A study of extreme CIII]1908 & [OIII]88/[CII]157 emission in Pox 186: implications for JWST+ALMA (FUV+FIR) studies of distant galaxies

Carbon spectral features are ubiquitous in the ultraviolet (UV) and far-infrared (FIR) spectra of galaxies in the epoch of reionization (EoR). We probe the ionized carbon content of a blue compact dwarf galaxy Pox 186 using the UV, optical, mid-infrared and FIR data taken with telescopes in space (Hubble, Spitzer, Herschel) and on the ground (Gemini). This local (z ∼ 0.0040705) galaxy is likely an analogue of EoR galaxies, as revealed by its extreme FIR emission line ratio, [O iii]88 µ m/[C ii] 157 µ m ( > 10). The UV spectra reveal extreme C iii] 𝜆𝜆 1907, 1909 emission with the strongest equivalent width (EW) = 35.85 ± 0.73 Å detected so far in the local (z ∼ 0) Universe, a relatively strong C iv 𝜆𝜆 1548, 1550 emission with EW = 7.95 ± 0.45Å, but no He ii 𝜆 1640 detection. Several scenarios are explored to explain the high EW of carbon lines, including high eﬀective temperature, high carbon-to-oxygen ratio, slope and upper mass of top-heavy initial mass function, hard ionizing radiation and in-homogeneous dust distribution. Both C iii] and C iv line proﬁles are broadened with respect to the O iii] 𝜆 1660 emission line. Each emission line of C iv 𝜆𝜆 1548, 1550 shows the most distinct double-peak structure ever detected which we model via two scenarios, ﬁrstly a double-peaked proﬁle that might emerge from resonant scattering and secondly a single nebular emission line along with a weaker interstellar absorption. The study demonstrates that galaxies with extreme FIR emission line ratio may also show extreme UV properties, hence paving a promising avenue of using FIR+UV in the local (via HST+Herschel/SOFIA) and distant (via JWST+ALMA) Universe for unveiling the mysteries of the EoR.


INTRODUCTION
Understanding the reionization of the Universe is one of the frontier goals of modern astronomy.Several theoretical and observational efforts have been made to answer the related pressing questions such as when and how first galaxies formed (e.g., Stark 2016) and whether these first galaxies reionized the intergalactic medium (IGM; e.g., Robertson et al. 2010).A first step to answer these questions is to search for the early galaxies in the epoch of reionization (EoR) and characterize their properties.
Deep imaging campaigns have been quite successful in searching for such sources.For example, the Hubble Space Telescope (HST) allowed us to compile large samples of early galaxies via deep nearinfrared (NIR) imaging programs such as the Great Observatories Origins Deep Survey (GOODS, Giavalisco et al. 2004), Extreme Deep Field (XDF, Illingworth et al. 2013) and Cosmic Assemble Near-Infrared Deep Extragalactic Legacy Survey (CANDELS, Gro-★ E-mail: kumari@stsci.edugin et al. 2011;Koekemoer et al. 2011).However, spectroscopy is essential to characterize the properties of these sources.Until the launch of the James Webb Space Telescope (JWST), the rest-frame ultraviolet (UV) spectroscopy was available for only a few EoR galaxies (e.g., Sobral et al. 2015;Stark et al. 2017;Topping et al. 2021;Hutchison et al. 2019).The JWST observations are significantly improving the dearth of UV spectroscopy for the EoR galaxies via the planned follow-up spectroscopic surveys of the Hubble deep fields (Robertson 2021;Curtis-Lake et al. 2022;Bunker et al. 2023).Similarly, the exquisite sensitivity of the Atacama Large Millimeter Array (ALMA) has made it possible to obtain the far-infrared (FIR) spectroscopy of EoR galaxies (e.g., Maiolino et al. 2015;Carniani et al. 2017;Smit et al. 2018;Bouwens et al. 2022;Witstok et al. 2022).The combined JWST+ALMA spectroscopic observations will significantly enhance our understanding of the EoR.
An indirect approach to probe the nature of reionization sources while efficiently using the two simultaneously-operating state-of-theart facilities, JWST and ALMA, is to perform detailed studies of the physical processes operating in local galaxies which might resemble EoR galaxies.Several different criteria have been devised so far to identify the local analogues of high-redshift galaxies, including gasphase metallicity, star-formation rate, compactness, stellar mass, UV luminosity, dust attenuation, Ly emission, colour, and ionization state among many others.Some established classes of local analogues of high-redshift galaxies are blue compact dwarf galaxies (BCD, Searle & Sargent 1972), green peas (GP, Cardamone et al. 2009) and blueberries (Yang et al. 2017), though it is not clear whether these galaxy populations also resemble the EoR galaxies, mainly because of the dearth of data available on EoR galaxies so far.
One of the goals of this paper is to demonstrate the use of FIR line ratio [O ] 88 µm/[C ] 157 µm for identifying the local analogues of the EoR galaxies.[O ]88 µm originate from the ionized gas, [C ] 157 µm may originate from both the ionized as well as neutral interstellar medium (ISM), and their relative strengths (i.e. [O ] 88 µm/[C ] 157 µm line ratio) can potentially tell us about the porosity of ISM (Chevance et al. 2016;Polles et al. 2019).The ALMA observations of EoR galaxies (Figure 1, blue points) reveal that [O ] 88 µm/[C ] 157 µm line ratio may vary in the range 1-10, indicating a highly porous ISM which will facilitate the leakage of ionizing photons required for reionization of the neutral IGM.The Herschel Dwarf Galaxies Survey (Madden et al. 2013;Cormier et al. 2015) reveal a large population of local dwarf galaxies with [O ] 88 µm/[C ] 157 µm > 1 (Figure 1, grey points), which are potentially the local analogues of EoR galaxies.
In an attempt to explore and establish using [O ] 88 µm/[C ] 157 µm as a criterion for identifying the EoR local analogues, we obtained HST UV and spatially-resolved optical spectroscopy of Pox 186, a unique dwarf galaxy showing the highest [O ] 88 µm/[C ] 157 µm ever detected in the local Universe (Figure 1, red point).Moreover, Ramambason et al. (2022) shows that this galaxy has an ionizing photon escape fraction of ∼ 40%, thus making it ideal for this study.Pox 186 was originally discovered in Kunth et al. (1981), and was thought to be a protogalaxy.Corbin & Vacca (2002) later shows that Pox186 is an ultra-compact galaxy still in the process of formation with a majority of star-formation concentrated in the central star cluster of mass 10 5 M .Figure 2 shows a narrow-band optical image of Pox 186 taken with Wide Field Planetary Camera2 onboard HST, along with the field-of-view (FOV) of instruments of primary observations used in this work.Table 1 lists some of the main physical properties of Pox 186, along with information about the UV and optical observing strategy.A typical UV spectrum of star-forming galaxies is known to show prominent spectral features such as Ly 1215, C  1548, 1550, He  1640, O ] 1660, 1666, [C ] 1907 and C ] 19091 , which have been used to infer information regarding hardness of radiation fields, ionization conditions, metal-content, wind properties within galaxies at all redshifts (e.g., Shapley et al. 2003;Senchyna et al. 2017;Nakajima et al. 2018;Schmidt et al. 2021).In this paper, we mainly focus on the ionized carbon spectral features, C  1548, 1550 and C ] 1907,1909, however, we complement the UV analysis with the spatially-resolved optical, mid-infrared (MIR) and FIR data.
The paper is organized as follows: Section 2 presents an overview of the data used in this work, including UV, optical, MIR and FIR.For UV and optical data, we explain the initial data reduction and processing.The MIR and FIR data are archival.In Section 3, we present the results of the multi-wavelength data analysis which includes the estimates of redshift, distance, flux and equivalent widths of detected emission lines and reddening.We also determine several physical properties of the ionized gas and the ionizing stellar population, such as electron temperature and density, gas-phase metallicity, ionization parameters, effective temperature and softness parameters.Section 4 presents a discussion focusing on UV carbon features including their large equivalent widths, line profiles and relative chemical abundance.We also discuss the implication of this study on future JWST+ALMA studies of reionization-era galaxies.Section 5 summarizes our main results.

HST/UV spectroscopy
The HST/COS observations were taken as part of the General Observing programs in HST Cycles 27 and 28 (GO: 16071 and 16445, PI: N Kumari) at lifetime adjustment position 4 (LP=4).Before taking each UV spectrum, the NUV target acquisition image is taken using the Mirror A. The FUV and NUV spectra were taken with the 2.5 arcsec diameter Primary Science Aperture (PSA) using the medium resolution gratings, G130M, G160M and G185M centred at 1291Å, 1623Å and 1913Å, respectively.We used all FP-POS positions for better spectral sampling and increased signal-to-noise (S/N).Table 1 lists the exposure times for gratings and Mirror A used within the two HST programs.All HST/COS data were processed with the standard data reduction pipeline CALCOS version 3.4.0.
Figure A1 shows the HST/COS NUV target acquisition image where the red circle denotes the 2.5 arcsec COS spectroscopic aperture.The wavelength settings allow us to cover several spectral features consisting of ISM (red), photospheric (purple), wind (yellow) and nebular (brown) lines as shown in Figures 3a and 3b.

GMOS-N optical spectroscopy
We obtained the spatially-resolved optical spectroscopy of Pox 186 using the GMOS (Hook et al. 2004) and IFU (GMOS-N IFU; Allington-Smith et al. 2002) at Gemini-North telescope in Hawaii, as part of two separate programs (PID: GN-2020A-FT-105, GN-2021A-FT-111, PI: N Kumari).The first program focussed on covering the optical wavelength range of ∼3500-8000Å, while the second program allowed us to cover the near-infrared (NIR) wavelength range of ∼8000-10000Å.The observations were taken in one-slit queue mode providing FOV of 3.5 ×5 , large enough to cover the entire galaxy (Figure 2).Along with the science exposures, standard calibration observations were obtained including GCAL flats, CuAr lamp for wavelength calibration and standard star HZ44 for flux calibration.We performed the basic steps of data reduction using the standard GEMINI reduction pipeline (version v1.15)   calibration, sky subtraction, and differential atmospheric correction finally producing the 3D data cubes, and have been described in detail in Kumari (2018).New GMOS-N detectors were installed in 2017, which further required the quantum efficiency correction for all flats.We also used the L. A.Cosmic (van Dokkum 2001) to remove the cosmic rays from the science exposures.We chose a spatial sampling of 0.25 for the final three-dimensional data cubes.We thus obtain three three-dimensional data cubes.We scale the flux of each of the three cubes using the methodology described in Appendix B.

Ancillary Mid-Infrared and Far-Infrared spectroscopy
MIR: Pox 186 was observed with the Spitzer telescope in the lowresolution (R ≈ 60-127) mode using long-slits of the InfraRed Spectrograph in the (IRS; Houck et al. 2004).The width of long slit is ∼ 3.6 , and its point-spread function (PSF) may extend beyond the

Source Redshift and Distance
We determine the source redshift by measuring the observed wavelength of four strong UV emission lines O ] and C ] as shown in Table 3.We do not use the strong C emission line because it is double-peaked (Section 4.2).We do not use the optical emission lines because the zero-point of the wavelength calibration of GMOS-N data is not very well-constrained (Kumari 2018).The redshift of

UV
Table 4 presents the fluxes and equivalent widths (EW) of the UV emission lines O ], C ] and C .We estimate fluxes by summing the fluxes in the spectral line after subtracting a local linear continuum fitted to either side of the three doublets.The continuum level at the central wavelength of each emission line is used to estimate their equivalent widths.We correct the UV emission line fluxes using the E(B-V) value determined from optical Balmer decrement which is described later in Section 3.3.

Optical
We want to compare the UV, optical, MIR and FIR properties of Pox186 together for which we need to take into account the varying aperture sizes or FOVs of the different instruments with which these four different datasets are acquired.The GMOS-IFU data allow us to probe the optical properties for different apertures and sizes.We chose to extract two sets of integrated spectra.The first one is obtained by integrating all GMOS spectra within a circular aperture of 1.25 radius centred on the brightest knot of Pox18, hence coinciding with the HST/COS aperture, and is referred to as 'COS-matched integrated' spectra.While the second set of spectra is obtained by integrating all spectra within the GMOS-IFU FOV and is referred to as 'Gemini-FOV integrated' spectra.The main difference between the two sets of spectra is that the COS-matched integrated spectra only include the compact core of Pox 186, and exclude its plume, unlike the Gemini-FOV integrated spectra which include both.Pox 186 : GMOS-N IFU Near-infrared (HST/COS aperture)  are the random measurement uncertainties.However, we also include a systematic flux uncertainty of 50% (see Appendix B) and propagate in the inferred properties whenever it becomes relevant in the analysis, and is explicitly mentioned in the paper.

Pox 186 7
Table 5. Optical emission line flux measurements (relative to H = 100) obtained from the two sets of integrated Gemini/IFU spectra: (i) the one overlapping with the COS aperture (ii) the other from the entire Gemini FOV.Observed line fluxes (  ) are extinction-corrected using E(B-V) to calculate the intrinsic line fluxes (  ).Notes: F(H ) in units of × 10 −15 erg cm −2 s −1 ....: S/N < 3 for a given line.-: For these lines, the covered Gemini FOV is different than for the rest of the emission lines, hence we do not report their values.

Reddening correction
We estimate the colour excess E(B-V), by using the attenuation curve of the Small Magellanic Cloud (SMC; Gordon et al. 2003) along with the observed Balmer decrement (H /H ) assuming a Case B recombination and an electron temperature and density of 10 4 K and 100 cm −3 , respectively.We estimate E(B-V) for both sets of integrated optical spectra, the one overlapping with COS and the other one corresponding to the entire GMOS FOV.The E(B-V) for the COS-matched integrated spectra is 0.053 ± 0.004 which is close to the E(B-V) of the Milky Way in the line-of-sight of Pox 186, i.e., 0.0385 ± 0.0016 (Schlafly & Finkbeiner 2011), hence indicating a very low amount of dust in the central region of Pox 186.
The intrinsic fluxes for the UV and optical lines are estimated by correcting the observed line fluxes using the E(B-V).No reddening correction is done for the MIR or FIR emission line fluxes.Table 4 presents the intrinsic fluxes of the emission lines of the UV COS spectra, while Table 5 shows the intrinsic fluxes for the COS-matched and Gemini-FOV integrated spectra.The uncertainties on the intrinsic fluxes are derived from propagating the random uncertainties on fluxes measured while fitting the emission lines.

Electron temperature and Density
The UV, optical and IR spectra have emission lines sensitive to the electron temperature (T  ) and density (N   T  .Similarly, the optical to IR line ratio, [O ] 5007/[O ] 88 µm is sensitive to T  , but also to N  (Dinerstein et al. 1985) .
Figure 5  which is in agreement with that obtained from optical emission line ratios.The electron temperatures derived from the COS-matched and Gemini-FOV integrated spectra agree with each other and are typical of the H regions within the star-forming galaxies (Kumari et al. 2017(Kumari et al. , 2018(Kumari et al. , 2019)).
We measure electron density using the density-sensitive line ratio [S ]  6717, 6731 line ratio and T  ([O ]) determined above for the COS-matched as well as the Gemini-FOV integrated spectra.For both datasets, the electron density indicates a low-density regime.We note that the density-sensitive line doublets [O ]  3727, 3729 could not be used for determining density, as the sensitivity of GMOS-IFU in the blue end has degraded over time, and hence the blue-end data are unusable.We do not estimate N  from the density-sensitive UV doublet C ] 1907, 1909 available from the COS spectra as the doublet is blended and asymmetric.

Chemical abundances
The chemical abundances of Pox 186 are only determined for the COS-overlapping central region, because of the non-detection of the necessary emission lines in the Gemini-FOV integrated spectra.
Gas  2+ .We also estimate C/O using the empirical method given in (Pérez-Montero 2017).The estimates of C/O from the direct and empirical method are in excellent agreement with each other (Table 6).

Radiation hardness
Hardness of the radiation field can be measured by the softness parameter log .It was initially defined in terms of optical ionic ratios  = (O + /O 2+ )/(S + /S 2+ ) (Vilchez & Pagel 1988), and can be estimated from the calibration provided in Kumari et al. (2021), i.e., log  = log  + 0.16/t + 0.22, where t = T  /10 4 and log  can be determined from the optical emission line ratios ( log  = We estimate log  = -0.47 ± 0.23 using MIR emission line fluxes (Table 2).We estimate log  = -0.14 ± 0.23 using T  from the Gemini-FOV integrated spectra and log  estimated earlier.We chose to use T  from the Gemini-FOV integrated spectra rather than COSmatched integrated spectra as the former covers the entire galaxy like MIR data.We do not use the optical emission line fluxes to determine log  as [O ]  3729, 3729 are not detected because of the decreased sensitivity of GMOS-N IFU in the blue wavelength end.Lower log  and log  indicate a hard radiation field.Pox 186 exhibits lower values of log  and log  compared to the average values exhibited by star-forming regions or galaxies in the local Universe (Kumari et al. 2021), thus indicating that the radiation field in Pox 186 is harder than average.
Table 6 summarizes all the physical properties derived in this section.The uncertainties on the derived quantities are estimated from the random uncertainties on the flux measured while fitting emission lines and excluding the systematic flux uncertainty.
We explore the cause for the extreme EW of carbon lines in the following:  Nakajima et al. (2018) states that EW(C ]) >30Å can be caused by blackbody with extremely high effective temperature (T eff ), i.e. > 6×10 4 K.We estimate T eff = 60 ± 18 kK assuming blackbody using the H -code (Section 3.4.3),thus indicating that the high effective temperature may be responsible for high carbon EW observed for Pox 186. ) where line flux is corrected using nebular E(B-V) while continuum is corrected using the stellar E(B-V).Similarly, the vertical dashed and dotted lines indicate the observed and reddening-corrected quantities, respectively.The reddening-corrected EW is obtained by using nebular E(B-V) for line fluxes and stellar E(B-V) for continuum, while the reddening-corrected emission line ratios are obtained by using nebular E(B-V) for both emission lines in the ratio.The right-ward pointing arrows in the middle panel indicates the lower limit on the C ]/He where a 2-upper limit on He line is considered.Jiang et al. (2021) suggests that the high EW(C ]) and EW(C ) could be due to a higher carbon abundance.

High carbon-to-oxygen ratio
We explore this via Figure 7, which shows that the C/O abundance of Pox 186 (red point) is higher than the average for galaxies found in the same metallicity range at z ∼ 0 (horizontal blue line) and z∼ 2 (horizontal green line) from Arellano-Córdova et al. (2022).It is also higher than that predicted by the best-fit line to C/O versus O/H for measurements of stars derived by Nicholls et al. (2017, solid purple curve) and for measurements of irregular dwarf galaxies derived from Garnett et al. (1995, dashed solid line).Thus, the higher C/O for Pox 186 supports the argument from Jiang et al. (2021) about higher carbon abundance causing the higher EW of carbon lines.
We note that the optical spectrum of Pox186 does not show any signature of the carbon-rich Wolf-Rayet (WR) stars either as red or blue WR bump which could lead to a direct enhancement of carbon.Schaerer et al. (1999) lists Pox 186 as a WR galaxy on the basis of a broad He  4686 at 0.8 above background reported in Kunth & Joubert (1985).The high-quality HST/COS and GMOS-IFU data allow us to exclude Pox 186 as a WR candidate.

Slope and upper mass of top-heavy initial mass function
To understand the origin of the extreme EW(C ]) measured in Pox186, we consider C models broadly similar to those in Witstok et al. (2022); here, we give a brief summary and highlight the differences with the models presented in Witstok et al. (2022).The incident radiation field of a single burst of star formation with varying ages (1 Myr to 100 Myr) is generated by v2.1 stellar population synthesis models including binary stars under a top-heavy initial mass function (IMF; with slope −2; Eldridge et al. 2017), ranging in stellar mass from 1 M to 300 M .Unlike in Witstok et al. (2022), calculations stop when a molecular fraction of 10 −6 is reached, such that the model does not extend into a photodissociation region beyond the central H region.
We considered models with a wide range of base (stellar) metallicities, tuning the gas-phase metallicity to match the observed values of Pox186.We introduce an additional nebular α/Fe enhancement, which is accomplished by increasing the nebular abundances of individual α elements (for details, we refer to Witstok et al. 2022).The nebular elemental abundances of the main α elements (C, O, Ne, Mg, Si, S) are scaled up by 4×, except for carbon which is increased by a factor of 2, so that the C/O ratio is approximately half the solar value as appropriate for Pox186 (see Section 4.1.2).Moreover, this implies our fiducial model with stellar metallicity  * = 0.025 Z has a nebular oxygen abundance of approximately 10% solar, as directly measured for Pox186 (see Section 4.1.2).We vary the ionisation parameter and hydrogen density between −4 < log 10  < −1 and 10 −1 cm −3 <  H < 10 4 cm −3 (as measured at the illuminated face of the cloud), respectively.
An overview of the models generated in C is shown in Figure 8.For simplicity, we only show models with a fixed density of  H = 10 2 cm −3 and stellar metallicities of 0.001 Z , 0.025 Z , and 0.2 Z .We find particularly the slope and upper mass of the IMF are restrictive in reproducing the extreme EWs of C ], as models with an upper mass of 100 M only reach EWs of approximately 20 Å 3 .However, none of these models could simultaneously reproduce the  Kewley et al. (2001, theoretical) and Kauffmann et al. (2003, empirical), respectively, distinguishing between the photoionized and non-photoionized regions.The blue dot denotes the corresponding emission line ratios obtained from the optical spectra integrated over the COS aperture, while the grey dots indicate the spatially-resolved emission line ratios.
observed EW of C ] and H emission lines, denoted by dashed horizontal and vertical lines, respectively, in Figure 8).
To explore this further, we considered the dust distribution which might affect the nebular C ] and the underlying stellar continuum differently.We estimate the UV continuum slope  = -0.36±0.02using the spectral windows from Calzetti et al. (1994) in the wavelength region of ∼1250-1850, which indicates a red continuum.If we use this  value to deredden the continuum using the SMC law from Reddy et al. (2018) and use E(B-V) derived from the optical data to deredden the emission line again using the SMC law, the EW(C ]) may decrease by a factor of 3. The reddening-corrected values are shown by dotted vertical and horizontal lines, which can not be reproduced by the models either.
The inability of the models to reproduce the observed or reddeningcorrected properties of Pox 186 could be due to the simplistic assumptions on the geometry and relative distribution of dust and gas within the photoionization models.In summary, it indicates the need to improve the existing population synthesis and photoionization models.

Hard Ionizing Radiation
High EW(C ]) is also proposed to be caused by the hard ionizing radiation from extreme stellar populations or AGN (Nakajima et al. 2018;Jiang et al. 2021) or shocks from the radio jets (Best et al. 2000).We rule out the possibility of an AGN/shocks causing the high EW(C ]) as figure 9 shows that the optical emission line ratios obtained from the integrated spectrum, [O ]/H versus [N ] /H (left-hand panel) and [O ]/H versus [S ] /H (righthand panel) do not occupy the AGN/shock region of the classical emission-line diagnostic diagrams (Baldwin et al. 1981;Veilleux & Osterbrock 1987).Figure 9 also shows a few spaxel-based line ratios lying beyond the photoionization region on the BPT diagrams; however, they are too few (∼2.5% and ∼3% for [N ] -BPT and [S ] -BPT diagrams, respectively) to be a conclusive indicator of hard AGN radiation.Hard ionizing radiation is also expected to produce He lines, so emission line ratio diagnostic diagrams including optical and UV He lines are also used to determine the presence of AGN (Feltre et al. 2016;Brinchmann et al. 2008).However, nei-ther He  1640 nor He  4686 lines are detected in the spectrum corresponding to the COS pointing.The above discussion shows that hard ionizing radiation from AGN is unlikely to be the cause of high EW(C ]).

Line profiles of carbon lines
Figure 10 (left-hand panel) shows that C line profile is broadened with respect to O ] line profile, which appears to be caused by collisional excitation.Berg et al. (2019) observed similar behaviour in a couple of local metal-poor galaxies and attributed this to the resonant scattering nature of C .However, for Pox 186, we find that C ] line is also broadened with respect to O ] (Figure 10, right-hand panel).Given that the C ] is not reported to come from resonant scattering, we rule out resonant scattering as the cause of broadening in carbon lines.
It is unlikely that outflows could cause broadening in carbon emission lines because outflows would lead to broadening in all emission lines in the same way, which would result in similar line profiles (i.e., including O ]).Still, we explore the outflows signatures in Figure 11, where we overplot the [O ] emission line (normalized by its peak flux) along with Si  1260 (normalized by the median flux within the local spectral region).Both O ] and Si  1260 lines are centered at ∼ 0 km s −1 , showing no signatures of outflow.However, a hint of ionized gas suffering turbulence is present in the line profile of H shown in Figure 12, which shows the presence of a broad underlying component of H along with a narrow component.
Figure 13 shows the velocity profile of the resonantly scattered C  1548,1550 doublet.The distinct blue and red peaks exhibited by both C emission lines are quite interesting, as no previous studies have ever resolved the two peaks in both emission lines of the C doublet, though the stronger C 1548 has been reported to have double peaks (Berg et al. 2019).The origin of double-peaked C is difficult to understand because C line profile will be impacted by the relative fraction of the gas emitting the narrow nebular emission and the foreground ISM resulting in absorption.We address two possible scenarios here: (1) a pure nebular emission  with no ISM absorption and (2) nebular emission along with ISM absorption.
For modelling a purely nebular C (Figure 13, left panel), we fit two Gaussian components to each C emission line consisting of a blue-shifted (dashed-blue fit) and a redshifted component (dashedred fit).The peak separation between the blue and the red peak ( V sep ) is ∼132±2 km s −1 for each C line, which is ∼25 km s −1 (on average) higher than those found by Berg et al. (2019) for two local dwarf irregular galaxies.We note that the redshifted component is broader than the blue-shifted component.
For modelling the second scenario comprising of nebular emission and interstellar absorption (Figure 13, right panel), we model the emission in each C line with a Gaussian profile (dashed-blue fit) and interstellar absorption via the Voigt profile (dashed-red fit).The Voigt profile corresponding to the ISM absorption is narrower compared to the broad Gaussian profile, pointing towards a lower fraction of foreground high ionization gas along the line of sight.

Stellar Winds
Stellar winds originate from hot and massive stars, i.e., with masses > ∼40 M and temperatures > 25,000 K, and lead to P Cygni-type nebular emission via single Gaussian (dashed red curve) and the interstellar absorption via Voigt profile (dashed blue curve).On both panels, the overall best fit is given by the solid green line.
UV line profiles, typically for the strongest lines, N  1240, Si  1400 and C  1550.The UV spectra of Pox 186 show a strong P-Cygni N  1400 feature (Figure 3a, 1st row), no Si  1400 and a weak P-Cygni C  1550 feature (Figure 3a, 3rd row).The weaker C  1550 compared to N  1240 at lower metallicities, is indicative of a lower wind density and velocity (Leitherer et al. 2001(Leitherer et al. , 2010)).It is likely that the early O main-sequence stars are the main constituent of Pox 186, since these stars do not display wind effects in Si  1400, but in N  1240 and C  1550 as we find in Pox 186 spectra.
Figure 14 shows the models (v2.1, IMF slope = -2.35 and upper stellar mass-limit = 300 M ) overlaid on C P-Cygni profile (normalized by the continuum at  ∼1560-1570Å) both for the single stars (left-hand panel) and binary stars (right-hand panel) pop-ulation for stellar ages varying from 1-2.5 Myr and metallicity of Z = 0.05Z .It appears that the stellar populations as young as 1.6 Myr are sufficient to reproduce the weak C P-Cygni profile.The inclusion of binary stars has no effect on the overall C profile, as the binary stellar population becomes more important only at later ages (3-5 Myr) (Eldridge et al. 2020).
The blue-ward absorption in the N P-Cygni is blended with the Ly absorption from Pox 186 which is itself blended with the Ly absorption from the MW.Before comparing the N P-Cygni with the stellar population syntheses models such as , it is necessary to model and remove the Ly absorption from Pox 186 and the MW, which requires careful modelling of the stellar continuum.We will present the detailed modelling of these components in a follow-up paper.

Apparent absence of Ly𝛼
The absence of resonantly-scattered Ly emission in spite of strong C ] and C emissions is worth-noting (Figure 3a, upper-panel).A positive correlation is suggested/expected between EW(C ]) and EW(Ly) by a few studies (e.g., Stark et al. 2014), though Rigby et al. (2015) suggest that a positive correlation exists for strong emitters with EW(C ])>5Å and EW(Ly) > 50Å, with correlation getting weaker for weaker emitters.Given that Pox 186 shows the highest EW(C ]) detected in the local Universe so far, we expect at least some Ly emission.Similarly, Berg et al. (2019) propose that the double-peaked structure of resonantly-scattered C emission could be associated with a double peak in Ly as well.Moreover, Pox186 shows log U = -2.4± 0.4 by or log (q/cm s −1 ) = 8.1 ± 0.4, which lies in the range of z∼2-3 Ly emitters (Nakajima & Ouchi 2014) further suggesting that Pox 186 could be a Ly emitter.To investigate this further, we inspected the UV spectrum of Pox 186 taken with Space Telescope Imaging Spectrograph (STIS/HST) dataset (PI: Corbin, PID: 8333) which indeed shows Ly emission (Figure E1).We note that the STIS and our COS pointings are offset by ∼ 2 arcsec (∼ 168 pc), which indicates that the region emitting Ly is not entirely overlapping with that emitting ionized carbon, and lies at the outskirts of Pox 186 probably because of Ly escaping due to the concentrated feedback from the star-formation (Heckman et al. 2011).Only a deep spatial map of Ly can help identify any potential Ly emission within Pox 186.Moreover, a statistically significant sample of galaxies such as Pox 186 is required to establish any spatial offset between Ly emission and carbon emission.Such spatial offsets (∼168 pc) between Ly emission and C ] or C emission may not be probeable/distinguishable within the reionization era galaxies, simply because of the angular resolution of the existing instruments.So, even if originating from different regions of galaxies, UV carbon emission lines, particularly the stronger C ] line, might still be a good indicator of Ly emission emerging from the ISM of galaxies even when Ly is unavailable at redshifts > 6 caused by a significantly large IGM neutral fraction (e.g., Fan et al. 2006).

Lyman Continuum escape fraction
Figure 15 shows C /C ] versus EW(C ]) for Pox 186 (red point), along with measurements for galaxies at different redshifts compiled in Schmidt et al. (2021).Schaerer et al. (2022) finds that a C /C ]> 0.75 (shaded orange region) is characteristic of strong Lyman Continuum (LyC) leakers, i.e. galaxies with LyC escape fraction f esc (LyC)> 0.1.Pox 186 exhibits C /C ]= 0.30±0.04,which is much smaller  Figure 15 also shows a couple of reionization era galaxies which have C /C ] below the threshold proposed by Schaerer et al. (2022), and close to Pox 186.This raises two questions: (1) Is C /C ] or  slope a good enough predictor for LyC leakers?(2) Were there EoR galaxies which did not contribute to the reionization of the Universe?Such questions can be addressed by making the simultaneous use of UV and FIR observations of large samples of EoR galaxies as well as their local analogues.It would be crucial to do follow-up UV+optical observations via JWST of EoR galaxies, which show extreme IR properties via ALMA, and vice versa.

SUMMARY
We investigate the ionized carbon within a local dwarf galaxy Pox 186 using the HST/COS UV data complemented by the GMOS-IFU optical data and Herschel FIR data.Our main results are summarized as follows: (i) Using the HST/COS UV emission lines, we measure a redshift Pox 186 15 z = 0.0040705 ± 0.000013 for Pox 186.This corresponds to a luminosity distance of 17.5 Mpc assuming a flat ΛCDM cosmology or 12.6 Mpc assuming Cosmicflows-3.
(iii) We explore several scenarios to explore the high EW of carbon lines, including a high effective temperature, higher than average carbon-to-oxygen ratio for a given gas-phase metallicity, photoionization models including binary stars, top-heavy IMF and nebular α/ enhancement and in-homogeneous dust-distribution.The photoionization models could not simultaneously reproduce all the observables irrespective of dust-reddening, which could be due to the simplistic assumptions of the model parameters.
(iv) The C ] and C lines also show broadening with respect to the O ] emission lines though the cause of this broadening remains unknown.We rule out outflows causing broad carbon emission lines as no outflow signatures are found in the velocity profiles of O ] and Si lines.
(v) Optical integrated spectrum coinciding with COS aperture, shows a broad and faint underlying component in H along with a narrow component.Ruling out outflows on the basis of UV data, the H velocity profile indicates a turbulent ISM.
(vi) The C doublet shows clearly distinct double peaks for each of the two emission lines, which can be explained via two scenarios, such as pure emission with no absorbing foreground ISM or nebular emission along with a little absorbing ISM in the foreground.
(vii) The high EW(C ]) and log U = -2.4± 0.4 suggests a high EW(Ly) for Pox 186; however, COS spectra do not show any signature of Ly though a spatially-offset STIS spectrum does show Ly emission.
(viii) We report an observed UV continuum slope  = -0.36±0.04 which corresponds to f esc (LyC)∼10 −4 , indicating that Pox 186 is not a LyC leaker.C /C ] is also below the threshold for LyC leakers suggested by Schaerer et al. (2022).This is in contrast with the extreme [O ]/[C ] FIR line ratio, 40% escape fraction (Ramambason et al. 2022)  This work shows that the extreme IR [O ]/[C ] emission line ratios could correspond to extreme UV properties such as high EW of carbon lines (C ] and C ), high carbon-to-oxygen ratio, broadened emission carbon line profiles and double-peak within the resonant carbon line doublet, C .However, the apparent absence of Ly emission and negligible LyC escape fraction (as estimated from UV slope and C /C ]ratio) within a dwarf galaxy with such extreme UV and IR properties are puzzling.This requires a similar investigation on a larger sample of similar galaxies with UV+FIR data.The combination of HST and Herschel data for the local Universe, and JWST and ALMA for the reionization era Universe are crucial in carrying out such studies and understanding the similarities and differences between the EoR galaxies and their local analogues.

APPENDIX A: TARGET ACQUSITION IMAGE
Figure A1 shows one of the two NUV target acquisition images taken with taking the spectroscopic COS observations.

APPENDIX B: OPTICAL FLUX SCALING AND SYSTEMATIC UNCERTAINTIES
Note that the flux calibration obtained from the spectrophotometric standard star is relative and not absolute.To determine the optical emission line fluxes accurately, we follow the procedure described below.We first compare the ground-based Gemini/optical spectroscopic fluxes with the space-based archival HST imaging taken with the F658N filter.To do so, we extract the flux within an aperture of 1.25 radius (matching the COS aperture) on the WFC3/F658N image, subtract a background and apply the aperture correction.This value is then multiplied with the inverse sensitivity and the FWHM of the filter (estimated from synthetic photometry package ) to get the total flux within the aperture in the cgs units.We note that the width of an HST/WFC3 filter can be defined in several ways; however, we chose to use FWHM following the methodology of Laigle et al. (2016) for generating large photometric catalogues.To get the WFC3/F658N equivalent from Gemini optical data, we multiply the COS-matched Gemini spectrum with the transmission curve of the WFC3/F658N filter available from the SVO Filter Profile Service (Rodrigo et al. 2012;Rodrigo & Solano 2020).We integrate the resultant spectrum to get the HST equivalent of Gemini flux.This gives us a scaling factor of ∼2.7 for the Gemini spectra taken with R831 grating at ∼ 6900Å.
Note that we have three sets of Gemini spectra taken B600 grating at ∼4650Å and R831 grating at 6900Å and 8900Å.The Gemini spectra taken with B600/4650Åand R831/8900Å could in principle be scaled with respect to the scaled Gemini spectra taken with R831/6900Å, however, the continuum is too faint to do such an exercise.So, we estimate the E(B-V) from the unscaled H/H ratio from the B600/4650Å, and then use it with the scaled Ha line flux to estimate the scaling factor for H and hence the corresponding spectra.A similar procedure based on E(B-V) is followed where we predict the H I P10 line flux obtained from the R831/8900Å spectra, thus determining the scaling factor of these spectra.
These scaling factors give us H flux value in reasonable agreement with that derived by Guseva et al. (2004) for different-sized slits.
To determine the systematic uncertainties on flux calibration, we compare the integrated Gemini/IFU spectra of the standard photometric star (HZ44) with that available from CALSPEC, which gives us a systematic uncertainty of 50%.

Figure 2 .
Figure 2. HST/WFPC2 image of Pox 186 taken in F555W filter (Prog id: 8333) on which we overlay FOVs of all primary observations used in this work.The red circle denotes the HST/COS aperture of 1.25 radius (PIDs: 16445, 16071).The yellow and cyan rectangles of size 3.5 ×5 denote the FOVs of GMOS-N IFU programs GN-2020A-FT-105 and GN-2021A-FT-111, respectively.

Figure 3a .
Figure 3a.HST/COS FUV spectra of Pox 186 taken with G130M/1291 (upper two panels) and G160M/1623 (lower two panels).These fluxes are smoothed by a boxcar filter of 6 pixels for better visualization.These spectra show several spectral features consisting of ISM (red), photospheric (purple), wind (yellow) and nebular (brown) lines.The milky way lines are marked in green.Note that not all marked lines are detected, and are provided here for reference.The geocoronal emission is marked by a grey shaded region in the first panel.

Figure 3b .
Figure 3b.HST/COS FUV spectra of Pox 186 taken with G185M/1913.These fluxes are smoothed by a boxcar filter of 2 pixels for better visualization.Note that the shown spectrum only represents the middle segment of the G185M/1913 grating, and the other two segments do not show any spectral feature.
Figure 4 shows the COS-matched integrated spectra along with several optical and NIR emission lines.We measure the emission line fluxes for the recombination and collisionally excited emission lines (except H and [O ] 5007) within the integrated spectra by fitting single Gaussian profiles after subtracting a linear continuum in the spectral region of interest via custom-written python codes using package (Newville et al. 2014).Equal weight is given to flux in each spectral pixel while fitting Gaussians and the fitting uncertainties on the Gaussian parameters are propagated to calculate the flux uncertainty.We also create emission line flux maps for all lines (including H and [O ] 5007) of the

Figure 4 .
Figure 4. GMOS-N IFU integrated spectra of Pox 186 obtained by summing the spatially-resolved spectra within a circular aperture of radius 1.25 overlapping with the COS aperture covering a rest-frame wavelength range of ∼ 3800-9900Å.The important optical emission lines are marked in blue.We have also marked the location of He II 4686, which remains undetected in the optical spectrum.

Figure 5 .
Figure 5. grids of [O ] 4363/[O ] 5007 versus [O ] 5007/[O ]88 µm for T  =8-20kK and N  =1-10000 cm −3 .The red point denotes the emission line flux ratios of Pox 186, where the optical [O ] emission line fluxes correspond to the entire Gemini-FOV, and the FIR [O ] line is taken from Cormier et al. (2015), hence both optical and FIR datasets shown here cover the entire galaxy.The uncertainties on the line ratio include the systematic uncertainties on the optical line fluxes (see Section B).
shows the optical [O ] lines flux ratio ([O ] 4363 / [O ] 5007) versus the optical-IR [O ] lines flux ratio [O ] 5007/[O ] 88 µm, where the grids are generated using the emissivities of the respective [O ] lines from the code for a set of T  and N  values.The grid lines are not orthogonal as [O ] 5007/[O ] 88 µm is sensitive to both T  and N  .We also show the optical [O ] lines flux ratio and optical-IR [O ] lines flux ratio [O ]of Pox 186, obtained from Gemini-FOV integrated spectra.We deem the use of Gemini-FOV integrated spectra for this comparison instead of the COS-matched integrated spectra because of the large FOV and PSF of the Herschel data (see Section 2.3) which includes not only the compact core of Pox 186 but also its plume.Note that the COS-matched integrated spectra miss the plume of Pox 186.The observed lines ratio lies on the grids and is in contradiction to that found by Chen et al. (2023).The optical [O ] emission line flux ratio for the Gemini-FOV integrated spectra corresponds to T  = 15000 ± 1300K.From the COS-matched optical spectra, we estimate T  ([O ]) = 16300±500K by using the emission line flux ratio of the auroral line [O ] 4363 and [O ] 4959, 5007.Since UV line O ]  1660, 1666 is also temperature-sensitive, we also estimate T  ([O ]) by using the emission line flux ratio of [O ] 5007/O ] 1660,1666, -phase metallicity: The gas-phase metallicity (12+log(O/H)) can be robustly estimated from the T  -base direct method, where the abundances of the dominant ionic states of oxygen (O + and O 2+ ) are first determined from the temperatures of their respective ionization zones and then combined to estimate the total oxygen abundance.The temperature of the high-ionization zone T  ([O ]) is determined as derived in Section 3.4.1 and is combined with N  ([S ] ) to estimate the temperature of the low-ionization zone by using the densitydependent calibration given in Pérez-Montero (2017).Like Kumari et al. (2019) where [O ]  3727, 3729 remain undetected, we measure O + /H + using [O ]  7320, 7330 and the low-ionization temperature T  ([O ] ) by employing the formula given in Kniazev et al. (2003).We measure O 2+ /H + using [O ] 4959, 5007 and T  ([O ]) in the formula given in Pérez-Montero (2017).The oxygen ionic abundances are combined to calculate the oxygen elemental abundance, 12 + log(O/H) = 7.87±0.04,for the region of Pox 186 probed by COS, and agrees within 3 with that derived by Guseva et al. (2004) for a larger region of this galaxy.Carbon-to-oxygen ratio: We follow the relations between T  ([O ]) and dereddened line ratios C ]/O ] and C /O ] provided in Pérez-Montero (2017), we estimate C 2+ /O 2+ and C 3+ /O 2+ .We estimate direct method C/O by combining C 2+ /O 2 + and C 3 +/O 2 +, assuming   =  2+ + 3+

For
estimating the ionization parameter (log U) and effective temperature (T eff ) of the central region matching COS-aperture, we use the publicly available code H -T (v5.3 Pérez-Montero et al. 2019) where we assume a blackbody model and use reddening-corrected optical emission line fluxes ([O ] 4959, 5007, [S ]  6717, 6731, He I 6678, Ar  7135, [S ]  9069, 9532) from the COS-matched integrated and the gas-phase metallicity (12 + log(O/H) = 7.87±0.04,Section 3.4.2),and run the code for plane-parallel and spherical geometry separately.We find that the choice of geometry has no effect on either log U or T eff for the central region of Pox186.We also estimate log U = -2.661± 0.014 using the calibration involving [S ]  9069,9532/[S ] 6717,6731 given byKewley et al. (2019), which agrees with that derived from HC -.For determining log U for the entire Pox186 galaxy, we use calibrations fromKewley et al. (2019) including the MIR line ratios (Table2), [Ne ] /[Ne ] which gives log U = -2.55 ± 0.22.Note that we do not use optical line ratio [S ] /[S ] mainly because [S ]  9069, 9530 lines cover a slightly different FOV than the rest of the optical emission lines including [S ] .It is for the same reason that we could not use HC -to estimate these two parameters for the Gemini FOV.

Figure 8 .
Figure 8. EW(C ]) plotted against EW(H ) (left-hand panel), C ]/He II 1640 line ratio (middle panel) and C /C ] (right-hand panel) using the models described in Section 4.1.3.The horizontal dashed line indicates the observed EW (C ]) while the horizontal dotted line indicates the reddening-corrected EW(C ]) where line flux is corrected using nebular E(B-V) while continuum is corrected using the stellar E(B-V).Similarly, the vertical dashed and dotted lines indicate the observed and reddening-corrected quantities, respectively.The reddening-corrected EW is obtained by using nebular E(B-V) for line fluxes and stellar E(B-V) for continuum, while the reddening-corrected emission line ratios are obtained by using nebular E(B-V) for both emission lines in the ratio.The right-ward pointing arrows in the middle panel indicates the lower limit on the C ]/He where a 2-upper limit on He line is considered.

Figure 9 .
Figure 9. Classical optical emission line ratio diagrams: [O ]/H versus [N ] /H (left-hand panel) and [O ]/H versus [S ] /H (right-hand panel).Black solid curve and dashed curve represent the maximum starburst line fromKewley et al. (2001, theoretical)  andKauffmann et al. (2003, empirical), respectively, distinguishing between the photoionized and non-photoionized regions.The blue dot denotes the corresponding emission line ratios obtained from the optical spectra integrated over the COS aperture, while the grey dots indicate the spatially-resolved emission line ratios.

Figure 10 .
Figure 10.Comparison of O ] 1666 line with respect to C ] 1907 (left-hand panel) and C  1548 (right-hand panel).In both panels, all emission lines (C  1548, O ] 1666 and C ] 1907) are normalized by their peak fluxes.

Figure 11 .
Figure 11.Comparison of velocity profiles of O ] 1666 emission line (blue curve) and Si  1260 absorption line (purple curve), where O ] 1666 is normalized by its peak flux, and Si  1260 is normalized by the median flux in the velocity range of -250 to 250 km s −1 .No signature of outflow is present.

Figure 12 .
Figure 12.A narrow (purple Gaussian) and a broad (red Gaussian) component is needed to reproduce the H line profile, hinting towards the ISM turbulence within Pox 186.

Figure 13 .
Figure 13.We model the resonantly scattered double-peaked C  1548,1550 doublet according to two possible scenarios: (1) Purely emission without any absorbing foreground interstellar medium in the line of sight (left-hand panel), for which we model the C doublet via multicomponent Gaussian fits to identify the peaks of the blue-shifted (dashed blue curve) and the red-shifted (dashed red) components.(2) Nebular emission along with interstellar absorption (right-hand panel), where we model the Cnebular emission via single Gaussian (dashed red curve) and the interstellar absorption via Voigt profile (dashed blue curve).On both panels, the overall best fit is given by the solid green line.

Figure 14 .
Figure 14.Comparison of observed C P-Cygni feature (grey spectrum) within Pox 186 normalized by the continuum level between 1560-1570 Å with the models at Z=0.05Z comprising single (left-hand panel) and binary stars (right-hand panel) at different ages, i.e., 1 Myr (brown), 1.3 Myr (magenta), 1.6 Myr (blue), 2 Myr (olive) and 2.5 Myr (yellow).The location of C emission and absorption from Pox 186 are marked in red.The C absorption originating from MW is marked in green.

Figure 15 .
Figure 15.C /C ] versus EW(C ]) for Pox 186 (red point) and for published data at different redshifts presented in Schmidt et al. (2021).The orange band represents the C /C ]> 0.75, which is suggested by Schaerer et al. (2022) as strong continuum leakers.Error bars are simply not shown for clarity.
and the high [O ]/[O ] values from the literature.This raises questions on the potential use of  or C /C ] as tracers of LyC leakers.

Figure A1 .
Figure A1.HST/COS NUV target acquisition image of Pox 186.The red circle denotes the 2.5 arcsec COS spectroscopic aperture and is centered at (RA, Dec): (13 25 48.641, -11 36 37.94).The compass on the bottom-right of the figure shows North and East on the HST image.At the luminosity distance of this galaxy (i.e.17.5Mpc), 1 arcsec corresponds to 84 pc.

Figures
Figures C1 and C3 show the emission line flux maps obtained by fitting a single Gaussian to each optical line.The orientation of the FOV is different in the two figures because each set of Figures here is obtained from different observing programs with different position angle (see Figure 2).A compass is shown in each figure to show the orientation.Figures C2 show the maps of optical emission line ratios used in the classical emission line ratio diagrams (or the so-called BPT diagrams).

Figure D2 .
Figure D2.models showing the variation of the equivalent width of the C ]emission lines for single stars (upper panel) and binary stars (lower panel), and for a metallicity of 0.05Z .Five values of log U are considered i.e., -1.5 (purple), -2.0 (yellow), -2.5 (blue), -3.0 (green), -3.5 (magenta).Similarly, we consider four values of hydrogen densities, log n  , i.e., 0 (solid curve), 1 (dotted curve), 2 (dashdot curve), 3 (dashed curve).The observed EW of C ] and the associated uncertainties are represented by the solid horizontal black line and the shaded olive green region.The models sufficient to produce high C are insufficient to reproduce the high C ].

Figure E1 .
Figure E1.A noisy yet prominent Ly emission amidst the damped Lyabsorption is detected within the STIS/FUV spectrum of Pox 186.The blue cross shows the location of geocoronal emission.

Table 3 .
Emission line redshift determinations

Table 4 .
Equivalent widths and emission line fluxes (observed   and intrinsic   ) of the UV nebular emission lines.Notes: Fluxes are in units of × 10 −14 erg s −1 cm −2 . 2 upper-limit Pox 186 is 0.0040705±0.000013from the presented HST/COS data and agrees with that of Guseva et al. (2004) and Eggen et al. (2021) within uncertainties.At H  = 70 km s −1 and Ω  = 0.3, the observed redshift corresponds to a luminosity distance of 17.5 Mpc, and an angular scale of 84.1 pc per arcsec.However, we derive a value of 12.6 Mpc if we use Cosmicflows-3 (Kourkchi et al. 2020).