We explore what dominant physical mechanism sets the kinetic energy contained in neutral , atomic ( H I ) gas .
Supernova ( SN ) explosions and magneto-rotational instability ( MRI ) have both been proposed to drive turbulence in gas disks and we compare the H I line widths predicted from turbulence driven by these mechanisms to direct observations in 11 disk galaxies .
We use high-quality maps of the H I mass surface density and line width , obtained by the THINGS survey .
We show that all sample galaxies exhibit a systematic radial decline in the H I line width , which appears to be a generic property of H I disks and also implies a radial decline in kinetic energy density of H I .
At a galactocentric radius of r _ { 25 } – often comparable to the extent of significant star-formation – there is a characteristic value of the HI velocity dispersion of 10 \pm 2 km s ^ { -1 } .
Inside this radius , galaxies show H I line widths well above the thermal value ( corresponding to \sim 8 km s ^ { -1 } ) expected from a warm H I component , implying that turbulence drivers must be responsible for maintaining this line width .
Therefore , we compare maps of H I kinetic energy to maps of the star formation rate ( SFR ) – a proxy for the SN rate – and to predictions for energy generated by MRI .
We find a positive correlation between kinetic energy of H I and SFR ; this correlation also holds at fixed \Sigma _ { HI } , as expected if SNe were driving turbulence .
For a given turbulence dissipation timescale we can estimate the energy input required to maintain the observed kinetic energy .
The SN rate implied by the observed recent SFR is sufficient to maintain the observed velocity dispersion , if the SN feedback efficiency is at least \epsilon _ { SN } \simeq 0.1 \times ( 10 ^ { 7 } { yr } / \tau _ { D } ) , assuming \tau _ { D } \simeq 10 ^ { 7 } yr for the turbulence dissipation timescale .
Beyond r _ { 25 } , this efficiency would have to increase to unrealistic values , \epsilon \gtrsim 1 , suggesting that mechanical energy input from young stellar populations does not supply most kinetic energy in outer disks .
On the other hand , both thermal broadening and turbulence driven by MRI can plausibly produce the velocity dispersions and kinetic energies that we observe in this regime ( \gtrsim r _ { 25 } ) .