EARLY OPTICAL SPECTRA OF NOVA V1369 CEN SHOW THE PRESENCE OF LITHIUM

The light elements deuterium, 3-helium, 4-helium, and 7-lithium are synthesized in non-negligible amounts during the first few minutes of the initial cosmic expansion (Kolb & Turner 1990). For Big Bang nucleosynthesis in the standard model of cosmology and particle physics (SBBN), the predicted abundances for these elements depends only on one parameter, the baryon-to-photon density ratio. Recently, Planck data have determined the baryon density to excellent precision, leading to a primordial lithium abundance in the range $A(\mathrm{Li})=2.66-2.73$ ¹² (Coc et al. 2014). This value is significantly larger than $A(\mathrm{Li})$ ∼ 2.1–2.3 obtained for old metal-poor ([Fe/H] $\leqslant -1.4$) halo stars, whose distribution in the lithium abundance–metallicity diagram is almost flat and defines the so-called "Spite plateau" (Spite & Spite 1982; Bonifacio et al. 2007). These stars were long thought to share a common, primordial, Li abundance. The discrepancy may be explained by diffusion and turbulent mixing in the stellar interiors (Korn et al. 2006), which corrodes the primordial Li abundance and/or by a non-standard BBN scenario (Iocco et al. 2009). Yet, another puzzle exists. The abundance of lithium observed in the upper envelope of young metal-rich ([Fe/H]¹³ > −1.4) stars traces a growth of the Li abundance from the Spite plateau value to the meteoritic value $A(\mathrm{Li})=3.26\pm 0.05$ (Lodders et al. 2009) and even higher, which clearly indicates that lithium enrichment mechanisms must occur on Galactic scales.

In recent decades, several potential Li producers have been proposed on theoretical grounds. Among the possible lithium farms in the Galaxy, the most plausible astrophysical sources are represented by asymptotic giant branch (AGB) stars (Iben 1973), novae (Starrfield et al. 1978), and Galactic cosmic-ray spallation (via fragmentation of material due to the impact of accelerated protons, leading to expulsion of nucleons; Lemoine et al. 1998). Galactic chemical evolution models (D'Antona & Matteucci 1991; Romano et al. 2001; Travaglio et al. 2001; Prantzos 2012) produce a convincing match to the observations when all lithium producers described above are included.

The main channel for thermonuclear production of lithium inside stars is based on a particular sequence of events, known as the hot bottom burning and beryllium transport mechanism (Cameron & Fowler 1971). At first, there is formation of ⁷Be via the reaction ³He $+\alpha$, which requires large temperatures $T\geqslant {10}^{7}\;{\rm{K}}$ and occurs in the deep interior of stars. Outward convection mechanisms must concurrently transport the newly formed ⁷Be to external cooler regions, where $T\approx {10}^{6}\;{\rm{K}}$. Here, beryllium is able to capture electrons (free and bound), giving rise to ⁷Li with the additional formation of 0.86 MeV neutrinos. More recent computations (Ventura & D'Antona 2010) suggest that a large abundance of ⁷Li can thus be achieved in AGB stars, particularly with masses larger than $7{M}_{\odot }$. However, large lithium yields and, consequently, a significant enrichment in Li of the surrounding interstellar medium, can be obtained from these sources only if extensive mass loss is associated with the phases of maximum Li production (Romano et al. 2001).

In nova systems, the abundance of lithium is related to the thermonuclear runaway (TNR) scenario, which accounts for the physical explanation of ⁷Li and its ejection from the binary system. In TNRs, the expanding gas reaches velocities of the order of 400–4000 km s⁻¹, which results in escape from the system and enrichment of the interstellar medium with elements synthesized in the nova outburst (José & Hernanz 1998). This scenario remains uncertain because of the lack of direct detection of ⁷Li during nova outburst, although its presence was proposed long ago (Starrfield et al. 1978). Indeed, ⁷Li is easily destroyed by proton fusion at temperatures greater than $2.6\times {10}^{6}\;{\rm{K}}$, giving rise to two helium atoms. For this reason, ⁷Li detection in stars must focus on the external regions, where the temperature is sufficiently cool to allow non-depletion of ⁷Li.

In the literature, only upper limits to the abundance of ⁷Li in selected novae have been reported, e.g., HR Del, IV Cep, and NQ Vul (Friedjung 1979), while direct detections of lithium in symbiotic novae (e.g., T CrB, RS Oph, and V407 Cyg) are characterized by velocities of a few dozens km s⁻¹, which implies that lithium is not associated with the WD ejecta, but rather originates in the secondary giant star (Wallerstein et al. 2008; Brandi et al. 2009; Shore et al. 2011).

Recently, Tajitsu et al. (2015) announced the detection of ⁷Be features in the near-UV late spectra of V339 Del. Indeed, the Cameron–Fowler mechanism is expected to occur during the TNR (Starrfield et al. 1978), leading to an overproduction of beryllium at the epoch of the nova outburst, with a consequent decay to lithium after ∼50 days. The amount of Be observed is larger than typical values deduced from theoretical predictions of CO novae from TNR models (Truran 1981), representing the very early phases of the nova outburst.

We report here the detection of ⁷Li i 6708 Å in the early (first three weeks) high-resolution spectra of the "slow" (${t}_{2}=40\pm 1$ days, where t₂ corresponds to the time during which the nova declines by 2 mag) nova V1369 Cen (Nova Cen 2013), when the nova was in the optically thick phase.¹⁴ Optical high-resolution spectral observations started 4 days after the initial outburst. Early spectral data were obtained on days 7, 13, and 21 with the Fiber-fed Extended Range Optical Spectrograph (FEROS; R ∼ 48,000) mounted on the ESO-MPG 2.2 m telescope located on La Silla (Kaufer et al. 1999), and on days 4, 11, 16, and 18 with the PUC High Echelle Resolution Optical Spectrograph (PUCHEROS; R ∼ 20,000) mounted on the ESO 0.5 m telescope located at the Pontificia Universidad Catolica Observatory in Santiago (Vanzi et al. 2012). The spectra are all characterized by the presence of bright Balmer and typical Fe ii emission lines, which suggests that the nova was engulfed in its "iron curtain" phase (Shore 2012). The presence of many absorption features in the range 3700–4600 Å in the first two weeks (days 4, 7, 11, and 13) complicates the identification of the most common transitions detected in the optically thick phase of novae, which are identified from their respective P-Cygni absorptions. We identify multiple expanding velocities for each transition. In the first week (days 4 and 7), we measure from the Balmer lines, O i 7773–7, 8446 Å, and Fe ii (multiplet 42) lines two expanding systems with mean velocities of ∼ −550 and −1350 (with a maximum value of −1400) km s⁻¹, while for Na i we see only the system expanding at ${v}_{\mathrm{exp}}\sim -550$ km s⁻¹. In the second week (days 11 and 13), we identify two prominent expanding components at ${v}_{\mathrm{exp}}$ ∼ −550 and −1100 km s⁻¹ from P-Cygni profiles of all main transitions (Balmer, O i, Fe ii, and Na i D), while an additional faster component is observed at ${v}_{\mathrm{exp}}\sim -1900$ km s⁻¹ in Hβ and Hα lines alone. In subsequent spectra (day $\geqslant$ 21), we identify a broad and structured absorption at ${v}_{\mathrm{exp}}\approx -1600$ km s⁻¹ and two additional components at ∼ −550 and −750 km s⁻¹. The evolution of the observed P-Cygni absorptions of Hβ and Na i D2 is shown in Figure 1.

Zoom In Zoom Out Reset image size

We have identified, via cross-correlation procedure between observed narrow features and laboratory wavelengths¹⁵ , 319 (over a total of 424 absorptions at day 7) low-excitation (${E}_{\mathrm{in}}\leqslant 6$ eV) transitions of singly ionized heavy elements (Ba, Cr, Fe, Mn, Sc, Sr, Ti, V, and Y; see, for example, Williams et al. 2008), all of them characterized by an expanding velocity of ${v}_{\mathrm{exp}}$ ∼ −550 km s⁻¹ (see Figure 2). This procedure resulted in some non-identifications and degenerate-identifications, which we estimate to represent a small fraction (less than 20%) of the entire data set. A complete analysis will be published elsewhere (L. Izzo et al. 2015, in preparation).

Zoom In Zoom Out Reset image size

Although from day 11 an additional expanding component is detected in the P-Cygni profiles of the Na i doublet (see right panel in Figure 1), no further expanding components are identified for these heavy elements absorbing systems: the cross-correlation method provides a minor number of identifications considering a unique expanding velocity of ∼1100 km s⁻¹. If we consider the presence of both expanding components for these heavy elements, the number of acceptable identifications for the higher velocity components is a small fraction, ∼1%, of the identified lower velocity systems.

Among the many narrow transitions, we have identified many low-ionization neutral elements transitions (Fe, Ca, K), all belonging to the one expanding absorption system at the same velocity that was observed in ionized heavy elements, with the only exception being Na i D lines, which show additional expanding components from day 11 (see Figure 1). In particular, we note the clear presence of the resonance transitions of ⁷Li i 6708, Ca i 4227, and K i 7699 (see Figure 3), all of them with an expansion velocity of ${v}_{\mathrm{exp}}$ ∼ −550 km s⁻¹, in the spectra of days 7–18.

Zoom In Zoom Out Reset image size

2.1. Possible Alternative Identifications

Here, we discuss several plausible alternatives to our ⁷Li identification.

• 1.  
  We can exclude the possibility that this feature is due to a diffuse interstellar band (DIB; Herbig 1995; Bondar 2012) because (i) no known DIBs are located at 6695.6 Å; (ii) all observed heavy element narrow absorptions vanish after ∼20 days from the initial outburst when the nova continuum is still bright and DIBs should persist, being related to interstellar medium located in between the background exciting radiation and the observer; (iii) we observe variations in the observed wavelengths of all these narrow absorptions, an effect that DIBs do not show.
• 2.  
  Another possibility that must be considered is whether or not the 6695.6 Å absorption line is due to metal lines excited by resonance absorption of UV radiation that is absorbed in the iron curtain phase and then reprocessed at longer wavelengths (Johansson 1983; Shore 2012). For example, each of the resonance transitions of Li, Ca, and K could be alternatively identified with a transition of Fe or an s-process element at that wavelength, assuming that they are pumped by UV radiation (either continuum or lines). Particularly for the Li i 6707.8 Å resonance doublet, a possible identification could be Fe ii 6707.54 Å. In this case, one should also expect absorption features from the same lower energy level configuration ($3{d}^{6}{(}^{5}D)5p$), and from the same term ($4{Fo}-4G$), which results in two additional possible transitions (Fe ii 6769.27, 6811.49 Å) in the observed spectra. We have checked for¹⁶ their presence in the spectra of days 7 and 13,¹⁷ but we did not find any absorption feature corresponding to Fe ii 6769.27 and 6811.49 Å. This evidence disfavors a UV-pumped origin for the absorption at 6695.6 Å, even if considering the coupling between the transitions and the ejecta velocity field. Conversely, it supports our initial identification as expanding Li i 6708 Å.
• 3.  
  We compare the red spectral region of Nova Cen with the same region observed in GW Ori, a T Tauri star characterized by the presence of lithium (Bonsack & Greenstein 1960). GW Ori was observed with FEROS, and we have selected the observation of 2010 November 26. In Figure 4, we show the comparison of these two spectra after correcting the nova spectrum for the expanding velocity of ${v}_{\mathrm{exp}}=-550$ km s⁻¹. We clearly see the presence of broad Ca i 6718 Å and Li i 6708 Å in GW Ori giving more support to the identification of absorptions at 6695.6 and 6705.3 Å detected in V1369 with Li i 6708 and Ca i 6718.

Zoom In Zoom Out Reset image size

Under the assumption that this absorption is ⁷Li, we have estimated its ejected mass following Friedjung (1979). Since lithium, sodium, and potassium are alkali metals with very similar Grotrian diagrams and with respective resonant transitions, differing by only 0.25 eV for K–Li and Li–Na, we can assume that resonance doublets form under similar conditions. This assumption implies that the ratio of their optical thickness ${\tau }_{i}$ is related to their abundance ratio multiplied by their respective gf ratio. Following Spitzer (1998, Equations (3)–(48)), for the case of Li/Na, we have

$$\begin{eqnarray}&&\displaystyle \frac{{A}_{m}(\mathrm{Li})}{{A}_{m}(\mathrm{Na})}=\displaystyle \left(\frac{{W}_{\mathrm{Li}\;6708}}{{6708}^{2}}/\displaystyle \frac{{W}_{\mathrm{Na}\;{\rm{D}}2}}{{5890}^{2}}\right)\times \displaystyle \frac{{{gf}}_{\mathrm{Na}\;{\rm{D}}2}}{{{gf}}_{\mathrm{Li}\;6708}}\times \displaystyle \frac{{u}_{\mathrm{Li}}}{{u}_{\mathrm{Na}}},\end{eqnarray} \tag{ 1 }$$

where ${W}_{{\lambda }_{i}}$ is the measured equivalent width at the transition wavelength ${\lambda }_{i}$ and u_i is the atomic mass of the corresponding element (in our cases, ${u}_{\mathrm{Li}}=7$, ${u}_{\mathrm{Na}}=23$, and ${u}_{{\rm{K}}}=39$). For the ⁷Li i doublet blend, we have considered the value of log gf = −0.174, whereas the single components have, respectively, log gf D1 = −0.00177 and log gf D2 = −0.3028 (Kramida et al. 2013). We have obtained ${A}_{m}(\mathrm{Li})/{A}_{m}(\mathrm{Na})=3.2/100$ and ${A}_{m}(\mathrm{Li})/{A}_{m}({\rm{K}})=6.5/100$ on day 7 and ${A}_{m}(\mathrm{Li})/{A}_{m}(\mathrm{Na})=2.4/100$ and ${A}_{m}(\mathrm{Li})/{A}_{m}({\rm{K}})=5.2/100$ on day 13. The lithium log overabundances are 4.8 and 3.9 with respect to sodium and potassium solar abundances (Lodders et al. 2009). This result implies an overabundance of lithium in the nova ejecta of the order of 10⁴, which is large enough to explain the Galactic ⁷Li enrichment (Friedjung 1979). The mass of sodium and potassium ejected in novae can be computed in terms of solar mass by using the results of different nova composition models, characterized by different WD masses, accretion rates, and mixing degrees (José & Hernanz 1998). We have considered the results obtained for CO nova models, where the mass of ejected sodium ranges from 3.4 × 10⁻⁵ to 2.0 × 10⁻⁴ the mass of ejected hydrogen, while the ejected potassium varies from 5.1 to 7.2 × 10⁻⁶. The hydrogen ejected mass can be estimated both from the intensity of Hβ when the ejecta is completely optically thin, e.g., in late nebular phases (Mason et al. 2005), and the observed t₂ value (Della Valle et al. 2002). Both methods converge toward the value of an ejected hydrogen mass of ${M}_{{\rm{H}},\mathrm{ej}}\approx {10}^{-4}$ ${M}_{\odot }$. After combining the Na and K ejecta measurements with the respective lithium mass abundance ratio, we find the mass of lithium ejected by V1369 Cen to be in the range ${M}_{\mathrm{Li},1}$ = 0.3–4.8 $\times {10}^{-10}$ ${M}_{\odot }$. The ratios Li/Na and Li/K were determined on days 7 and 13 when only a small amount of ⁷Be decayed to ⁷Li. Hence, the measured amount of ⁷Li may be only a lower limit. However, TNRs that produce ⁷Be could start even years or months (see Starrfield, Iliadis, and Hix in Bode & Evans 2008) before the nova was discovered. Therefore, the most plausible scenario for V1369 Cen is the one in which TNRs started weeks/months before the optical detection, and therefore the measured abundance of lithium can be considered a good approximation of the total amount of lithium actually produced by the nova (or a firm lower limit). The crucial quantity needed to compute the global Li yield of the galactic nova population is the Galactic nova rate, which is known within a factor of two, R_N = 20–34 events yr⁻¹ (Della Valle & Livio 1994; Shafter 1997). With the yield obtained above, we derive for the lithium mass injected in the Milky Way by nova systems ${M}_{\mathrm{Li},\mathrm{tot}}$ ∼ 2–45 ${M}_{\odot }$ Gyr⁻¹. However, it is well known that the ejecta of "slow" novae are more massive than ejecta of "fast" novae, by an order of magnitude, ∼${10}^{-4}$ ${M}_{\odot }$ versus ∼${10}^{-5}$ ${M}_{\odot }$ (Della Valle et al. 2002); therefore, fast novae, which can form ∼30% of the nova population of the Milky Way (Della Valle & Duerbeck 1993), should contribute only marginally to the global ${}^{7}\mathrm{Li}$ yield. The above reported range of lithium mass decreases to ${M}_{\mathrm{Li},\mathrm{tot}}\leqslant 17$ ${M}_{\odot }$ Gyr⁻¹, for a rate of "slow" novae of 15–24 events yr⁻¹, in good agreement with the theoretical predictions (José & Hernanz 1998).

Figure 5 shows the A(Li) versus [Fe/H] observed relation compared with Galactic chemical evolution model results. The Galactic chemical evolution model used is an updated version of the chemical evolution model of Romano et al. (1999, 2001), which is based on the two-infall model of Chiappini et al. (1997) and includes AGB stars (Karakas 2010), super-AGB stars (Doherty et al. 2014), Galactic cosmic rays (Lemoine et al. 1998), and novae, as Li producers. In this model, the Galactic inner halo and thick disk form by accretion of gas of primordial chemical composition on a short timescale (∼1 Gyr). The gas is efficiently turned into stars as long as its density is above a critical threshold, below which the star formation stops. The thin disk forms out of a second episode of infall of gas of mainly extragalactic origin on longer timescales (7–8 Gyr in the solar neighborhood) and with lower star formation efficiency. It is worth emphasizing that the adopted mass assembly history is consistent with what is obtained for Milky Way–like galaxies in a full cosmological framework (Colavitti et al. 2008). The black continuous line in Figure 5 is the best fit to the data obtained with a model with all Li sources, starting from a primordial Li abundance of $A(\mathrm{Li})=2.3$ and by assuming that each "slow" nova (the adopted current slow novae rate is 17 events yr⁻¹) ejects on an average ${M}_{\mathrm{Li}}$ = 2.55 $\times {10}^{-10}$ ${M}_{\odot }$ in agreement with the measurement of ${M}_{\mathrm{Li}}$ = 0.3–4.8 $\times {10}^{-10}$ ${M}_{\odot }$ presented in this paper. The black dashed line shows the predictions of the same model when a high primordial Li abundance is adopted (see the Section 1). The red line is the best model with all Li factories but novae: it is clearly seen that novae are necessary to explain the late rise from the plateau value. The gray area indicates the uncertainties in the model predictions due to uncertainties in both the estimated Li yield from novae and the current slow nova rate: the upper (lower) boundary refers to the upper (lower) limit to the Li yield estimated in this paper and a maximum (minimum) current slow nova rate of 24(15) events yr⁻¹. The light green area similarly indicates the uncertainties in the model predictions when the maximum and minimum Li yields from José & Hernanz (1998) are assumed. Though the two areas partly overlap, it is clearly seen that the theoretical nova yields tend to underproduce Li in the Galaxy, while the semi-empirical yields estimated in this Letter give a better match with observed data points. Should the result presented here be confirmed by further observations of ⁷Li, classical novae would stand as one of the major Li producers on a Galactic scale.

Zoom In Zoom Out Reset image size

We thank the referee for constructive comments/criticisms that have improved the paper and Steven Shore, Marina Orio, and Paolo Molaro for useful discussions. We are grateful to Roland Gredel for DDT programme 091.A-9032 B.