The Sunburst Arc with JWST: II. Observations of an Eta Carinae Analog at $z=2.37$

The NIRCam data weakens the SN scenario suggested by Vanzella et al. (2020). Figure 2 shows the April 2023 NIRCam observation, and Godzilla still maintains similar brightness to that of the Sunburst LCE (clump 1, marked with red squares). It expands the observed duration of Godzilla from previously 2 years to 5 years. While Diego et al. (2022) included the original discovery ESO New Technology Telescope (NTT) observations in 2014 (Dahle et al. 2016), we exclude this as there Godzilla is not clearly resolved. Counting from the February 2018 HST observation, Godzilla has maintained its approximate luminosity for $\sim$1.5 years in the rest-frame.

Vanzella et al. (2020) argued that Godzilla (“Tr” in their terminology) is a Type IIn SN, a type of SN with narrow lines in their spectra. These narrow lines are believed to arise when a massive star explodes into a dense CSM. Although Type IIn SNe can be observed more than a few decades after explosion (Immler et al. 2005; Milisavljevic et al. 2012), it does not mean that it maintains high luminosity for that long time. One sub-type of Type IIn SNe, IIn-P (Mauerhan et al. 2013), is a type that shows a lasting, luminous plateau phase. A recent theoretical work however predicted the duration of this plateau phase to around 100 days at most (Khatami & Kasen 2023). Thus, a luminous plateau lasting for at least 1.5 years observed in Godzilla is highly unlikely to happen in any kind of SN known so far.

Godzilla is also unlikely to be some other kind of rare transient that shows an approximately flat light curve for 1.5 years. As summarized in Section 1, such a scenario contradicts the maximum time delay predicted from the lens model by Sharon et al. (2022), as a sudden increase in luminosity should have been observed in counter images in other arcs.

Godzilla and a pair of clumps next to it (the “P knots” following Diego et al. 2022) are believed to be nearby in the source plane (Diego et al. 2022). An extreme magnification can happen if Godzilla lies on the critical curve, as suggested in Diego et al. (2022). In Figure 10, we compared emission line profiles normalized by the flux of [O iii]$\lambda$5008 at each spectrum for the P knots, Godzilla and the Sunburst LCE. The left and right clumps in the P knots, P1 and P2, show nearly identical line profile traits such as redshift, line ratios and line width in various emission lines, and are clearly distinguished from Godzilla. It suggests that the P knots are likely to be the mirrored image of the same object different from Godzilla, and strengthens the hypothesis of a perturbing mass in the foreground generating a critical curve crossing the top of Godzilla and between the P knots (Diego et al. 2022, see Figure 6).

The plausibility of this lensing scenario is strengthened by the NIRCam detection of a low surface brightness galaxy centered only $0.5\mathrm{″}$ from Godzilla. As shown in Figure 11, this galaxy has a clumpy structure and may be responsible for creating the small-scale perturbations needed to shift the critical curve to the locations suggested by Diego et al. (2022).

Diego et al. (2022) suggested images 5.1 - 5.4 (marked with green circles in Figure 2) and 4.7 as candidate counterimages of Godzilla based on that they are located between images of clumps 1 and 2. On the other hand, the lens model from Sharon et al. (2022) predicts that Godzilla and the P knots together make up the highly resolved 8th counterimage of Clump 4 (depicted with purple circle labeled “4.8” in Figure 2).

To identify possible counterimages of Godzilla, which should have a similar SED as Godzilla, we examined the SEDs for Godzilla, the P knots, and the images of clumps 1, 4, and 5. Using Photutils (Bradley et al. 2024), we extracted flux from 0.06 arcsec radius circular apertures, using the images taken with 6 NIRCam filters (F115W, F150W, F200W, F277W, F356W, F444W). A color-color diagram was then created by combining the filters that displayed the most distinct differences in SED shape (F200W-F277W versus F277W-F356W), as shown in Figure 12. The stellar continuum properties of Godzilla appear to be closer to clump 4 than to clump 5. The P knots also occupy a position on the color-color diagram that is indistinguishable from clump 4.

The unusually bright O i $\lambda$8449 in Godzilla provides an additional method for investigating its counterimages. Figure 13 shows the O i $\lambda$8449/[O iii]$\lambda$5008 map for the north arc and northwest arc, revealing that strong O i $\lambda$8449 emission is a unique feature of only Godzilla, P knots and clump 4, aside from the bright areas corresponding to image 8.4 in the pointing 1. Of the clump 5 images, only image 5.4, located next to image 1.4, falls within the NIRSpec pointing area and does not show strong O i $\lambda$8449. To test whether this is due to suppressed [O iii]$\lambda$5008 caused by high density, we have normalized O i $\lambda$8449 with H$\beta$ (Figure 20). We observe an O I excess in the same regions as seen in the [O iii]$\lambda$5008 normalization map, and image 5.4 still shows no elevation. Thus, the elevation in O i $\lambda$8449 is not the result of suppressed [O iii]$\lambda$5008 in the high-density region. We believe that the [O iii]$\lambda$5008 emission observed in the Godzilla region originates from high-density gas with $n_{\rm e}\gtrsim 10^{6}\>\rm cm^{-3}$. As shown in Figure 8, [O iii]$\lambda$5008 probes regions with $n_{\rm e}\gtrsim 10^{6}\>\rm cm^{-3}$, regardless of temperature. While the O i/H$\beta$ ratio is also influenced by oxygen abundance, the O i $\lambda$8449/[O iii]$\lambda$5008 map, which is independent of metallicity, shows that the elevated O i $\lambda$8449 is not due to high oxygen abundance. O i $\lambda$8449 exhibits an elevation due to a unique Ly$\beta$-pumping mechanism, which helps constrain the counterimage of Godzilla.

Through photometric color comparison (Figure 12) and the O i $\lambda$8449 maps (Figure 13, Figure 20), we concluded that images of clump 4 and P knots are counterimages containing Godzilla, as they share a similar stellar continuum color and characteristic nebular emission.

The magnification factor of the candidate counterimages is known from the gravitational lens model; therefore, by comparing the flux of the candidate counterimages with that of Godzilla, we can determine Godzilla’s magnification factor. However, simply comparing the stellar continuum brightness is not sufficient. Sharon et al. (2022) considers Godzilla and the P knots as an enlarged version of clump 4, which makes direct flux comparison difficult if Godzilla is indeed a part of clump 4. Since O i $\lambda$8449 is a very unique line, if we assume that the only source of O i $\lambda$8449 observed in clump 4 is inside Godzilla and that there are no other major O i $\lambda$8449 sources, comparing the brightness of O i $\lambda$8449 between the clump 4 images and Godzilla would provide a more accurate comparison.

We extracted one-dimensional spectra from Godzilla, the two P knots (P1, P2), and images 4.4 and 4.9. We did not include image 4.10, as its close proximity to the bright image 1.10 might cause contamination. To measure the total O i $\lambda$8449 flux for Godzilla, we re-extracted its spectrum over a $7\times 5$ pixel area that includes the five regions (Center, NE, SE, NW, SW) shown in Figure 3. The continua have been removed following the method described in Section 3.1, then we summed the flux of the O i $\lambda$8449 line within the extraction area. We also compared the flux of the stellar continuum using photometry measurement employed for SED and color comparison in the previous section. The results are shown in Table 4.

Godzilla exhibits a large discrepancy between its continuum flux ratio and its O i $\lambda$8449 flux ratio. Compared to image 4.4, Godzilla’s continuum flux is 7.6 times brighter, while its O i $\lambda$8449 flux is much brighter than that of image 4.4 with the ratio of 35. Assuming that Godzilla is the sole O i $\lambda$8449 source, a flux ratio of 35 would be a more likely measure of the true magnification ratio between the two images. Under the assumption of a homogeneous magnification across the Godzilla region, the fact that the stellar continuum is only 7.6 times brighter suggests that the Godzilla region contains $\sim 22\%$ ($7.6/35\times 100$) of the stars compared to image 4.4. Meanwhile, P1 and P2 do not show significant differences between their continuum flux ratios and O i $\lambda$8449 flux ratios when they were compared to image 4.4. From this, we can infer that P1 and P2 represent the entirety of image 4.4, rather than just a part of it. The magnification factor for image 4.4 has derived to be $16.2^{+1.7}_{-1.8}$ from Pignataro et al. (2021) and $15.7^{+1.3}_{-2.7}$ from Sharon et al. (2022). Combining these magnification factors with the O i flux ratio, the magnification factors for Godzilla, P1, and P2 are inferred to be $\mu\approx 560$, $\mu\approx 20$, and $\mu\approx 25$, respectively.

Comparing to image 4.9 presents a slightly different picture. The difference between the stellar continuum and O i $\lambda$8449 flux in Godzilla is even more pronounced in this comparison, where the continuum flux ratio is $\sim 9.5$ and the O i $\lambda$8449 flux ratio soars to $\sim 94$. Interpreting this with the same logic as before suggests that Godzilla contains only $\sim 10$ % of the stars present in clump 4. In the meanwhile, P1 and P2 show discrepancy of only factor of 2. The magnification factors for image 4.9 vary significantly between the two models presented. Pignataro et al. (2021) reported the magnification factor for image 4.9 to be $71.1^{+7.6}_{-6.4}$ while Sharon et al. (2022) reported it as $31.1^{+8.9}_{-3.1}$, noting the magnifications of images 1.9, 1.10, 4.9, and 4.10 should not be taken at face value. The magnification factors for Godzilla are then inferred to be $\mu\approx 6700$ and $\mu\approx 2900$ based on the models from Pignataro et al. (2021) and Sharon et al. (2022). This is all summarized in Figure 14, which clearly displays the variations in Godzilla’s magnification factor and the percentage of stellar components based on different images and models.

Such widely varying values indicate that the current lens model has not reached a consensus. In image 4.4, predictions from the two models were similar and the critical line in that area was well constrained. However, it is still unclear if the magnification factor for image 4.4 is indeed more accurate than that for image 4.9. The stellar continuum flux in image 4.4 is $1.2\pm 0.09$ times brighter than that of 4.9, yet the lens models suggest magnification factor for image 4.4 to be $\sim 1/2-1/5$ smaller than that of 4.9, indicating an internal inconsistency. Regarding the differences between models, despite symmetry between images 9 and 10, a model by Sharon et al. (2022) does not place the critical line between them. This results in a lower magnification factor than Pignataro et al. (2021), which used a perturber (galaxy 1298 in that study) to bend the critical line. Recently, the updated lens model by Solhaug et al. (2024) proposed a new perturber location (Perturber I therein) that shifts the critical line to run between images 9 and 10. This work reports a magnification factor $\sim 270$ for image 4.9 (E. Solhaug, private communication), which leads to $\mu\approx 25000$ for Godzilla. This is an extreme number compared to $\mu\approx 600-7000$ derived by Diego et al. (2022) or $\mu\approx 190-5000$ by Pascale & Dai (2024) based on the same counterimage comparison methods. It should also be noted that results from Diego et al. (2022) are based on the assumption that clump 5 is the counterimage of Godzilla. While Pascale used clump 4 instead, both Diego et al. (2022) and Pascale & Dai (2024) lacked observational data on O i $\lambda$8449 and thus compared only the photometric brightness of Godzilla and its counterimage. They assumed that the total brightness of the counterimage corresponds to the total brightness of the Godzilla region and did not consider the possibility that Godzilla corresponds to only a part of the clump.

These two studies also aimed to estimate the maximum magnification of Godzilla. Diego et al. (2022) suggested that while magnifications up to $10^{5}$ due to small-scale perturbers such as stars are theoretically possible, there are two key limitations: (1) such extreme magnifications cannot be sustained over long periods due to the motion between the star and the caustic, and (2) microlenses within the cluster can reduce the maximum achievable magnification, especially when the magnification is very high and caustics are overlapping ($\tau_{\text{eff}}\gg 1$). Based on these factors, the highest sustained magnification was estimated to be a few times $10^{4}$. On the other hand, Pascale & Dai (2024), based on microlensing simulations and the fact that Godzilla shows flux variations of less than 3%, concluded that the magnification factor of Godzilla is unlikely to exceed 2000. However, their approach relies on the value $\log(\mu M_{\star}/M_{\odot})=9.3$. Adopting their magnification factor, it means that the stellar mass of the Godzilla should be comparable to that of the LCE cluster. Comparing the rest-frame optical images of clump 4 and the LCE cluster with similar magnification (e.g. image 1.4 and 4.4 where $\mu=15.3^{+7.7}_{-6}$ and $\mu=15.7^{+1.3}_{-2.7}$, respectively, according to Sharon et al. (2022)), their relative brightnesses are not consistent with being of similar mass, and it makes their maximum magnification estimation less convincing.

These contradictory results from different studies illustrate how the estimated magnification factor of Godzilla can vary significantly depending on the chosen lens model. Given the current lens models, we cannot rule out the scenario in which Godzilla is magnified by a very high factor of $\sim 10^{4}$, which represents the highest sustainable magnification.

Our main interest lies not in the components contributing to Godzilla’s ordinary stellar continuum, but in identifying the source responsible for the O i $\lambda$8449 emission and other unique characteristics of Godzilla. Spatially analyzing the surrounding region around the brightest Center may help understand this source. As seen in Table 3, the broadest component at Center, NE, and SE regions show a high value of $\mathrm{FWHM}\sim 435-547$ $\rm{km\,s^{-1}}$. It is larger than the expansion velocity of most local LBVs (a few to 100 $\rm km\,s^{-1}$), and similar to that of $\eta$ Car (Weis & Bomans 2020). While NW and SW regions do not have the very broad component in H$\alpha$, these regions show strong outflow features in many lines. In the G140H/F100L grating, NW and SW regions show broad components blue-shifted compared to the narrow component by $143\pm 8$ $\rm{km\,s^{-1}}$ and $153\pm 12$ $\rm{km\,s^{-1}}$ in the rest-frame, respectively.

One distinguishing feature of Godzilla is that it is very dusty. The Sunburst Arc is in general not so dusty (Mainali et al. 2022). The dust cover in this region seems extremely uneven; in the spatial dust map of the whole Sunburst Arc, Godzilla and counterimages of clump 4 stand out with a significantly higher $E(B-V)$ than typical values for the galaxy (Rivera-Thorsen et al. in prep.). As shown in Table 5, the $E(B-V)$ reaches $\sim 0.24-0.78$ , depending on the region. The NW and SW regions, where strong outflow features are observed (Figure 3), are less dusty. Interestingly, Mainali et al. (2022) find that their MagE spectrum containing Godzilla have a slightly lower $E(B-V)$ compared to other regions based on the H$\alpha$/H$\beta$ Balmer decrement (0.02 for the M-3 pointing containing Godzilla vs. 0.04–0.15 for the other pointings). The $E(B-V)\approx 0.02$ for the M-3 pointing, which includes Godzilla, is significantly lower than the value derived in this study. This discrepancy is likely due to aperture dilution. The Balmer emission line map shows strong emission along the core of the arc, with a significantly lower H$\alpha$/H$\beta$ value than in Godzilla (Rivera-Thorsen et al. in prep.). Given that the M-3 slit includes the central spine of the arc adjacent to Godzilla, we anticipate a substantial dilution effect, lowering the calculated E(B-V) value. Moreover, unlike the NIRSpec spectra, MagE spectra are subject to air blurring as they are ground-based, which would further enhance the aperture dilution effect.

Assuming $R_{V}=4.05$, we obtain $A_{V}=1.8$ for the Center, a value strikingly close to $A_{V}\approx 2.0$ derived for the Weigelt blobs (Davidson et al. 1995; Hamann et al. 1999). Despite significant dust reddening, strong UV emission lines are observed in the Weigelt blobs, as they are for Godzilla. This may suggest that the dust is mixed with the nebular gas, allowing emission from the outer layers of the gas to escape with minimal attenuation, rather than being absorbed by a dust screen external to the nebular gas.

Given the amount of dust Fe is typically expected to be depleted, yet we detect abundant Fe lines. This phenomenon is also observed in $\eta$ Car, and Smith & Ferland (2007) suggested that Fe-bearing grains can be selectively destroyed while other dust grains remain intact; a similar process might be occurring in this system.

The surrounding regions are also variant in their gas temperature and density. Looking at the temperature and density diagnostics diagram in the surrounding regions (Figure 19), we can see that the density is higher ($n_{\rm e}\gtrsim 10^{3}\>\rm cm^{-3}$) in the SE and SW regions compared to the NE and NW regions ($n_{\rm e}<10^{3}\>\rm cm^{-3}$). The cross section of the density diagnostics and low temperature diagnostic line ([O ii]$\lambda$$\lambda$7322,7332) is also more biased to the higher density in the SE and SW regions.

Spatially variant kinematics, dust properties, and gas temperature and density clearly show that the central bright source is surrounded by multi-phase, inhomogeneous gas. Depending on the magnification, if it is closer to 7000, this may imply a circumstellar medium and dust ejected through stellar winds or previous eruptions. If the magnification is closer to 600 and Godzilla displays a broader region, this could reveal spatial variation within a larger nebular region.

Astronomical objects that emit O i $\lambda$8449 through the Ly$\beta$-pumping mechanism are limited. We exclude reflection nebulae and H ii Regions since The O i $\lambda$8449 line observed there is typically produced by stellar continuum pumping or recombination, not by Ly$\beta$-pumping. For example, Grandi (1975) demonstrated that starlight continuum fluorescence is the preferred excitation mechanism for the O i line in the Orion Nebula. We also exclude supernova remnants (SNRs). Certain supernova remnants, especially those interacting with surrounding interstellar material, can exhibit O i $\lambda$8449 line emissions. However, O i $\lambda$8449 emission through Ly$\beta$-pumping is rare in SNRs because they typically lack the necessary combination of dense neutral oxygen and intense UV radiation near the remnant. For example, from the ratio of O i $\lambda$7776 and O i $\lambda$8449, Winkler & Kirshner (1985) and Itoh (1986) have shown that O i $\lambda$8449 observed in SNR Puppis A and Cassiopeia A mainly results from recombination. Very Massive Stars (VMS) and Wolf-Rayet (WR) stars are also easily excluded since they are not known for strong Bowen fluorescent lines. Moreover, we detect neither nebular nor broad stellar He ii and there is no sign of broad stellar C iv emission lines at 5808 Å  which effectively excludes a VMS and WR stars. We also exclude classical Be stars. Classical Be stars are B-type stars of luminosity classes V-III that exhibit prominent Balmer emission lines (Jaschek et al. 1981). Be stars are the rapidly rotating stars and we can observe emission lines from the disks or ring-like envelopes surrounding the low-latitude regions (Kogure & Leung 2007). Although Mathew et al. (2012b, a) have argued that Ly$\beta$-pumping plays an important role in the excitation of O i $\lambda$8449 lines observed in classical Be stars, Figure 4 of Mathew et al. (2012b) shows that there is still a significant contribution from collisional excitation, unlike in Godzilla. B[e] stars, a subgroup of peculiar Be stars that exhibit IR excess and forbidden lines (Allen & Swings 1976), are also excluded due to the lack of prominent Ly$\beta$-pumped O i $\lambda$8449 lines. For example, the spectrum of B[e] star HD 45677 shows O i $\lambda 7776/\lambda 8449\approx 0.3$ (de Winter & van den Ancker 1997, Fig. 4) which can be explained by collisional excitation.

Following is a list of objects that could emit strong O i $\lambda$8449 mainly through Ly$\beta$-pumping:

• •

  Broad-line regions (BLRs) of active galactic nuclei (AGNs) and quasars: High UV fluxes from the central black hole can ionize the surrounding gas and produce Bowen fluorescence lines. After analyzing the spectra of 16 Seyfert galaxies, Grandi (1980) concluded that 13 of them exhibited broad O i $\lambda$8449 emission, with Ly$\beta$-pumping identified as the excitation mechanism. Godzilla does not display line widths approaching those typically found in BLRs of several thousand km/s; and does not show high-excitation emission lines such as He ii, N v, or O vi which are typical in BLRs of AGNs. In Figure 15, we show the position of Center of Godzilla in the N2, S2, and O1 BPT diagrams; it clearly falls within the stellar region, without any evidence for high-energy excitation or strong shocks, ruling out the black hole scenario discussed in Diego et al. (2022).

• •

  Planetary nebulae: In dense regions close to the central star of planetary nebulae, Ly$\beta$ photons can excite neutral oxygen efficiently. Ly$\beta$-pumped O i $\lambda$8449 emission has been observed in several compact, high-density planetary nebulae, such as NGC 7027 (Rudy et al. 1992), IC 5117 (Rudy et al. 2001), and IC 4997 (Rudy et al. 1989; Feibelman et al. 1992). However, due to the hot central star, a highly ionized spectrum is generally observed. The estimated temperature for the central stars of NGC 7027, IC 5117 and IC 4997 is 219000 K (Zhang et al. 2005), 120000 K (Hyung et al. 2001) and 47000 - 59000 K (Feibelman et al. 1979), respectively. One can easily see that high-ionization lines such as He ii or [Ne v] detected in spectra of planetary nebulae listed above, are not observed in Godzilla.

• •

  Symbiotic stars: Symbiotic stars refer to binary systems consisting of a cool giant star transferring mass to an accompanied hot star, mostly white dwarfs. Ly$\beta$ photons from the hot star can excite neutral oxygen atoms in the surrounding gas and induce Bowen fluorescence. Ly$\beta$-pumped O i $\lambda$8449 emission has been observed in many symbiotic starts such as RR Telescopii (RR Tel) (Thackeray 1955; Damineli 2001), AG Pegasi (AG Peg) (Ciatti et al. 1974; Tomov et al. 2016), V1016 Cygni (V1016 Cyg) (Strafella 1981), HM Sagittae (HM Sge) (Ciatti et al. 1977), and BX Monocerotis (BX Mon) (Anupama et al. 2012). Shore & Wahlgren (2010) also thoroughly examined the O i $\lambda$1302 and O i] $\lambda$1641 observed in EG And, Z And, V1016 Cyg, and RR Tel, along with nova RS Oph 1985 in outburst, and concluded that the line strength variation is related to the light curve and outburst activity. Spectra of symbiotic stars listed above all show high-excitation lines including He ii originating from hot white dwarfs, which are not observed in Godzilla.

• •

  Nova ejecta: During a nova eruption, a white dwarf accretes material from its companion star until a thermonuclear runaway occurs. The ejected material is intensely ionized, and as it cools, it produces various emission lines including O i $\lambda$8449. Ly$\beta$-pumped O i $\lambda$8449 has been observed in several novae in the outburst phase, such as nova Cygni 1975 (Strittmatter et al. 1977) and nova V4643 Sgr (Ashok et al. 2006). In their outburst, novae typically exhibit high-excitation lines such as He ii and O vi which are not observed in Godzilla and show rapid spectral evolution on a timescale of days to months.

• •

  Herbig Ae/Be (HAeBe) Stars: These young, pre-main sequence stars, defined by Herbig (1960), are surrounded by accretion disks. UV radiation from these stars, as well as shocks within the accretion disks, can induce the Bowen fluorescence. Mathew et al. (2018) argued that Ly$\beta$-pumping is the primary mechanism responsible for O i $\lambda$8449 line observed in HAeBe stars. Although HAeBe stars show many spectral similarities to Godzilla, they have less in common compared to $\eta$ Car and the Weigelt blobs, which will be discussed next. Figure 2 of Mathew et al. (2018) shows that in HAeBe stars, the O I $\lambda 8449/\lambda 7776$ ratio can reach up to about $\sim 25$ – 50 in extreme cases, whereas in Godzilla, this ratio is $\sim 110$. While O I $\lambda$7776 is fairly prominent in many HAeBe stars, this line is not detected in Godzilla. Furthermore, due to the presence of a stellar disk, H$\alpha$ in HAeBe stars often shows a double-peak profile, and even when a single peak is present, it lacks the broad wings observed in Godzilla (Carmona et al. 2010, Fig. 3).

We find that the peculiar O i $\lambda$8449 source in Godzilla is best explained as an analog of the Luminous Blue Variable (LBV) star $\eta$ Car and the Weigelt blobs, given its spectral characteristics. In Section 4.4, we have shown that the strong, narrow permitted oxygen line O i $\lambda$8449 has most likely arisen from a Ly$\beta$ pumping mechanism. In Section 5.3.2, we have reviewed a number of known astronomical sources of O i $\lambda$8449 emission, and have shown that the combination of a strong O i $\lambda$8449 line and the absence of the O i $\lambda$7776 line is a rare feature, observed only in Weigelt blobs among objects that do not exhibit high-ionization lines. Moreover, as mentioned in Section 4.3, we know that dense gas of $n_{\rm e}\gtrsim 10^{6}\>\rm cm^{-3}$ exists in close vicinity of Godzilla (Vanzella et al. 2020), although our optical density diagnostics only offer limited constraints on low density gas. Existence of high density gas and the strong fluorescent line emission is evocative of the Weigelt blobs in the $\eta$ Car systems explained in Section 1.

The Weigelt blobs are dense gas condensations that are primarily neutral, with an ionized surface layer facing the stars (See Figure 1. in Johansson & Letokhov 2005, for a sketched out overview). The O i in the neutral condensations is shielded by hydrogen from ionisation, but exposed to Ly$\beta$ photons emitted from the H ii surface of the blob, as well as the stellar wind, populating the 3d³D level of the neutral oxygen atoms (Johansson & Letokhov 2005, Figure 4). This level has an excitation energy very close to the resonant energy of Ly$\beta$ photons and can be pumped by it, if Ly$\beta$ is sufficiently strong - a mechanism known as PAR. Shore & Wahlgren (2010, Figure 1) show a different view of this same cascade. Once the 3d³D level is populated, it decays to 3p³P level through a $\lambda=11286$ Å transition which is outside the wavelength range of these observations. Subsequently, they continue to cascade down from 3p³P to 3s³S₁ through the $\lambda=8449$ Å transition, and from there into O i]$\lambda$1641 and the $\lambda=$ 1302, 1304, 1306 triplet. We note that all of these lines are strongly detected in the MagE and NIRSpec observations of Godzilla. In the Weigelt blobs, as oxygen in the interior of the blob remains neutral, it results in a special situation where we observe strong Ly$\beta$-pumped O i $\lambda$8449, but not the recombination line O i $\lambda$7776. The clear detection of the full Ly$\beta$-pumped cascade, combined with the non-detection of O i $\lambda$7776 (Section 4.4, Figure 9) is the smoking gun that permitted O i lines in Godzilla has arisen from dense, Weigelt blob-like gas condensations.

In the Weigelt blobs, [Ne iii] is only observed at times when the blob is exposed to the secondary star, which is hotter than the LBV star. We can explain the significant dust attenuation, the strong Bowen fluorescent emission, and the [Ne iii] emission line simultaneously, if Godzilla is a binary system including a massive evolved LBV-like star and a hotter star, analogous to $\eta$ Car. It is a plausible scenario considering that 70% of massive stars are affected by binary interaction, and 50% have companions (Weis & Bomans 2020). It is also noteworthy that Smith & Tombleson (2015) and Smith (2019) suggested that LBVs result from close binary evolution. Alternatively, the properties listed above can be explained without invoking a binary scenario, if the gas condensations in Godzilla are exposed to more intense radiation of an LBV star compared to the Weigelt blobs. Assuming the LBV star has a similar temperature to $\eta$ Car, this would be possible if the blobs were located more closely to the star than the Weigelt blobs to $\eta$ Car. These two scenarios could potentially be distinguished by observing periodic variability in line emission. However, depending on the inclination and viewing angle, spectral variability might not be detected, even if it is indeed a binary system.

We would also like to discuss the magnification and brightness issues raised in previous studies. Diego et al. (2022) has noted that with $\mu\approx 7000$, the true brightness of Godzilla would match that of $\eta$ Car during its Great Eruption. Despite the high resolution of JWST/NIRSpec, we do not detect any signs of an eruption such as a very broad H$\alpha$ ($\sim 10^{4}$ $\rm{km\,s^{-1}}$) or P Cygni profiles. If we adopt the values suggested by Diego et al. (2022), this implies that the magnification factor for Godzilla must exceed 7000 if the source has the same magnitude as $\eta$ Car in its non-eruption phase. However, it should be noted that the analysis by Diego et al. (2022) is based on the assumption that Godzilla is a single stellar source. If Godzilla is indeed composed of multiple stars, the constraints on the magnification factor could be relaxed.

The overall picture we propose is as follows: Godzilla is part of clump 4, comprising 10 - 25% of its stellar light, depending on the lens models and the images in comparison. It could be a small group of just a few stars, or tens or even hundreds of stars. These stars contribute significant or even dominating fraction of the stellar continuum. Within this group of stars, a peculiar O i $\lambda$8449 source is present, which is best explained as a source similar to the Weigelt blobs in the $\eta$ Car system. The spectra of Godzilla resemble those of $\eta$ Car in its quiescent phase, with nearby dense gas condensations exposed to a hotter source, either due to a presumable hotter companion or their potentially closer location to the star. We believe that the emission lines observed in Godzilla are primarily from this dense gas condensation, though some lines, such as redshifted [O ii] lines, may originate from other regions (Figure 8). Our scenario is a kind of hybrid model, differing from both Diego et al. (2022)’s interpretation of Godzilla as a single object and Pascale & Dai (2024)’s interpretation as a star cluster, in the sense that the stellar continuum is affected by multiple stars, while a single object analogous to $\eta$ Car dominates the emission line. In the next section, we further examine the similarities and differences between O i $\lambda$8449 in Godzilla and the Weigelt blobs.

Johansson & Letokhov (2005) discuss the possibility of astrophysical laser effects of O i $\lambda$8449 in the Weigelt blobs. As the decay of the lower 3s³S₁ level to the ground level is faster than the decay of the upper 3p³P level, population inversion occurs. O i $\lambda$8449 from Godzilla is due to Ly$\beta$ pumping, so we expect it to show an inverted population, but it does not necessarily mean that it is a laser since we cannot guarantee stimulated emission without information on the size of the blob and amplification of the O i $\lambda$8449 line. For observational confirmation, Johansson & Letokhov (2005) suggests looking for sub-Doppler width, i.e. the line width narrower than the width caused by Doppler broadening, of the 8449 line.

Ly$\alpha$-pumped iron lines can provide us an additional clue about the sub-Doppler width of the O i $\lambda$8449 line. Along with the Ly$\beta$-pumped O i $\lambda$8449, we detect Ly$\alpha$-pumped Fe ii $\lambda$8490 with S/N of 8 (see Table 2). Interestingly, another Ly$\alpha$-pumped iron line, Fe ii $\lambda$8453, which arise from the same mechanism thus should appear is not detected in the fits using [O iii]$\lambda$$\lambda$4960,5008 as kinematic template (Table 2). This line is blended with O i $\lambda$8449, and we suspect that its flux may have been included in the O i $\lambda$8449 flux since the O i $\lambda$8449 line is narrower than other lines. We fitted the fluorescent lines again, using only one component and relaxing the width constraint. The result is shown in Table 2 with asterisk. We measured FWHM of 106$\pm$5 $\rm{km\,s^{-1}}$ for the fluorescent lines and it increased S/N of Fe ii $\lambda$8453 from $\sim 1.8$ to $\sim 12$. However, along with the lack of information on the thermal broadening in this object, the spectral resolution of NIRSpec does not allow sufficiently strong constraints on the line profile to confirm or falsify if the O i $\lambda$8449 emission observed in Godzilla is an astronomical laser.

The relative strengths of Ly$\alpha$-pumped Fe ii lines compared to the Ly$\beta$-pumped O i $\lambda$8449 line is significantly smaller than observed in the Weigelt blobs (see Davidson & Humphreys 2012, Figure 5.4). In the Weigelt blobs, Fe ii fluorescent lines are much stronger than O i $\lambda$8449. This might imply an abundance difference, which is not surprising considering that Godzilla is at $z=2.37$ and the Weigelt blobs are in our Galaxy, where the metal enrichment has been enhanced by Type Ia SN. Several observations have already suggested significant alpha enhancement, i.e. iron deficiency, in $z\sim 2-3$ galaxies (Steidel et al. 2016; Topping et al. 2020a, b; Cullen et al. 2021). Alternatively, it might be a circumstantial evidence of a O i laser and significant amplification in that line.