radius slightly decreases with redshift,contrary to common expectations of galaxies becoming progressively more dominated bymolecular hydrogen at high redshifts. We discuss possible implications of this trend on ourunderstanding of the internal working of high redshift galaxies. Key words: galaxies: evolution – galaxies: star formation – galaxies: spirals
Our understanding of galaxy formation and evolution is strictly con-nected to the accretion of cold gas on galaxies across cosmic time:this gas coming from the cosmic web cools down to form atomichydrogen (HI) first, and then molecular hydrogen (H ), that caneventually collapse under gravitational instability to form new stars.Feedback from star formation also plays a crucial role, as it is anecessary ingredient to ensure a low efficiency of the star formationprocess itself: without feedback the gas in a galaxy would be con-sumed almost completely over a free-fall time, turning most baryonsinto stars, as opposed to the ∼
10 per cent of baryons being lockedinto stars as actually observed in the local Universe (e.g. Bigiel et al.2008; Krumholz et al. 2012; Hayward & Hopkins 2017). Feedbackfrom star formation includes photo-dissociation of H into HI dueto the radiation emitted by young stars (e.g. Allen et al. 2004; Stern-berg et al. 2014). Therefore, HI is not only an intermediate gasphase towards star formation, but also one of its products, and it iskey in establishing the self-regulating nature of the star formationprocess. Unfortunately, till now our knowledge of the HI content inindividual galaxies is restricted to the low redshift Universe, whereHI is detected in emission via the 21cm line. Several surveys havetargeted HI in galaxies at 𝑧 < .
05: HIPASS (Meyer et al. 2004),ALFALFA (Giovanelli et al. 2005), xGASS (Catinella et al. 2018), ★ E-mail: [email protected]
HI-MaNGA (Masters et al. 2019). At higher redshift, the HIGHzsurvey (Catinella & Cortese 2015) targeted the HI emission of mas-sive galaxies at 𝑧 ∼ .
2, while the CHILES survey pushed the limitof individual detections up to 𝑧 ∼ . 𝑧 ∼ . 𝑧 ∼
1. Damped Ly 𝛼 orMgII absorption line systems give us the chance to estimate the HIcontent at 𝑧 (cid:38) content of five nearby, grand-design, massivemain sequence (MS) galaxies on scales of ∼ /HI increases with gassurface density, and at fixed total gas surface density it decreases(increases) for regions with a higher (lower) specific star formationrate (sSFR). In this paper we exploit tight correlations to estimatethe evolution with redshift of the H /HI mass ratio within galax-ies. It is generally assumed that this ratio increases with redshift,because galaxies are more gas rich and as the gas surface densityincreases, recombination is favored. However, galaxies at high red-shift are also more star forming, and higher levels of star formationfavor photo-dissociation of the H molecule, hence it is not a prioriobvious which trend would dominate over the other. © a r X i v : . [ a s t r o - ph . GA ] J a n Morselli L. et al. ∗ , SFR, H AND HI AT 500 pc RESOLUTION
The methodology to retrieve estimates of the stellar mass ( 𝑀 ★ ),SFR, HI mass ( 𝑀 HI ) and H mass ( 𝑀 H ) is detailed in Eniaet al. (2020) and M20. Briefly, starting from the DustPedia archive(Davies et al. 2017; Clark et al. 2018) we built a sample of fivenearby, face-on, grand design spiral galaxies with stellar mass inthe range 10 . − . 𝑀 (cid:12) , that lie on the MS relation at 𝑧 =
0. Thesesources have been observed in at least 18 bands from the far ultravi-olet (FUV) to the far infrared (FIR). We used the photometric datafrom FUV to FIR to run SED fitting with MAGPHYS (da Cunhaet al. 2008) on cells of 500pc × UV ) and obscured (SFR IR ) contribu-tions. To this aim, SFR UV and SFR IR have been computed using thescaling relations of Bell & Kennicutt (2001) and Kennicutt (1998),respectively, where the UV and IR luminosities ( 𝐿 UV and 𝐿 IR ) areevaluated from the best-fit SED (see Enia et al. 2020). Finally, asthese sources are included in the HERACLES (Leroy et al. 2009)and THINGS (Walter et al. 2008) surveys, they have been observedin CO(2-1) and HI at 21 cm. Hereafter, we make use of the H es-timated using 𝛼 CO = 4.3 𝑀 (cid:12) ( K · km · s − pc ) − (e.g. Bolatto et al.2013). Details on how the HI and H maps at 500pc resolution wereobtained can be found in M20, where the consistency of the resultsusing a constant or metallicity-dependent 𝛼 CO is discussed. /HI MASS RATIO AT HIGH REDSHIFT In this paper we exploit local correlations observed at 500pc resolu-tion to estimate the redshift evolution of the H /HI ratio. An impor-tant caveat of this procedure is the validity on galactic scales of cor-relations observed on sub-galactic scales or, in other words, whetherintegrated quantities can be estimated from spatially-resolved rela-tions. In recent years several studies have indeed revealed that the"main" correlations involved in the star formation process, the MSof star forming galaxies and the molecular gas Main Sequence(MGMS, e.g. Lin et al. 2019) have very similar slopes when ana-lyzed on sub-galactic or galactic scales (e.g. Hsieh et al. 2017; Linet al. 2019; Cano-Díaz et al. 2019; Enia et al. 2020). To estimate the redshift evolution of 𝑀 H / 𝑀 HI in MS galaxies weproceed as follows. We define the variable 𝑌 as the log of 𝑀 H / 𝑀 HI and express it as a function the the total gas mass ( 𝑀 gas = 𝑀 H + 𝑀 HI ) and SFR : 𝑌 = log 𝑀 H 𝑀 HI = 𝑓 ( 𝑀 gas , SFR ) . (1)It follows that: 𝑑𝑌𝑑 log ( + 𝑧 ) = 𝑚 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) + 𝑚 𝑑 log ( SFR ) 𝑑 log ( + 𝑧 ) , (2)where: 𝜕𝑌𝜕 log 𝑀 gas (cid:39) 𝑚 and 𝜕𝑌𝜕 log ( SFR ) (cid:39) 𝑚 , (3)with 𝑚 describing the conversion of HI into H and 𝑚 the oppositeconversion from H to HI due to photo-dissociation. From Tacconiet al. (2018) we have that, at fixed stellar mass, 𝑑 log 𝑀 H 𝑑 log ( + 𝑧 ) = . 𝑀 H , not to 𝑀 gas . For the redshift evolution ofthe SFR (at fixed stellar mass) we adopt the scaling from Speagleet al. (2014): 𝑑 logSFR 𝑑 log ( + 𝑧 ) = . . (5)Therefore, Equation (2) becomes: 𝑑𝑌𝑑 log ( + 𝑧 ) = 𝑚 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) + . 𝑚 . (6)As a next step, we need to derive 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) . Since we have:log 𝑀 HI = log 𝑀 H − 𝑌, (7)then: 𝑀 gas = 𝑀 H × ( + − 𝑌 ) , (8)and the derivative becomes: 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) = 𝑑 log 𝑀 H 𝑑 log ( + 𝑧 ) + 𝑑 log ( + − 𝑌 ) 𝑑 log ( + 𝑧 ) = . − (cid:18) + 𝑀 H 𝑀 HI (cid:19) − 𝑑𝑌𝑑 log ( + 𝑧 ) (9)where the first derivative is given by Equation (4). Therefore, usingEquation (9), Equation (6) becomes: 𝑑𝑌𝑑 log ( + 𝑧 ) = − 𝑚 (cid:18) + 𝑀 H 𝑀 HI (cid:19) − 𝑑𝑌𝑑 log ( + 𝑧 ) + . 𝑚 + . 𝑚 (10)Now we integrate the left and right sides of Equation (10) between 𝑧 = 𝑧 : ∫ 𝑧 (cid:18) + 𝑚 + 𝑌 (cid:19) 𝑑𝑌 = ( . 𝑚 + . 𝑚 ) ∫ log ( + 𝑧 ) 𝑑 log ( + 𝑧 ) (11)By solving the integrals of the left and right sides of Equation (11)we get: ( + 𝑚 )( 𝑌 𝑧 − 𝑌 ) + 𝑚 ( log ( + 𝑌 ) − log ( + 𝑌 𝑧 )) = ( . 𝑚 + . 𝑚 ) log ( + 𝑧 ) , (12)where the subscript 0 ( 𝑧 ) refers to the values at redshift 0 ( 𝑧 ).Thus, this equation is meant to describe the redshift evolution ofthe H /HI mass ratio at fixed stellar mass. To proceed with thenumerical solution of Equation (12), we need the values of 𝑚 and 𝑚 that we obtain from Figure 8 of M20, reported here in the leftpanel of Figure 1. This Figure shows how the ratio of molecular toatomic hydrogen varies as a function of the total gas surface densityand distance from the spatially resolved MS relation, Δ MS , which isdefined as the difference between log(SFR) of a region and its MSvalue at the same stellar mass. Inside galaxies, the H /HI mass ratiois very strongly correlated with the total gas surface density andanticorrelated with the local SFR, as quantified by Δ MS . In M20 weinterpret this anticorrelation as evidence that the UV radiation fromrecently formed, massive stars has the effect of photo-dissociatingmolecular hydrogen, a manifestation of the self-regulating nature ofthe star formation process.We estimate 𝑚 by fitting the relation between log ( H / HI ) and log Σ gas along the MS ( Δ MS ∼ 𝑚 , we calculate the slope of the log ( H / HI )− Δ MS relation at fixedlog Σ gas , considering a narrow range of log Σ gas values where data MNRAS000
0. Thesesources have been observed in at least 18 bands from the far ultravi-olet (FUV) to the far infrared (FIR). We used the photometric datafrom FUV to FIR to run SED fitting with MAGPHYS (da Cunhaet al. 2008) on cells of 500pc × UV ) and obscured (SFR IR ) contribu-tions. To this aim, SFR UV and SFR IR have been computed using thescaling relations of Bell & Kennicutt (2001) and Kennicutt (1998),respectively, where the UV and IR luminosities ( 𝐿 UV and 𝐿 IR ) areevaluated from the best-fit SED (see Enia et al. 2020). Finally, asthese sources are included in the HERACLES (Leroy et al. 2009)and THINGS (Walter et al. 2008) surveys, they have been observedin CO(2-1) and HI at 21 cm. Hereafter, we make use of the H es-timated using 𝛼 CO = 4.3 𝑀 (cid:12) ( K · km · s − pc ) − (e.g. Bolatto et al.2013). Details on how the HI and H maps at 500pc resolution wereobtained can be found in M20, where the consistency of the resultsusing a constant or metallicity-dependent 𝛼 CO is discussed. /HI MASS RATIO AT HIGH REDSHIFT In this paper we exploit local correlations observed at 500pc resolu-tion to estimate the redshift evolution of the H /HI ratio. An impor-tant caveat of this procedure is the validity on galactic scales of cor-relations observed on sub-galactic scales or, in other words, whetherintegrated quantities can be estimated from spatially-resolved rela-tions. In recent years several studies have indeed revealed that the"main" correlations involved in the star formation process, the MSof star forming galaxies and the molecular gas Main Sequence(MGMS, e.g. Lin et al. 2019) have very similar slopes when ana-lyzed on sub-galactic or galactic scales (e.g. Hsieh et al. 2017; Linet al. 2019; Cano-Díaz et al. 2019; Enia et al. 2020). To estimate the redshift evolution of 𝑀 H / 𝑀 HI in MS galaxies weproceed as follows. We define the variable 𝑌 as the log of 𝑀 H / 𝑀 HI and express it as a function the the total gas mass ( 𝑀 gas = 𝑀 H + 𝑀 HI ) and SFR : 𝑌 = log 𝑀 H 𝑀 HI = 𝑓 ( 𝑀 gas , SFR ) . (1)It follows that: 𝑑𝑌𝑑 log ( + 𝑧 ) = 𝑚 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) + 𝑚 𝑑 log ( SFR ) 𝑑 log ( + 𝑧 ) , (2)where: 𝜕𝑌𝜕 log 𝑀 gas (cid:39) 𝑚 and 𝜕𝑌𝜕 log ( SFR ) (cid:39) 𝑚 , (3)with 𝑚 describing the conversion of HI into H and 𝑚 the oppositeconversion from H to HI due to photo-dissociation. From Tacconiet al. (2018) we have that, at fixed stellar mass, 𝑑 log 𝑀 H 𝑑 log ( + 𝑧 ) = . 𝑀 H , not to 𝑀 gas . For the redshift evolution ofthe SFR (at fixed stellar mass) we adopt the scaling from Speagleet al. (2014): 𝑑 logSFR 𝑑 log ( + 𝑧 ) = . . (5)Therefore, Equation (2) becomes: 𝑑𝑌𝑑 log ( + 𝑧 ) = 𝑚 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) + . 𝑚 . (6)As a next step, we need to derive 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) . Since we have:log 𝑀 HI = log 𝑀 H − 𝑌, (7)then: 𝑀 gas = 𝑀 H × ( + − 𝑌 ) , (8)and the derivative becomes: 𝑑 log 𝑀 gas 𝑑 log ( + 𝑧 ) = 𝑑 log 𝑀 H 𝑑 log ( + 𝑧 ) + 𝑑 log ( + − 𝑌 ) 𝑑 log ( + 𝑧 ) = . − (cid:18) + 𝑀 H 𝑀 HI (cid:19) − 𝑑𝑌𝑑 log ( + 𝑧 ) (9)where the first derivative is given by Equation (4). Therefore, usingEquation (9), Equation (6) becomes: 𝑑𝑌𝑑 log ( + 𝑧 ) = − 𝑚 (cid:18) + 𝑀 H 𝑀 HI (cid:19) − 𝑑𝑌𝑑 log ( + 𝑧 ) + . 𝑚 + . 𝑚 (10)Now we integrate the left and right sides of Equation (10) between 𝑧 = 𝑧 : ∫ 𝑧 (cid:18) + 𝑚 + 𝑌 (cid:19) 𝑑𝑌 = ( . 𝑚 + . 𝑚 ) ∫ log ( + 𝑧 ) 𝑑 log ( + 𝑧 ) (11)By solving the integrals of the left and right sides of Equation (11)we get: ( + 𝑚 )( 𝑌 𝑧 − 𝑌 ) + 𝑚 ( log ( + 𝑌 ) − log ( + 𝑌 𝑧 )) = ( . 𝑚 + . 𝑚 ) log ( + 𝑧 ) , (12)where the subscript 0 ( 𝑧 ) refers to the values at redshift 0 ( 𝑧 ).Thus, this equation is meant to describe the redshift evolution ofthe H /HI mass ratio at fixed stellar mass. To proceed with thenumerical solution of Equation (12), we need the values of 𝑚 and 𝑚 that we obtain from Figure 8 of M20, reported here in the leftpanel of Figure 1. This Figure shows how the ratio of molecular toatomic hydrogen varies as a function of the total gas surface densityand distance from the spatially resolved MS relation, Δ MS , which isdefined as the difference between log(SFR) of a region and its MSvalue at the same stellar mass. Inside galaxies, the H /HI mass ratiois very strongly correlated with the total gas surface density andanticorrelated with the local SFR, as quantified by Δ MS . In M20 weinterpret this anticorrelation as evidence that the UV radiation fromrecently formed, massive stars has the effect of photo-dissociatingmolecular hydrogen, a manifestation of the self-regulating nature ofthe star formation process.We estimate 𝑚 by fitting the relation between log ( H / HI ) and log Σ gas along the MS ( Δ MS ∼ 𝑚 , we calculate the slope of the log ( H / HI )− Δ MS relation at fixedlog Σ gas , considering a narrow range of log Σ gas values where data MNRAS000 , 1–6 (2021) edshift Evolution of 𝑀 H / 𝑀 HI Figure 1.
Left panel : log ( H / HI ) - log Σ gas plane, adapted from Figure 8 of M20. Each cell is color-coded according to the average value of Δ MS . The bluesolid line is the best fit to the cells having an average value of Δ MS in the range [-0.2,0.2]; the slope of this best fit is 𝑚 . The gray shaded area includes thevalues for which we compute Δ MS as a function of H /HI ratio, as shown in the right panel : the slope of the best fit (blue solid line) give us 𝑚 . exist over the widest range of the H / HI mass ratio (the vertical greyregion in the left panel), hence offering the best possible estimate ofthis derivative. The best fit returns a slope of − .
55 (right panel ofFigure 1). We adopt these two derivatives as proxies for 𝑚 and 𝑚 as defined by Equations (3), based on the aforementioned similaritybetween the corresponding spatially resolved and global relations.For simplicity, in the following we assume 𝑚 =1.5 and 𝑚 = − . 𝑑𝑌𝑑 log ( + 𝑧 ) = − . + . (cid:16) + 𝑀 H2 𝑀 HI (cid:17) − . (13)Equation (13) implies that the redshift derivative of 𝑌 is alwaysnegative, i.e., the phase equilibrium shifts in favour of HI in highredshift galaxies. This comes from the SFR increasing with redshiftfaster than the molecular gas mass, see the above Equations (4) and(5). Let us consider three limiting cases. If 𝑀 H largely dominatesover 𝑀 HI , then the denominator in Equation (13) is ∼ − .
35. If 𝑀 HI largely dominates, the derivative be-comes -0.54. Finally, if the two phases are nearly equal in mass thedenominator is ∼ − . 𝑚 = . 𝑚 = − . 𝑀 H , / 𝑀 HI , = 1/3, 1 and 3, i.e., three typical values of theH /HI mass ratio within the optical radius of MS galaxies in thelocal Universe (Casasola et al. 2020). Galaxies that at 𝑧 = 𝑀 H / 𝑀 HI , implying that by 𝑧 ∼ dominatedwill tend to show a slightly steeper evolution, to reach 𝑀 H / 𝑀 HI ∼ 𝑧 =
2. We notice that lower values than 3.5 in Equation (5)can be found in the literature: they would imply a flatter evolutionof 𝑀 H / 𝑀 HI compared to our results. With the data for the five galaxies in the sample of M20 we an-alyze how 𝑀 HI and 𝑀 H are linked on scales of 500 pc. We ob- serve a slightly super-linear correlation between log( 𝑀 HI / 𝑀 ★ ) andlog( 𝑀 H / 𝑀 ★ ), characterized by a slope of 1.13, a Spearman coef-ficient of 0.62 and 𝑝 -value ∼ 𝑀 HI 𝑀 ★ ∝ .
13 log 𝑀 H 𝑀 ★ (14)and the correlation is shown in Figure 3. We note that one of ourfive galaxies, NGC5194 (M51), has a significantly flatter slope andsmaller Spearman coefficient, and interestingly is the only galaxy inthe sample to be experiencing an interaction (with M51b) as well asthe only one to have T-type = 4 (while the rest of the galaxies haveT-type between 5.2 and 5.9). We decided to keep NGC5194 in oursample for consistency with Method 1, but noting that the slope forthe remaining four galaxies is slightly steeper (1.24). This correla-tion gives us the possibility to estimate the evolution of 𝑀 HI / 𝑀 ★ with 𝑧 just by considering the evolution of the molecular gas (atfixed stellar mass), expressed in Equation (4). Hence, Equation (14)becomes: 𝑀 HI 𝑀 ★ ∝ (cid:18) 𝑀 H 𝑀 ★ (cid:19) . ∝ ( + 𝑧 ) . × . (15)and thus: 𝑀 H 𝑀 HI ∝ ( + 𝑧 ) . × ( + 𝑧 ) − . ∝ ( + 𝑧 ) − . . (16)The trend expressed by Equation (16) is shown in Figure 2 (dashedlines) for the three values of 𝑀 H / 𝑀 HI at 𝑧 = 0 used in Method 1:1/3, 1 and 3. The two methods appear to give basically consistentresults, with only a modest evolution of 𝑀 H / 𝑀 HI with redshiftin favor of HI, which is more pronounced in Method 1 (we notethat a steeper slope than the one expressed in Equation (15) wouldincrease the consistency between the two methods). This agreementmay not be surprising, as the two methods are in fact more similarthan they appear. Indeed, in Method 1 the effect of the SFR on 𝑀 H / 𝑀 HI is treated explicitly, whereas in Method 2 it is implicit inthe 𝑀 H - 𝑀 HI correlation. MNRAS , 1–6 (2021)
Morselli L. et al.
Figure 2.
Redshift evolution at fixed stellar mass of the H /HI mass ratio, obtained applying Method 1 (solid lines) and Method 2 (dashed lines), for threedifferent values of ( 𝑀 H / 𝑀 HI ) 𝑧 = = 1/3 (turquoise), 1 (gray) and 3 (black). The values obtain from the HI detection of C20 at z=1.04 are marked with thewhite-to-black colored bar, with the gradient indicating variations of the fraction of HI inside the optical radius. The values estimated from the correlations ofZhang et al. (2020) at z=0, 0.83 and 1.23 are indicated with the yellow-to-purple colored bar, with the gradient indicating the variations in stellar mass. We compare these trends with the recent detection of HI in emissionin 𝑧 ∼ . · 𝑀 (cid:12) , the mean HI mass is 1 . · 𝑀 (cid:12) . Tocompute the mean H /HI mass ratio in the galaxies observed by C20we proceed as follows. We consider the mean molecular-to-stellarmass ratio in the local Universe for galaxies with 𝑀 ★ ∼ 𝑀 (cid:12) to be ∼ . · 𝑀 (cid:12) is ∼ . · 𝑀 (cid:12) . By applying the scaling fromTacconi et al. (2018), expressed by Equation (4), the expected mean 𝑀 H in 𝑧 = ∼ . · 𝑀 (cid:12) , hence: (cid:18) 𝑀 H 𝑀 HI (cid:19) 𝑧 = = . · . · = . . (17)This value is obtained assuming that the HI detected by C20 ( 𝑀 HI tot )lies completely within the optical radius ( 𝑅 ) of the galaxies in thesample (i.e., 𝑓 R25 = 𝑀 HIR25 𝑀 HItot =
1, with 𝑀 HI R25 the HI mass within 𝑅 ). The average beam of the observations described in C20 isbetween 30 and 60 kpc, thus it is likely that a certain fraction of theobserved HI lies outside the optical radius, resulting in an under-estimation of the H /HI mass ratio within the optical radius. Thewhite-to-black bar in Figure 2 represents the estimate of 𝑀 H / 𝑀 HI at 𝑧 =1, assuming that C20 have sampled a fraction of HI insidethe optical radius, varying from 100 per cent (white) to 25 percent (black). In particular, we find that 𝑀 HI > 𝑀 H at 𝑧 ∼ 𝑓 R25 > .
4. For 𝑓 R25 < .
4, a value consistent with the 𝑧 = 𝑀 H > 𝑀 HI at 𝑧 ∼
1, but even when this fraction is only 20%, 𝑀 H is only a factor of 2higher than 𝑀 HI .In Figure 2 we also include, with yellow-to-purple vertical bars,the recent estimates of 𝑀 HI obtained by Zhang et al. (2020) fromthe local correlations between log 𝑀 HI 𝑀 ★ and the ( 𝑁𝑈𝑉 − 𝑟 ) color,which is a proxy for the specific SFR. We report their results atthree different redshifts: 𝑧 = 0, 0.83 and 1.23. As above, we use theevolution of 𝑀 H with redshift as given by Equation (4) to estimate 𝑀 H at the three redshifts, while 𝑀 HI is obtained for M ★ varyingbetween 10 𝑀 (cid:12) (in yellow in Figure 2) and 10 . 𝑀 (cid:12) (in purplein Figure 2). It is worth noting that the HI estimates used in Zhanget al. (2020) do not refer to values within the optical radius; this isclear at 𝑧 =
0, where the estimates of 𝑀 H / 𝑀 HI are significantlysmaller than those of Casasola et al. (2020) computed within theoptical radius.While our two methods and the one of Zhang et al. (2020) yieldsimilar results (in that they suggest a non-vanishing HI contributionat high redshift), it is worth recapping the underlying physical mo-tivations of each of them. Method 1 is built on the observed scalingof the 𝑀 H / 𝑀 ★ ratio with redshift, Eq. (4), and attempts to includethe effect of photo-dissociation of the H molecules by young stars.Method 2 assumes that the local correlation between 𝑀 H and 𝑀 HI holds at all redshifts, and the rationale of it is that if galaxies havemore H they must have also more HI, which is the necessary step toform H . The method of Zhang et al. (2020) assumes that the localcorrelation between 𝑀 HI and the ultraviolet-optical colour (a SFRproxy) holds also at all redshifts: as the SFR increases with redshift,so has to do 𝑀 HI as well. We notice that only our two methods usethe observed increasing trend with redshift of the H to stellar massratio. MNRAS000
0, where the estimates of 𝑀 H / 𝑀 HI are significantlysmaller than those of Casasola et al. (2020) computed within theoptical radius.While our two methods and the one of Zhang et al. (2020) yieldsimilar results (in that they suggest a non-vanishing HI contributionat high redshift), it is worth recapping the underlying physical mo-tivations of each of them. Method 1 is built on the observed scalingof the 𝑀 H / 𝑀 ★ ratio with redshift, Eq. (4), and attempts to includethe effect of photo-dissociation of the H molecules by young stars.Method 2 assumes that the local correlation between 𝑀 H and 𝑀 HI holds at all redshifts, and the rationale of it is that if galaxies havemore H they must have also more HI, which is the necessary step toform H . The method of Zhang et al. (2020) assumes that the localcorrelation between 𝑀 HI and the ultraviolet-optical colour (a SFRproxy) holds also at all redshifts: as the SFR increases with redshift,so has to do 𝑀 HI as well. We notice that only our two methods usethe observed increasing trend with redshift of the H to stellar massratio. MNRAS000 , 1–6 (2021) edshift Evolution of 𝑀 H / 𝑀 HI Figure 3.
Correlation between log( 𝑀 HI / 𝑀 ★ ) and log( 𝑀 H / 𝑀 ★ ) at 500 pcresolution, for the 5 galaxies of M20. The best fit correlation (solid orangeline) has a slope of 1.13 and a Spearman coefficient of 0.62. All the above results rely on extrapolations from local trends thatmay or may not hold when applying them to high redshift, thus atthis stage we consider the results tentative. Yet, in all methods theH /HI mass ratio is expected to decrease with redshift, contraryto the notion that it would increase, with H dominating at highredshift. Thus, these results suggest that HI cannot be neglectedat high redshift and we discuss below some implications for ourunderstanding of high redshift galaxies.The first one concerns star formation, namely the gas depletiontime 𝑀 gas /SFR and the star formation efficiency (SFE). For lack ofdirect evidence on the HI mass, the H mass has been generally usedas a proxy for the total gas mass. If our projections are correct, and ifsome (if not all) of the HI observed within the optical disk of galaxiescomes from H photo-dissociation, then the total gas depletion timeshould be at least a factor of ∼ to stars conversion is not a one-way process inside galaxies,but rather a cycle in which part of the H in converted back to HI.Thus, the total gas depletion time is a more informative quantitycompared to the molecular gas depletion time.The second implication concerns the contribution of HI to thetotal baryonic mass inside the stellar disk of high redshift galaxies.Even when ignoring HI, spatially-resolved dynamical studies haveshown that 𝑧 ∼ , as it is at 𝑧 ∼
0, then these galaxiesmay turn out even more baryon dominated than estimated thusfar. Similarly, a higher gas fraction due to the addition of the HIcomponent would lower the Toomre parameter, making disks moreprone to clump formation instabilities.For a direct assessment of the HI content of star forming galax-ies at high redshifts we will have to wait for the planned surveyswith the Square Kilometer Array (SKA). Indeed, ultra-deep SKA1surveys may probe massive galaxies (with 𝑀 HI > ∼ 𝑀 (cid:12) ) up to 𝑧 < ∼ . ) that will detect thementioned amount of HI up to 𝑧 ∼ . ) that will reach 𝑧 ∼ .
7. These observations should be amplysufficient to check the extent to which our projections are correct.
ACKNOWLEDGMENTS
We are grateful to the anonymous referee for a careful considerationof our manuscript, to Leslie Hunt for useful comments on an earlyversion, and to Lucia Rodríguez-Muñoz, Arianna Renzini, BhaskarAgarwal and Hannah Übler for fruitful discussion and valuable in-puts. LM acknowledges support from the BIRD 2018 research grantfrom the Universit`a degli Studi di Padova. AE and GR acknowledgethe support from grant PRIN MIUR 2017 - 20173ML3WW 001.
DATA AVAILABILITY
The derived data underlying this article will be shared on reasonablerequest to the corresponding author.
REFERENCES
Allen R. J., Heaton H. I., Kaufman M. J., 2004, ApJ, 608, 314Bell E. F., Kennicutt Robert C. J., 2001, ApJ, 548, 681Bigiel F., Leroy A., Walter F., Brinks E., de Blok W. J. G., Madore B.,Thornley M. D., 2008, AJ, 136, 2846Blyth S., et al., 2015, in Advancing Astrophysics with the Square KilometreArray (AASKA14). p. 128 ( arXiv:1501.01295 )Bolatto A. D., Wolfire M., Leroy A. K., 2013, ARA&A, 51, 207Cano-Díaz M., Ávila-Reese V., Sánchez S. F., Hernández-Toledo H. M.,Rodríguez-Puebla A., Boquien M., Ibarra-Medel H., 2019, MNRAS,488, 3929Casasola V., et al., 2020, A&A, 633, A100Catinella B., Cortese L., 2015, MNRAS, 446, 3526Catinella B., et al., 2018, MNRAS, 476, 875Chowdhury A., Kanekar N., Chengalur J. N., Sethi S., Dwarakanath K. S.,2020, Nature, 586, 369Clark C. J. R., et al., 2018, A&A, 609, A37Davies J. I., et al., 2017, PASP, 129, 044102Enia A., et al., 2020, MNRAS, 493, 4107Fernández X., et al., 2016, ApJ, 824, L1Genzel R., et al., 2017, Nature, 543, 397Genzel R., et al., 2020, ApJ, 902, 98Giovanelli R., et al., 2005, AJ, 130, 2613Hayward C. C., Hopkins P. F., 2017, MNRAS, 465, 1682Hsieh B. C., et al., 2017, ApJ, 851, L24Hunt L. K., Tortora C., Ginolfi M., Schneider R., 2020, A&A, 643, A180Kanekar N., Sethi S., Dwarakanath K. S., 2016, ApJ, 818, L28Kennicutt Robert C. J., 1998, arXiv.org, pp 189–232Krumholz M. R., Dekel A., McKee C. F., 2012, ApJ, 745, 69Leroy A. K., et al., 2009, AJ, 137, 4670Lin L., et al., 2019, ApJ, 884, L33Masters K. L., et al., 2019, MNRAS, 488, 3396Meyer M. J., et al., 2004, MNRAS, 350, 1195Morselli L., et al., 2020, MNRAS, 496, 4606Scoville N., et al., 2017, ApJ, 837, 150Speagle J. S., Steinhardt C. L., Capak P. L., Silverman J. D., 2014, ApJS,214, 15Sternberg A., Le Petit F., Roueff E., Le Bourlot J., 2014, ApJ, 790, 10Tacconi L. J., et al., 2018, ApJ, 853, 179Walter F., Brinks E., de Blok W. J. G., Bigiel F., Kennicutt Robert C. J.,Thornley M. D., Leroy A., 2008, AJ, 136, 2563Zhang W., Kauffmann G., Wang J., Chen Y., Fu J., Wu H., 2020, arXive-prints, p. arXiv:2011.04500da Cunha E., Charlot S., Elbaz D., 2008, MNRAS, 388, 1595MNRAS , 1–6 (2021)
Morselli L. et al.
This paper has been typeset from a TEX/L A TEX file prepared by the author. MNRAS000