Using one-dimensional hydrodynamic simulations including interstellar heating , cooling , and thermal conduction , we investigate nonlinear evolution of gas flow across galactic spiral arms .
We model the gas as a non-self-gravitating , unmagnetized fluid , and follow its interaction with a stellar spiral potential in a local frame comoving with the stellar pattern .
Initially uniform gas with density n _ { 0 } in the range 0.5 { cm } ^ { -3 } \leq n _ { 0 } \leq 10 { cm } ^ { -3 } rapidly separates into warm and cold phases as a result of thermal instability ( TI ) , and also forms a quasi-steady shock that prompts phase transitions .
After saturation , the flow follows a recurring cycle : warm and cold phases in the interarm region are shocked and immediately cool to become a denser cold medium in the arm ; post-shock expansion reduces the mean density to the unstable regime in the transition zone and TI subsequently mediates evolution back into warm and cold interarm phases .
For our standard model with n _ { 0 } = 2 { cm } ^ { -3 } , the gas resides in the dense arm , thermally-unstable transition zone , and interarm region for 14 % , 22 % , 64 % of the arm-to-arm crossing time .
These regions occupy 1 % , 16 % , and 83 % of the arm-to-arm distance , respectively .
Gas at intermediate temperatures ( i.e .
neither warm stable nor cold states ) represents \sim 25 -30 % of the total mass , similar to the fractions estimated from H i observations ( larger interarm distances could reduce this mass fraction , whereas other physical processes associated with star formation could increase it ) .
Despite transient features and multiphase structure , the time-averaged shock profiles can be matched to that of a diffusive isothermal medium with temperature 1 , 000 { K } ( which is \ll T _ { warm } ) and “ particle ” mean free path of l _ { 0 } = 100 { pc } .
Finally , we quantify numerical conductivity associated with translational motion of phase-separated gas on the grid , and show that convergence of numerical results requires the numerical conductivity to be comparable to or smaller than the physical conductivity .