EDP Sciences
Free Access
Issue
A&A
Volume 610, February 2018
Article Number L16
Number of page(s) 4
Section Letters to the Editor
DOI https://doi.org/10.1051/0004-6361/201732395
Published online 01 March 2018

© ESO 2018

1 Introduction

High-resolution multi-object spectrographs have made it possible to retrieve accurate information about the kinematics, chemical composition, ages and other fundamental physical properties for a very large ensemble of stars, putting strong constraints on the formation and evolution of our galaxy (e.g. Bensby et al. 2014; Hayden et al. 2015; Lindegren et al. 2016). Nevertheless, as soon as we enter into the broad field of extragalactic astronomy, our comprehension about how galaxies have formed and evolved with time becomes more uncertain, mostly because their global observed physical properties have usually been inferred from their integrated stellar light or emission lines.

Following the pioneering works by Garnett (1990) and Vila Costas & Edmunds (1993), emission line diagnostics are capable of measuring C, N, and O abundances, and are being applied to extragalactic systems, providing also the spatial distribution of the elemental abundances (e.g. Pilyugin et al. 2010; Sánchez-Menguiano et al. 2016; Belfiore et al. 2017; Toribio San Cipriano et al. 2017; Amorín et al. 2017). This blossoming of observational data will open a new window of interest for galactic astroarchaeology studies. In this respect, chemical evolution models – when included within cosmological hydrodynamical simulations and combined with spectrophotometric stellar population synthesis codes – can uniquely provide a powerful tool to unveil the past evolutionary history of galaxies. Only by dealing with the abundances of specific chemical elements (and hence only by shelving the generic observable “metallicity”), can we define sensitive chemical diagnostics of the entire star formation history (SFH) of the galaxy.

In this Letter, we present the results of our study on a sample of star-forming disc galaxies within an up-to-date cosmological hydrodynamical simulation, including a detailed routine for the galaxy chemical evolution (Kobayashi et al. 2007). We focus our analysis on the evolution of three examples of disc galaxies, representative of the different stellar systems within our catalogue, and show how their gas-phase chemical abundances of C, N, and O evolve as functions of the galaxy SFH. In Sect. 2, we briefly summarise the main characteristics and the setup of our simulation. In Sect. 3, we present the inhomogeneous enrichment and the global gas-phase CNO abundance evolution of our disc galaxies. Finally, in Sect. 4, we draw our conclusions.

2 Simulation model and assumptions

All the results of this Letter are obtained by running and analysing the outcome of a cosmological hydrodynamical simulation based on the GADGET-3 code (Springel 2005), which uses the smoothed particle hydrodynamics (SPH) method. Our code was developed by Kobayashi et al. (2007) to include the relevant physical processes related to the star formation activity and chemical evolution of galaxies (see also Taylor & Kobayashi 2014 for the details).

Chemical enrichment. The chemical evolution routine is the same as the one originally developed by Kobayashi (2004) for another SPH code, and gives an excellent agreement with the elemental abundance distributions as observed in the Milky Way (MW; Kobayashi & Nakasato 2011). In our model, we adopt the stellar yield set by Kobayashi et al. (2011) with failed Supernovae (SNe) at mass m ≥ 25 M and metallicity Z ≥ 0.02, which are required from the observations of nearby SNe (e.g. Smartt 2009; see also the theoretical study of Müller et al. 2016). In failed SNe, all synthesised O and heavier elements fall back into the black hole, except for H, He, C, N, and F, which are synthesised in the outermost layers of the SN ejecta (Kobayashi et al., in prep.). Following Kobayashi et al. (2006), half of stars with m ≥ 25 M are assumedto explode as hypernovae. For the asymptotic giant branch (AGB) stars, we assume the same yields as in Kobayashi et al. (2011). Finally, we assume the Kroupa (2008) initial mass function (IMF), defined in the mass range 0.01≤ m ≤ 120 M, and the same stellar lifetimes as in Kobayashi (2004).

The photo-chemical code. In order to compute the average luminosity-weighted ages and metallicities of the galaxies in the simulation, as well as their half-light radii, we have developed a stellar population synthesis model1, which is based on the one originally developed by Vincenzo et al. (2016b), assuming the PARSEC stellar evolutionary tracks (Bressan et al. 2012).

Simulation setup. Our assumed cosmological model is the standard Λ -cold dark matter Universe, with the same cosmological parameters as given by the nine-year Wilkinson Microwave Anisotropy Probe (Hinshaw et al. 2013). We follow the evolution of a cubic volume of the Universe, with periodic boundary conditions and side U = 10 Mpc h−1, in comoving units. The initial conditions are the same as in Kobayashi et al. (2007), with improved resolution; NDM = Ngas = 1283 both for dark matter (DM) and gas particles. The feedback from the star formation activity and chemical evolution (i.e. the thermal energy and the nucleosynthetic products from stellar winds and SN explosions) is distributed to a number NFB = 576 of neighbour gas particles, weighted by the smoothing kernel. These parameters are determined to match the observed cosmic star formation rate (SFR; Hopkins & Beacom 2006; Madau & Dickinson 2014). Our simulation has the following mass resolutions: MDM ≈ 3.097 × 107 h−1 M and Mgas = 6.09 × 106 h−1 M, for the DM and gas mass components, respectively. Finally, we assume a gravitational softening length ϵgas ≈ 0.84 h−1 kpc, in comoving units.

3 Results

We analyse the outcome of our simulation, by creating a catalogue of stellar systems at redshift z = 0. We firstly make use of the friend-of-friends group-finding algorithm with adaptive hierarchical refinement ROCKSTAR (Behroozi et al. 2013) to create a catalogue of DM halos at redshift z = 0; from this catalogue, we isolate and analyse all the embedded stellar systems which are spatially clustered and contain a sufficient number of both star and gas particles, so as to draw well sampled chemical abundance patterns. We exclude all the stellar systems which are undergoing a major merger at the present time. In total, we select ~ 35 galaxies with the aforementioned broad and simple criteria, lying within DM halos with virial mass in the range 11.0 ≤ log (MDM/M) ≤ 13.0 dex (Vincenzo et al., in prep.).

In this Letter, we present the results of our analysis for three examples of star-forming disc galaxies, which have total stellar masses roughly of the same order of magnitude but are representative of distinct characteristic SFHs that are seen in the other disc galaxies in the simulation. In Table 1, we summarise the redshift evolution of the main predicted characteristics of our reference galaxies. The predicted stellar age distribution function of our galaxies is shown in Fig. 1. When passing from Galaxy C to Galaxy A, the emitted light at the present time is dominated by older stellar populations; the chemical enrichment also differs from galaxy to galaxy and all our reference galaxies end up at the present time by having almost the same average metallicity (⟨Z⟩~ 0.01), following the observed mass-metallicity relation.

In Fig. 2, we show our predictions at redshift z = 0 for the gas-phase log(N/O) versus log(O/H) + 12 abundance pattern in Galaxy A. Our results with failed SNe (light blue points) are compared with the same simulation but with the original Kobayashi et al. (2011)’s yields without failed SNe (orange points); each simulation point in the figure corresponds to one location within the galaxy. To understand the abundance patterns, we briefly recall that nitrogen is mostly synthesised by intermediate-mass stars by hot-bottom burning during the AGB phase. Most of the nitrogen from AGB stars is secondary and its stellar yields steadily increase as functions of the initial stellar metallicity. All mass ranges of stars can also produce primary N, converted from 12C by the CNO cycles (see also Vincenzo et al. 2016a and the references therein). Finally, oxygen is mostly produced by core-collapse SNe with very short lifetimes ( ~106 yr).

In Fig. 2, with failed SNe, we find an increasing trend (of “secondary” production) at the metal-rich side of the N/O versus O/H diagram, predicting a similar slope as in observations. At very high metallicity, the increase may be slightly smaller than in the observed relation, which may be because stellar yields from super-solar AGB stars are not available (Vincenzo et al. 2016a). We also note that N and O might also be affected by dust depletion, which is not taken into account in our work. If we do not assume failed SNe, we have a much flatter slope at high metallicity, which is inconsistent with observations.

In Fig. 3, we show how the N/O–O/H relation evolves with redshift in our reference galaxies. There is a large scatter particularly at low O/H, which is due to inhomogeneous enrichment; in fact, there are gas particles that have gathered the nucleosynthetic products of only a few core-collapse SNe or AGB stars in the past (see Kobayashi & Nakasato 2011 for more details). When the galaxy is still experiencing its earliest evolutionary stages (M ~ 108 M), the gas-phase N/O ratios are almost constant with metallicity, suggesting the presence of a plateau of “primary” production (z ~ 4 for Galaxy A; z ~ 2 for Galaxy B). This is not produced in classical one-zone chemical evolution models with instantaneous mixing approximation, which systematically underestimated N/O at very low metallicity and had to assume an artificial primary N production by massive stars (e.g. Matteucci 1986; Chiappini et al. 2005; Vincenzo et al. 2016a) to reproduce the observations at low metallicity in dwarf galaxies (Pilyugin et al. 2010; Berg et al. 2012; Izotov et al. 2012; James et al. 2015), damped-Lyα (DLA) absorption systems (e.g., Pettini et al. 2002, 2008), and MW halo stars (e.g., Spite et al. 2005).

We note that our predicted N/O plateau resides at larger O/H than in the HII regions of nearby dwarf galaxies, meaning that our reference galaxies were already more metal-rich at high redshifts than nearby dwarf galaxies. In the adopted yields, N/O from core-collapse SNe is much lower than in our high-redshift galaxies ( ~–1.45 dex), but in the simulation there is already a significant contribution from AGB stars, which is the difference from the one-zone models. At lower redshifts (M ≳ 109 M), the plateau exists but the value quickly reaches log(N/O) ~ 1.2 dex with a larger contribution from AGB stars. Galaxy C is already at this phase when it is formed by a major merger at z ~ 2. We note that the plateau value would be even higher with the effects of stellar rotation.

In Fig. 4, we show how the average SFR-weighted gas-phase CNO abundances vary as functions of redshift for our reference galaxies; this diagram can be used for constraining the galaxy SFH. We only show the predictions of our simulation for the evolutionary times when the galaxy total stellar mass M(z) ≥ 108 M. In Fig. 4a the increase of the average N/O ratios turns out to be larger, corresponding to a more intense star formation episode (high-z in Galaxy A; low-z in Galaxy C). This is the typical imprint of strong chemical enrichment of secondary N from AGB stars.

In Fig. 4b and c, we show how the average SFR-weighted gas-phase log(C/O) and log(C/N) evolve with redshift. Carbon is mostly synthesised by low-mass stars with masses in the range 1 ≲ m ≲ 3 M (e.g. see Kobayashi et al. 2011). As for nitrogen, carbon can also be produced by core-collapse SNe, on short time-scales after the star formation event. With the IMF weighting, half of the C producers in galaxies are low-mass stars. The metallicity dependence of the C yields is opposite with respect to the metallicity dependence of the secondary N; while the produced secondary N steadily increases as a function of metallicity, the C yield steadily decreases.

In Fig. 4c, at high redshift, in the earliest galaxy evolutionary stages, we mostly witness at the chemical enrichment from massive stars with high C/N. The averageC/N is subsequently predicted to sharply decrease at the onset of massive AGB stars. Then, C/N temporarily increases due to the C production of low-mass AGB stars; finally, C/N decreases again because of the opposite metallicy-dependence of the C and N stellar yields. Currently, observations at high redshifts are available for C/O ratios (Amorín et al. 2017), which roughly overlap withour model curves. The very low C/O ratios in Amorín et al. (2017) are consistent with the rapid and intense star formation in Galaxy A, but at much lower redshifts.

Table 1

Main properties of our reference galaxies as functions of redshift.

thumbnail Fig. 1

Distribution of the present-day total stellar mass as a function of the star particle age. Each panel corresponds to one of our reference galaxies.

Open with DEXTER
thumbnail Fig. 2

Predicted gas-phase log(N/O) vs. log(O/H) + 12 abundance pattern in Galaxy A with failed SNe (light blue points; this work) and without failed SNe (orange points). The solid grey line corresponds to the average relation obtained from various observations by Dopita et al. (2016, see references therein), the grey points with error bars toa compilation of data of individual HII regions in blue diffuse dwarf galaxies (Berg et al. 2012; Izotov et al. 2012; James et al. 2015), and the dark grey crosses to Belfiore et al. (2017) from a sample of spatially resolved galaxies from the MaNGa survey.

Open with DEXTER
thumbnail Fig. 3

The redshift evolution of the predicted gas-phase log(N/O) vs. log(O/H) + 12 abundance patterns within our three reference galaxies. Galaxies A, B, and C clearly show different chemical evolution of the ISM abundances with redshift. We compare our simulation with the observed average N/O–O/H relation from Dopita et al. (2016, solid grey line).

Open with DEXTER
thumbnail Fig. 4

(a–c) average SFR-weighted gas-phase log(N/O), log(C/O) and log(C/N) as functions of redshift. We only show the predictions of our simulation for the redshifts when the galaxy stellar mass M (z) ≥ 108 M. The black star symbol on each trackmarks the redshift when M ≈ 109 M. The grey triangles with the error bars correspond to the observational data of Amorín et al. (2017).

Open with DEXTER

4 Conclusions

In this Letter, we have shown how the gas-phase C, N, and O abundances evolve as functions of the galaxy SFH. All our results have been obtained by analysing three examples of star-forming disc galaxies within a full cosmological hydrodynamical simulation, including up-to-date nucleosynthetic yields and a standard (non-variable) IMF.

Our simulation predicts an almost flat trend of N/O versus O/H at high redshifts, which is caused by an inhomogeneous enrichment of the galaxy ISM with a significant contribution of AGB stars at low metallicity. Our simulation represents an improvement with respect to previous chemical evolution models, both because the N/O–O/H relation was studied with simple one-zone models with instantaneous mixing approximation and becausewe did not assume any artificial primary N production by massive stars (Matteucci 1986; Chiappini et al. 2005; Vincenzo et al. 2016a). We note that the N/O plateau would be even higher if rapidly rotating massive stars were included.

We conclude that failed SNe are a very promising scenario to explain the observed trend of N/O vs. O/H. The inclusion of super-solar AGB stars might represent a further improvement of our model, since they can produce even larger amounts of secondary N as the metallicity increases, helping to explain the further increase of N/O at very high O/H.

The evolution of the gas-phase CNO abundances is driven by the chemical enrichment from AGB stars and core-collapse SNe, which create an increasing trend of the N/O and C/O towards higher metallicities by present. The predicted increase of the average N/O and C/O with redshift turns out to be faster within galaxies experiencing a more intense star formation activity during the considered redshift interval. Finally, in the earliest stages of galaxy evolution, we predict the largest C/N ratios and the lowest N/O ratios.

It is not easy to obtain both C and N for the same sample of galaxies with current observations, since the N/O and C/O ratios can only be estimated for z ≲ 2.5 (optical rest frame) and z ≳ 2 − 3 (UV rest frame), respectively; in the near future, however, it will be possible to have measurements of C/N within galaxies at z ≲ 6 with the Near-Infrared Spectrograph on the James Webb Space Telescope.

Acknowledgments

We thank V. Springel for providing GADGET-3, and R. Maiolino, R. Amorín, P. Taylor, and S. Smartt for stimulating discussions. We thank an anonymous referee for his/her comments. FV acknowledges funding from the UK STFC through grant ST/M000958/1. This research has made use of the DiRAC HPC facility in Durham, UK, supported by STFC and BIS.

References


All Tables

Table 1

Main properties of our reference galaxies as functions of redshift.

All Figures

thumbnail Fig. 1

Distribution of the present-day total stellar mass as a function of the star particle age. Each panel corresponds to one of our reference galaxies.

Open with DEXTER
In the text
thumbnail Fig. 2

Predicted gas-phase log(N/O) vs. log(O/H) + 12 abundance pattern in Galaxy A with failed SNe (light blue points; this work) and without failed SNe (orange points). The solid grey line corresponds to the average relation obtained from various observations by Dopita et al. (2016, see references therein), the grey points with error bars toa compilation of data of individual HII regions in blue diffuse dwarf galaxies (Berg et al. 2012; Izotov et al. 2012; James et al. 2015), and the dark grey crosses to Belfiore et al. (2017) from a sample of spatially resolved galaxies from the MaNGa survey.

Open with DEXTER
In the text
thumbnail Fig. 3

The redshift evolution of the predicted gas-phase log(N/O) vs. log(O/H) + 12 abundance patterns within our three reference galaxies. Galaxies A, B, and C clearly show different chemical evolution of the ISM abundances with redshift. We compare our simulation with the observed average N/O–O/H relation from Dopita et al. (2016, solid grey line).

Open with DEXTER
In the text
thumbnail Fig. 4

(a–c) average SFR-weighted gas-phase log(N/O), log(C/O) and log(C/N) as functions of redshift. We only show the predictions of our simulation for the redshifts when the galaxy stellar mass M (z) ≥ 108 M. The black star symbol on each trackmarks the redshift when M ≈ 109 M. The grey triangles with the error bars correspond to the observational data of Amorín et al. (2017).

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.