Dense gas in nearby galaxies XIII. CO submillimeter line emission from thestarburst galaxy M 82

(CO)-C-12 J = 1-0, 2-1, 4-3, 7-6, and (CO)-C-13 1-0, 2-1, and 3-2 line emis sion was mapped with angular resolutions of 13 " - 22 " toward the nuclear region of the archetypical starburst galaxy M 82. There are two hotspots on either side of the dynamical center, with the south-western lobe being sli ghtly more prominent. Lobe spacings are not identical for all transitions: For the submillimeter CO lines, the spacing is similar to 15 ", for the mil limeter lines (CO J = 2-1 and 1-0) the spacing is similar to 26 ", indicati ng the presence of a 'low' and a 'high' CO excitation component. A Large Velocity Gradient (LVG) excitation analysis of the submillimeter li nes leads to inconsistencies, since area and volume filling factors are alm ost the same, resulting in cloud sizes along the lines-of-sight that match the entire size of the M 82 starburst region. Nevertheless, LVG column dens ities agree with estimates derived from the dust emission in the far infrar ed and at submillimeter wavelengths. 22 " beam averaged total column densit ies are N(CO) similar to 510(18) and N(H-2) similar to 10(23) cm(-2); the t otal molecular mass is a few 10(8) M.. Accounting for high UV fluxes and variations in kinetic temperature and ass uming that the observed emission arises from photon dominated regions (PDRs ) resolves the problems related to an LVG treatment of the radiative transf er. Spatial densities are as in the LVG case (n(H-2) similar to 10(3.7) cm( -3) and similar to 10(3)cm(-3) for the high and low excitation component, r espectively), but (CO)-C-12/(CO)-C-13 intensity ratios greater than or simi lar to 10 indicate that the bulk of the CO emission arises in UV-illuminate d diffuse cloud fragments of small column density (N(H-2) similar to 5 10(2 0) cm(-2)/km s(-1)) and sub-parsec cloud sizes with area filling factors mu ch greater than 1. Thus CO arises from quite a different gas component than the classical high density tracers (e.g. CS, HCN) that trace star formatio n rates more accurately. The dominance of such a diffuse molecular interclu mp medium also explains observed high [C I]/CO line intensity ratios. PDR m odels do not allow a determination of the relative abundances of (CO)-C-12 to (CO)-C-13. Ignoring magnetic fields, the CO emitting gas appears to be c lose to the density limit for tidal disruption. Neither changes in the C-12 /C-13 abundance ratio nor variations of the incident far-UV flux provide go od fits to the data for simulations of larger clouds. A warm diffuse ISM not only dominates the CO emission in the starburst regi on of M 82 but is also ubiquitous in the central region of our Galaxy, wher e tidal stress, cloud-cloud collisions, shocks, high gas pressure, and high stellar densities may all contribute to the formation of a highly fragment ed molecular debris. (CO)-C-12, (CO)-C-12/(CO)-C-13, and [C I]/CO line inte nsity ratios in NGC 253 (and NGC 4945) suggest that the CO emission from th e centers of these galaxies arises in a physical environment that is simila r to that in M 82. Starburst galaxies at large distances (z similar to 2 2- 4 7) show (CO)-C-12 line intensity ratios that are consistent with those ob served in M 82. PDR models should be applicable to all these sources. (CO)- C-12/(CO)-C-13 line intensity ratios much greater than 10, sometimes observ ed in nearby ultraluminous mergers, require the presence of a particularly diffuse, extended molecular medium. Here [C I]/CO abundance ratios should b e as large or even larger than in M 82 and NGC 253.