2 Observations and data reduction

We observed I Zw 36 in the [C II] $158\ \mu \rm m$ line with the Long-Wavelength Spectrometer (LWS; Clegg et al. 1996) on board the Infrared Space Observatory (ISO; Kessler et al. 1996). The observations are two concatenated 3-point raster scans: the Target Dedicated Time (TDT) numbers 19401278 and 19401279. Both the rasters are centered on the optical center of the galaxy, listed in Table 1 with other observational parameters. The other 4 observed positions have offsets of $\pm 3\hbox{$^\prime$ }$ in the north-south (TDT 19401278) and east-west (TDT 19401279) directions, relative to the center position. These off-position observations enable us to evaluate any possible contamination by the foreground Galactic emission. We adopted the Astronomical Observation Template (AOT) LWS02 mode: line observations with a medium-resolution ( $\Delta \lambda = 0.6\ \mu \rm m$) grating spectroscopy. The observations at the galaxy position consisted of 40 grating scans, with 16 seconds of total integration time for each grating position. Twenty grating scans (8 s of total integration) were carried out at each of the off-positions. The grating positions were spaced at 1/4 of the spectral resolution for both the on- and off-position observations. The beam size of the LWS derived from observations of Mars was 68'' in FWHM at the wavelength of the [C II] line (Gry et al. 2000).

We subtracted the detector dark currents from the Standard Processed Data (SPD) of the Off-Line Processing (OLP) version 7 products, using the LWS Interactive Analysis (LIA) version 7.3. The ISO Spectral Analysis Package (ISAP) version 1.6a was used for the data reduction afterwards. The data affected by cosmic-ray hits were manually removed, when they had not been discarded automatically in the SPD. The individual spectra were averaged at each of the observed positions, and then the spectra at the 4 off-positions were averaged resulting in a single off-source spectrum. No difference was seen among the spectra at the off-positions. Finally, we derived the foreground-subtracted spectrum by subtracting the off-source spectrum from the on-source spectrum. We tried to fit the on-source, off-source, and foreground-subtracted spectra with single-Gaussian line profiles and linear baselines. The widths of the Gaussian profiles were assumed to be equal to the instrumental resolution because the ¹²CO J = 1-0 line width of the galaxy was small ( $\Delta v_{\rm CO} = 32\ \rm km\ s^{-1}$; Young et al. 1995) relative to the resolution.

For faint sources, the uncertainty in the LWS flux calibration predominantly results from difficulties in the subtraction of the detector dark currents. These difficulties prevented us from deriving a correct continuum flux density from the LWS02 spectra of I Zw 36. However, line flux is less affected by this problem, especially for brighter emission such as the [C II], which generally have a flux density comparable to that of the continuum on a galactic scale at this wavelength and spectral resolution (Crawford et al. 1985). Thus, for the [C II] line, we adopt the nominal LWS calibration uncertainty of the OLP version 7 products for faint sources, 50%.

Figure 1: Observed flux density (solid curve) of I Zw 36 as a function of wavelength. A linear baseline was subtracted. The vertical bars indicate the uncertainties ($1 \sigma$) based on the scatters among the individual spectral scans. The dashed curve indicates the [C II] line profile expected when the Galactic [C II]/CO J = 1-0 intensity ratio (Nakagawa et al. 1998; Dame et al. 1987) is applied to CO observations (Young et al. 1995) of I Zw 36. The heliocentric radial velocity of the CO emission is $v_{\rm hel} = 288\ \rm km\ s^{-1}$at the line center. The width of the expected line profile is determined predominantly by the instrumental resolution

Figure 2: Observed [C II]/12CO J = 1-0line intensity ratio of galaxies (filled circles) vs. interstellar oxygen abundance, $\rm 12 + \log[ O / H ]$. The ratio for I Zw 36 indicates the upper limit (Table 1). The horizontal bars indicate the observation uncertainties for I Zw 36 and NGC 1569. For the inner Galaxy, the LMC, and IC 10, the horizontal and vertical bars represent the spatial variations within the galaxies ($1 \sigma$ except the horizontal bar for our Galaxy; see text). The dashed line indicates the upper limit for starburst and normal spirals (Stacey et al. 1991); the main-beam temperature scale is adopted for the CO line in this figure instead of the $T_{\rm R}^\ast$ scale