We analyze the transit timing variations ( TTV ) obtained by the Kepler mission for 22 sub-jovian planet pairs ( 19 published , 3 new ) that lie close to mean motion resonances .
We find that the TTV phases for most of these pairs lie close to zero , consistent with an eccentricity distribution that has a very low root-mean-squared value of e \sim 0.01 ; but about a quarter of the pairs possess much higher eccentricities , up to e \sim 0.1 - 0.4 .
For the low-eccentricity pairs , we are able to statistically remove the effect of eccentricity to obtain planet masses from TTV data .
These masses , together with those measured by radial velocity , yield a best fit mass-radius relation M \sim 3 M _ { \oplus } ( R / R _ { \oplus } ) .
This corresponds to a constant surface escape velocity of \sim 20 km / s .

We separate the planets into two distinct groups , “ mid-sized ” ( those greater than 3 R _ { \oplus } ) , and “ compact ” ( those smaller ) .
All mid-sized planets are found to be less dense than water and therefore must contain extensive H/He envelopes that are comparable in mass to that of their cores .
We argue that these planets have been significantly sculpted by photoevaporation .
Surprisingly , mid-sized planets , a minority among Kepler candidates , are discovered exclusively around stars more massive than 0.8 M _ { \odot } .
The compact planets , on the other hand , are often denser than water .
Combining our density measurements with those from radial velocity studies , we find that hotter compact planets tend to be denser , with the hottest ones reaching rock density .
Moreover , hotter planets tend to be smaller in size .
These results can be explained if the compact planets are made of rocky cores overlaid with a small amount of hydrogen , \leq 1 \% in mass , with water contributing little to their masses or sizes .
Photoevaporation has exposed bare rocky cores in cases of the hottest planets .
Our conclusion that these planets are likely not water-worlds contrasts with some previous studies .

While mid-sized planets most likely accreted their hydrogen envelope from the proto-planetary disks , compact planets could have obtained theirs via either accretion or outgassing .
The presence of the two distinct classes suggests that 3 R _ { \oplus } could be identified as the dividing line between ‘ hot Neptunes ’ and ‘ super-Earths . ’