Changes in the Na D₁ Absorption Components of η Carinae Provide Clues on the Location of the Dissipating Central Occulter

Luminous blue variables (LBVs) are evolved massive stars that are near the Eddington Limit and are variable on timescales of days to centuries. The variability of these stars is such that when a star is observed being hotter, it is fainter in the optical and brighter in the ultraviolet. The majority of active LBVs tend to fluctuate between spectral types of late-O or early B- and F-type supergiants, while maintaining a nearly constant bolometric luminosity, with only small changes (see for example the work on AG Car; Groh et al. 2009, 2011). These stars have been observed to have stellar eruptions where they eject up to several solar masses of material, creating circumstellar ejecta around many of these objects. The population of LBVs was recently documented in Richardson & Mehner (2018) and their properties have been reviewed in the past by Humphreys & Davidson (1994) and van Genderen (2001).

η Carinae (η Car; HD 93308; HR 4210; previously η Argus) is a massive binary star within the Carina Nebula and is sometimes considered a prototype of the LBVs. The system is relatively close to us with a distance of approximately only 2.3 kpc (Smith 2006). η Car has undergone two observed eruptions, causing significant amounts of gas and other debris to obscure the system itself, although there may have been additional eruptions in the past (Kiminki et al. 2016). This material has resulted in a bipolar shell of ejecta surrounding the system from the Great Eruption in the 1840s and from the second Lesser Eruption in the 1890s, known as the Homunculus and the Little Homunculus, respectively. The binary is in a long-period 5.54 yr orbit (2022.7 ± 0.3 days; Damineli 1996; Damineli et al. 1997; Teodoro et al. 2016) that is highly eccentric (e = 0.9; e.g., Damineli et al. 2000; Grant et al. 2020; Grant & Blundell 2022). The primary star provides some ionization to the surrounding gas, such as causing Fe ii emission lines, while the secondary star further ionizes the Homunculus and the Little Homunculus. During the system's periastron passage, the secondary star passes into the far side of the optically thick wind of the primary star, which causes the ionization of the gas in our line of sight to drop to normal levels for the primary star. This variability, where the ionization of the gas temporarily drops, has been referred to as spectroscopic events in the literature. As noted in several studies (e.g., Melnick et al. 1982; Damineli et al. 1998; Davidson et al. 2005; Johansson et al. 2005; Gull et al. 2006; Richardson et al. 2010, 2015), the change in ionization causes increased absorption of low-ionization species from the ejecta, especially at velocities associated with the Little Homunculus (near −145 km s⁻¹).

In recent years, the central source of the η Car system has been observed to be brightening. Unlike most normal LBV activity, such as the S Doradus cycles, the central source is brightening in both the optical and the ultraviolet in tandem. There are two competing hypotheses related to this variability. The first is that the central star is quickly evolving as it settles from the Great Eruption (Mehner et al. 2010, 2011, 2015, 2019; Davidson et al. 2018; Martin et al. 2021). This suggestion relies on wind lines, such as Hα (Mehner et al. 2010), changing with a long-term evolution, which could be interpreted as changes in the mass-loss properties. The second hypothesis is that there was an clump in the line of sight that has been dissipating and/or moving out of our line of sight (Hillier & Allen 1992; Hillier et al. 2001; Damineli et al. 2019, 2021). This suggestion is supported by the large amount of added extinction that is needed in the models of the stellar spectra by both Hillier et al. (2001, 2006) and Groh et al. (2012). As this extinction was fairly gray, a change in the extinction should allow most wavelengths to change at similar rates.

We aimed to use the variability of the Na D absorption component associated with the Lesser Eruption to understand how it relates to the documented changes in the spectrum of η Car. The ionization potential of the neutral Na is ≈ 5.14 eV, which corresponds to a wavelength of ∼2412 Å. Thus the ionization of neutral sodium is achieved by radiation in the near-ultraviolet. However, the ground state photoionization cross section of Na i has a Cooper minimum ¹¹ near 2000 Å, ¹² and consequently a large fraction of the ionizations will occur from radiation short of 1800 Å. Given the large ultraviolet flux of η Car, we expect Na to mostly be singly ionized. As a consequence, the time variation of the Na i equivalent widths will be primarily governed by the timescale for the flux variations, and thus the orbital timescale around the periastron. Thus we would expect the Na equivalent width (∝column density) variations and the ultraviolet flux variations to track each other. In Section 2, we describe 12 yr of observations of the system from the Cerro Tololo Inter-American Observatory (CTIO) 1.5 m telescope and its echelle spectrographs, along with some complementary observations from the Hubble Space Telescope and the Space Telescope Imaging Spectrograph (HST/STIS). Section 3 details the measurements we made of three absorption components in the Na D spectrum. Section 4 discusses these findings, and we conclude this study in Section 5.

Our observations come from a time series of high-resolution spectroscopy of η Car taken with the CTIO 1.5 m Small and Moderate Aperture Research Telescope System (SMARTS) telescope. All data were taken with a fiber centered on the object, with a projected angular size of 27. The first observations were taken from 2008–2010 with the fiber-fed spectrograph having a resolving power of R ∼ 40,000, and were described by Richardson et al. (2010, 2015). Since the 2009 periastron passage, a time series of higher signal-to-noise and higher-resolution spectra were obtained with the SMARTS telescope and the new CHIRON spectrograph with modes having a resolving power of 80,000–90,000 for these observations (Tokovinin et al. 2013; Paredes et al. 2021). For the CHIRON time series, the data were described by Richardson et al. (2016), with more observations having the same characteristics taken in the intervening time, especially around the 2020 periastron passage. In total, we analyze 418 spectra taken with the CTIO 1.5 m telescope in this study.

Data were previously reduced for all of the data previously published using techniques described in Richardson et al. (2010, 2015) for the fiber echelle. This was done with standard techniques using IRAF. ¹³ For the CHIRON data, the data were processed with the standard pipeline described recently by Paredes et al. (2021). The observations also included a standard observation of HR 4468 (B9.5Vn) or μ Col (O9.5V), which was used to empirically derive the blaze function and then remove it through the division with custom scripts in either IDL or with python. This was done in orders without spectral lines, while the remaining orders (primarily around Hα or Hβ) were interpolated to match adjacent orders.

The observations of the most recent periastron passage began with CTIO/CHIRON data collected on 2019 October 15 and concluded with data collected on 2020 March 18 when the telescope was closed due to the COVID-19 pandemic. These new observations did provide higher signal-to-noise Na D profiles than ever before, owing to the ongoing increase in the visual brightness of η Car (e.g., Damineli et al. 2019, 2021).

In order to study the Na D complex, we first had to remove contamination from water in Earth's atmosphere. We utilized the telluric measurements from Wallace & Livingston (2003) to construct a telluric template at the same resolution and sampling rate as the CHIRON data. This template could then be artificially strengthened or weakened to match each spectrum, under the assumption that the relative strengths of different lines remains fixed. This method was used over the entire Na D region, including regions with no intrinsic features from η Car to ensure that the remaining spectrum was intrinsic to the source and the intervening material from the ejecta and the interstellar medium.

When dealing with spectra of η Car, it is important to use a reliable clock with which we can compare observations to the binary ephemeris. To that end, we adopt the ephemeris from Teodoro et al. (2016), which should be close to the timing of the periastron in 2014, and is given by the equation relating the phase ϕ to the Heliocentric Julian Date (HJD) by

$$\begin{eqnarray*}&&\phi =\displaystyle \frac{({HJD}-2,456,874.4)}{2022.7}+13.\end{eqnarray*}$$

In this context, the addition of the 13 in the equation means that the 2014 periastron was the thirteenth observed spectroscopic minimum since the first reported by Gaviola (1953).

We found an absorption component that we suspected to be interstellar in its origin, and we collected spectra of neighboring stars with the same high-resolution mode of the CHIRON spectrograph in 2021 July. The objects observed were HD 93204, HD 93205, HD 303308, and CPD 59° 2628. Using the same methodology as with η Car, the spectra were reduced and then telluric corrected.

We also used archived spectra taken with the HST/STIS in the E140M echelle mode with a nominal resolving power, R = 45,800. ¹⁴ The three spectra were recorded at phases ϕ = 11.119, 13.194, and 14.170, which is in the early recovery to the high-ionization state of the nebular structures. These data were pipeline-processed and then downloaded from MAST.

In Figure 1, we show selected spectra from our time series of the system. There are several features in this figure, which we will describe in detail in the following section. The interstellar components near 0 km s⁻¹ (and near −300 km s⁻¹ for the Na D₂ component in the shown Na D₁ reference frame) are both very close to black in our data, but some small uncorrectable amount of scattered light in the spectrographs makes the profile appear to have a minimum near 1% of the normalized continuum light.

Zoom In Zoom Out Reset image size

There is a dramatic disappearance of a feature at −145 km s⁻¹ over our full data set. This and other velocities were fit for a radial velocity by means of a Gaussian profile or by measuring the point of minimum flux on select data sets with high signal-to-noise. We caution that these profiles are not always simple Gaussians, so we adopt these velocities to better describe the variations of the individual components. Richardson et al. (2010) described a similar component in the Hα profile that appeared and strengthened across the 2009 periastron passage, and Richardson et al. (2015) noted changes in the Na D complex similar to those of Hα during the 2009 periastron passage. In the most recent periastron passage, the feature's center has moved to bluer velocities, and the profile has bifurcated so that a second component at −168 km s⁻¹ is now visible near the periastron. Furthermore, there is a component at +87 km s⁻¹ that is seen in all of the CHIRON spectra, without much variability present. For our analysis, we concentrate on the Na D₁ λ5896 line, as the Na D₂ λ5890 line is strongly blended with both the D₁ components originating in the Homunculus and the He i λ5876 line.

For the features that we discuss in this paper, we measure the equivalent width defined as

$$\begin{eqnarray*}&&{W}_{\lambda }=\int \displaystyle \frac{{F}_{C}(\lambda )-{F}_{\lambda }(\lambda )}{{F}_{C}(\lambda )}d\lambda ,\end{eqnarray*}$$

where F_C is the continuum flux and F_λ is the flux as a function of wavelength (λ). The continuum around these features is difficult to define as the normalized spectrum shows broad Na D emission from the outer portions of the stellar wind. Therefore, we renormalized each observation around the feature to measure the strength at that time of the observation. In order to estimate the errors in W_λ of the measurements, we adopted the method of Vollmann & Eversberg (2006), where the error depends on the dispersion of the spectrum (Δλ), the number of spectral pixels integrated over (n), and the signal-to-noise ratio (S/N) of the spectrum per resolution element shown by the equation

$$\begin{eqnarray*}&&\sigma =\displaystyle \frac{n{\rm{\Delta }}\lambda -{W}_{\lambda }}{S/N}.\end{eqnarray*}$$

The errors for the weak feature at −168 km s⁻¹ appear to be a bit larger than the scatter around the trends in the points even when we carefully measured the quantities needed in the equation above. Such a trend could mean that we overestimated the errors; however, since we had to locally normalize the data around the feature, within the emission component of the Na D complex, there are additional errors involved in that process. There are also sources of error that are not accounted for, such as how the strength could change on a changing slope of the emission profile. We have therefore kept these error bars at this level, since a measurement to a precision of <5 mÅ is difficult to achieve with the star and this instrument.

3.1. The Absorption Component at −145 km s⁻¹

Richardson et al. (2015) found that the absorption component near −145 km s⁻¹ increases near the periastron passage, similar to the Hα component that strengthens at the same time (Richardson et al. 2010; Damineli et al. 2021). We measured these spectra, along with the CTIO/CHIRON data, in the region surrounding the variable −145 km s⁻¹ feature. The exact range changed with time as the feature moved to bluer velocities in more recent observations, but the range of integration was typically about 25 km s⁻¹ wide and moved with the absorption with time. The equivalent width measurements are tabulated in the Appendix (Table 1). Gull et al. (2006) reported that ultraviolet features of this velocity seemed to appear stronger at the periastron, yet remain fairly consistent in equivalent width (W_λ ) throughout the remainder of the orbit.

From our long time series of measurements in Figure 2, the absorption at −145 km s⁻¹ is observed to be weakening over this time period. In Figure 1 we see that this feature seems to move to bluer velocities over this same time period, while the interstellar absorption components stay constant in velocity as expected. By a time shortly after the 2014 periastron event, this feature was indiscernible in the spectra despite the higher signal-to-noise (S/N ∼ 150–200 per pixel). We see similarities in the two events for 2009 and 2014 when phased to the binary clock (Figure 3). The change in the ultraviolet flux during the periastron passage is about a factor of 3 according to Gull et al. (2021), consistent with the changes seen in this line. The 2020 event is discussed in the following subsection.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It was reported in Gull et al. (2006) that the −145 km s⁻¹ features in the ultraviolet would change intensity across the spectroscopic minimum. The absorption feature in Fe ii, as measured at ϕ = 0.739, appears during the broad maximum. However, when at a minimum, the feature appeared stronger and was also accompanied by two more features at velocities near −145 km s⁻¹. As noted by Gull et al. (2006), this indicates a brief drop in ionization of the Little Homunculus in our line of sight caused by the shadowing of that region by the primary's wind intervening in front of the secondary star.

3.2. The Absorption Component at −168 km s⁻¹

During the recent 2020 periastron passage, an absorption component at −168 km s⁻¹ appeared. This absorption may have been present in earlier periastra, but would have been blended with a much stronger component at −145 km s⁻¹. It was also seen in the Hα line as observed by Damineli et al. (2021), although at the velocity of −180 km s⁻¹. This change in the Hα and Na D absorptions associated with the periastron passages is likely due to a changing ionization structure in the Little Homunculus. A similar velocity component was reported in the low-ionization ultraviolet lines by Gull et al. (2006) during the low state corresponding to the periastron passages. This component, while independent of the −145 km s⁻¹ feature, seems to have become similar in strength to its changing velocity counterpart at −145 km s⁻¹, and could have been undetected in previous cycles due to the stronger component at −145 km s⁻¹. We compare the blended component to that of the −145 km s⁻¹ component in Figure 3, where we see that the beginning of the increase in absorption shows a similar time frame from the pre-event minimum to maximum as the previous events.

While not as prominent as the −145 km s⁻¹ feature had been, the −168 km s⁻¹ feature became significantly strong enough to take equivalent width measurements during the last periastron. Measurements for the −168 km s⁻¹ feature were taken between the wavelengths of 5892.45 Å and 5893.20 Å, corresponding to the radial velocities of −180.82 km s⁻¹ and −142.62 km s⁻¹, respectively. This integration includes the smaller contribution of the −145 km s⁻¹ component remnants, which can both be seen in the spectrum at phase 14.010 in Figure 1. We attempted to measure the two features at −145 and −168 km s⁻¹ independently, but the features were blended, weak, and non-Gaussian in nature. The resulting measurements were deemed unreliable by both their component errors and visual inspection of the spectra compared with the measurements.

As seen in Figure 4 and tabulated in Table 3, there is variability in W_λ of the −168 km s⁻¹ feature. This originates in a decreasing level of ionization near the periastron. The increased activity near ϕ = 14.00 resulted in the steady climb in magnitude of W_λ , which then begins to decrease near ϕ = 14.018. This process is similar to that of the −145 km s⁻¹ feature. The timing of the disappearance of the −168 km s⁻¹ feature was interrupted by the temporary closure of the facilities at CTIO.

Zoom In Zoom Out Reset image size

3.3. The Absorption Component at +87 km s⁻¹

Following the telluric correction, a previously unnoticed absorption component was recognized at ∼+87 km s⁻¹. This component, seen in Figure 1, indicates that material is moving toward η Car and would represent an inflow of material if originating in the gas surrounding η Car. A second, identical component at −217 km s⁻¹ in the Na D₁ reference frame is also observed and represents the Na D₂ transition at the same velocity.

Equivalent width measurements for the +87 km s⁻¹ feature were taken between the wavelengths of 5897.55 Å and 5897.74 Å, corresponding to the radial velocities of 78.87 km s⁻¹ and 88.53 km s⁻¹ and are shown in Figure 5 and tabulated in Table 2. Each measurement was made within the bounds of 0.19 Å or 9.66 km s⁻¹. We notice a small change around the 2014 periastron passage, which may be explained by the change in the V-band brightness across the periastron, such as that observed by Damineli et al. (2019). In this regard, we see an increase in continuum flux leading up to the event, followed by a dip in the flux after the periastron. With the equivalent width depending on the continuum, as shown by the associated equivalent width equation earlier in this section, we see that the small differences at that time can be explained by the change in flux. Furthermore, oscillations at times away from the periastron may reflect changes in the photometric brightness, such as those recently documented by Richardson et al. (2018). However, as noted by Richardson et al. (2010), usually changes in the equivalent widths due to the continuum are primarily a concern for large flux lines, such as the Hα line. It is also possible that these small changes are due to changes in the Na D emission in the region of this absorption component, which could alter our measurements. The exact amount of variation that could be intrinsic to this component should therefore be very small when compared to the large changes in the underlying flux from the system along with the variability of the broad Na D wind emission features present in this region of the spectrum.

Zoom In Zoom Out Reset image size

Our measurements of Na D₁ absorption components from η Car can be used to illuminate some recent changes in the spectrum and the photometric record. In particular, Damineli et al. (2019, 2021) found that there are long-term changes related to the circumstellar extinction in the line of sight that should conclude in ∼2032. Our results show that the Na D absorption from the Little Homunculus is much weaker and likely absent at times away from the periastron, in broad agreement with the results from Damineli et al. (2019). The spectral modeling of the system, first accomplished with HST spectroscopy and the CMFGEN non-LTE radiative transfer code by Hillier & Miller (1998) and Hillier et al. (2001), required 2.0 mag of gray extinction in the line of sight to model both the stellar (wind) spectrum and achieve the correct line strengths for the narrow line emission from the surrounding Weigelt knots. The system also had 5 mag of extinction in the V band in the line of sight toward the binary. We show a cartoon diagram of the most likely geometry that could accommodate the observations in Figure 6.

Zoom In Zoom Out Reset image size

If we assume that the component near −145 km s⁻¹ originates in the Little Homunculus as noted in several studies (e.g., Melnick et al. 1982; Damineli et al. 1998; Davidson et al. 2005; Johansson et al. 2005; Gull et al. 2006; Richardson et al. 2010, 2015), then we can begin to interpret the variability and character of the components at −145 and −168 km s⁻¹. First, we consider the variability of the component at −145 km s⁻¹. Richardson et al. (2010) describe the variability of the component at the same velocity for the Hα transition. For Hα, Richardson et al. (2010) postulate that during the spectroscopic minimum near the periastron passage, the secondary star would pass behind the primary star, and thus drop the ionizing radiation in our line of sight. This drop in ionization would then relax the gas and allow lower ionization transitions to be observed in absorption.

These same principles can explain the changes we see in both the −145 km s⁻¹ component (Figure 2) across the 2009 and 2014 periastron passages and the −168 km s⁻¹ component (Figure 4) during the recent 2020 periastron event. The −168 km s⁻¹ absorption was only previously observed in six low-ionization lines (Mg i, Mg ii, Cr ii, Mn ii, Fe ii, and Ni ii) in the near-ultraviolet by Gull et al. (2006) and only during the low-ionization state of the binary. In contrast, the −145 km s⁻¹ was observed in the same lines, as well as in He i, Ti ii, V ii, Co ii, and Ni ii, although Gull et al. (2006) reported it as −146 km s⁻¹.

In the far-ultraviolet, several transitions have been observed by HST/STIS, allowing us to make comparisons to the Na D₁ observations described here. For this discussion, we limit ourselves to the Al ii λ1671, Al iii λ1855, Si ii λ1527, and Si iv λ1394 transitions that were observed during the previous two periastron passages where we also have CHIRON spectroscopy. All of these ultraviolet transitions and Na D are resonance lines. In Figure 7, we show these transitions at phases 11.119, 13.194, and 14.170 so that the binary-induced phase differences are minimal, and we can examine the long-term evolution of the profiles. Each profile was normalized to the "continuum" flux level near 1483 Å recorded at phase ϕ = 14.170. Gull et al. (2022) demonstrate that normalization of the "continuum" is valid since virtually all continuum originates from deep within η Car-A.

Zoom In Zoom Out Reset image size

These profiles show that these species have evolved to show less absorption at −145 km s⁻¹, although these changes are amplified because of the normalization of these data. This shows a similar trend as the Na D observations, where recent observations since phase ∼13.5 have had no absorption present at that velocity except for a brief appearance at the cycle 14 periastron event. During the recent periastron, this absorption was seen at higher velocity and with a companion absorption component at −168 km s⁻¹. In the ultraviolet profiles, the lower-ionization Al ii and Si ii lines show that the absorption components at these velocities are also seen at higher velocities than in earlier cycles, such as just following the cycle 11 periastron passage.

We postulate that the absorption component was formed in a spatially unresolved dense clump within the Little Homunculus. This clump would respond to the ionization changes induced by the companion's orbit around the primary star. In this situation, the clump would be either moving out of our line of sight or dissipating with time. Both scenarios allow for a weaker absorption profile now compared to in the past, consistent with the observations of the ultraviolet resonance lines and the Na D profiles.

We used a curve of growth method to find the column density of the absorbing Na atoms during the times just prior to the 2009 event (phases near 11.95), near phase 12.5, and after the 2014 event at phase 13.1. We used these times to avoid times when the ionization changes, which would therefore change the column density of neutral sodium in our line of sight. At these three epochs, the equivalent widths were roughly 70 ± 12 mÅ, 52 ± 2.5 mÅ, and 14.7 ± 2.2 mÅ, respectively. With the known oscillator strength and wavelength of Na D₁, we can then calculate the values of W_λ /λ to obtain the values of F(τ) using the curve of growth from Spitzer (1978) with an assumed line width of 10 km s⁻¹, which is complementary to techniques used for the local interstellar medium (Welsh et al. 1990). Then, we derive a column density of 4.9 × 10¹¹ cm⁻², 3.5 × 10¹¹ cm⁻², and 0.8 × 10¹¹ cm⁻² for these three epochs, respectively, confirming the lower column density in recent times. This is in agreement with the changes in the ultraviolet flux, which has increased by about a factor of 10 in the same time period (Gull et al. 2021).

With a clump in the line of sight, we can also imagine that the angular size was relatively small to us. The neighboring Weigelt knots, separated from the central source by only about 02 (Weigelt & Ebersberger 1986), did not undergo the same amount of extinction as the central source according to the models of Hillier et al. (2001). As a result, an angularly small clump similar to the Weigelt knots, either within the Little Homunculus or in our line of sight, could provide the necessary extinction to provide the absorption lines and the added extinction toward the binary. As Damineli et al. (2019, 2021) have noted, the central occulter must be dissipating to explain the object's brightening and the lack of observed changes in the object's spectrum. As a result, the simplicity of a dissipating or moving clump in our line of sight would help explain many of the unusual observations of this system, although neither explanation can be fully refuted with the current data.

The component at −145 km s⁻¹ is also seen to shift in its velocity during the last several cycles as it was dissipating. The movement of the clump out of the line of sight or its thicker parts dissipating on the interior of the Little Homunculus shell, as illustrated in our Figure 6, would allow for an apparent shift in the radial velocity of the component by having the densest part of the clump being at different points in the outflow of the Little Homunculus. Furthermore, as the thickest part of the clump at the central velocity dissipated or moved out of the line of sight, the changing ionization front would penetrate into the Little Homunculus allowing for the weaker −168 km s⁻¹ to be visible during the most recent periastron passage while not observed in Na D in past events.

Moreover, Gull et al. (2016) recently reported on spatially resolved HST/STIS spectroscopy of the binary across an entire orbit of the system. The results indicate that the Weigelt knot B (A4 in the analysis of Weigelt & Ebersberger 1986) has fully dissipated or disappeared in the surrounding region of η Car. If the clump is similar ejecta as the Weigelt knots that were ejected during the Lesser Eruption, then the clump is traversing the region rather than dissipating, then it could now be providing additional extinction toward this clump. However, Weigelt B could also be dissipating with time, meaning that the clumps left over from the Lesser Eruption of η Car could all be beginning to dissipate around this time, as the Weigelt knots have been shown to be associated with the Lesser Eruption of η Car (Weigelt & Kraus 2012).

The origin of the component at +87 km s⁻¹ is simpler to explain. With an apparent lack of variability (Figure 5), it is hard to envision a situation in which that component would actually represent inflowing material from the surrounding winds and ejecta. We anticipated this being interstellar in origin so we tested our hypothesis in two ways. First, we see in the low-ionization lines of Al ii and Si ii that there is a component at roughly the same velocity as the +87 km s⁻¹ component of Na D₁ and Na D₂. This component is not seen to vary in its absolute absorption with the HST observations recorded over an interval of 16.5 yr (Figure 7). The +87 km s⁻¹ component is not present in Si iv λ1394, Al iii λ1855, nor in Si ii λ1533. The absorbing cloud contains only singly ionized ions and is sufficiently cool that no Si ii ions populate the low-lying energy level at 287 cm⁻¹.

These features were reminiscent of the ultraviolet lines of other stars in the Carina Nebula complex observed by Walborn et al. (2002). Strong absorption has been previously observed in the nearby CPD −59°2603, as noted in Danks et al. (2001). Small line widths of high-velocity features, as well as column density variations, indicate structures with transverse velocities when compared to radial velocities. Walborn et al. (2002) depict these absorptions from both 1997 and 1999 observations. These data show, in total, nine strong, low-ionization absorptions within our line of sight. The strongest features appear in lower-velocity regions, but it should be noted that dominant components of high-ionization states appear at intermediate and negative velocities. The high-ionization absorption corresponds to the global expansion of the H ii region. This expansion is further investigated in Walborn et al. (2007). These observations of neutral sodium at varying velocities for stars in the Carina Nebula raise the possibility that they are formed in low-density outflowing gas from the region, perhaps originating from a past supernova or the collective winds of the cluster.

To further test the interstellar origin of this component, we observed three of the Carina stars observed by Walborn et al. (2002) with the CHIRON spectrograph at the same resolving power as our η Car observations, along with CPD −59° 2628, which is very close to η Car in the sky. ¹⁵ For these observations of Na D shown in Figure 8, we see many similarities in the interstellar lines.

Zoom In Zoom Out Reset image size

The primary interstellar lines near 0 km s⁻¹ all show similar components for these stars. As noted by Walborn et al. (2002), many of these stars have lines at high velocities. The Na D components of the neighboring CPD −59°2603 show a positive velocity component at a lower velocity than that of η Car, with a similar profile showing a double absorption. The observations of η Car show a similar double component of absorption. These similarities with other stars within the Carina Nebula show that they could be explained by a complicated foreground absorption complex, perhaps from a past supernova or cluster wind that shows complexity across the region, but the physical cause for these complex interstellar absorptions is beyond the scope of this paper.

Our study of the Na D complex of η Car has found several interesting trends that relate to other observations of the system.

• 1.  
  The absorption component near −145 km s⁻¹ has been observed moving to higher negative velocities over the past decade. Simultaneously, the strength of this absorption has weakened overall, becoming indiscernible by 2018. This happened parallel to the brightening of the star while the nebula has remained more consistent in its brightness. This points to the central occulter being a spatially unresolved clump in the Little Homunculus.
• 2.  
  We find a new absorption in the spectra that is near +87 km s⁻¹, which seems to be interstellar in origin. Similarly shaped structures are seen in neighboring stars, but at different velocities. The variability of this component can likely be attributed to changes in the continuum flux and Na D wind emission during the periastron passages.
• 3.  
  The interstellar feature at +87 km s⁻¹ is also seen in some low-ionization ultraviolet resonance lines.
• 4.  
  To further disentangle if the occulter is/was in the Little Homunculus or in another clump interior to the Little Homunculus, future observations of the Na profile at high resolution and in reflected light from various lines of sight in the Homunculus could illuminate this discussion.

Many measurements in this project were originally made by Lucas St-Jean, and we thank him for these early contributions that led to these results. The CTIO observations are the result of many allocations of telescope time. We thank internal SMARTS allocations at Georgia State University (in 2008–2009), as well as NOIR Lab (formerly NOAO) allocations of NOAO-09B-153, NOAO-12A-216, NOAO-12B-194, NOAO-13B-328, NOAO-15A-0109, NOAO-18A-0295, NOAO-19B-204, NOIRLab-20A-0054, and NOIRLab-21B-0334. C.S.P. and A.L. were partially supported by the Embry-Riddle Aeronautical University Undergraduate Research Institute. E.S. acknowledges support from the Arizona Space Grant program. N.D.R., C.S.P., A.L., E.S., and T.R.G. acknowledge support from the HST GO Programs #15611 and #15992. A.F.J.M. is grateful to NSERC (Canada) for financial aid.