Reference paper to support information -- energy relation $\left(kT \ln2 \rm\frac{J}{bit}\right)\;.$ In answer to Maxwell's Demon Constant (Information-Energy equivalence) there is stated that one bit of information allows to perform $kT \cdot \ln2$ Joules of work. Which paper supports the thesis? (there are many publications on Maxwell daemon, Szilard engine, Landauer's principle).
 A: See e.g. page 3 of 

http://arxiv.org/abs/0707.3400

It's nonsensical to attribute this simple particular insight to a "discoverer"; all these considerations should be associated with Ludwig Boltzmann who knew the answer even though the information in physics was considered continuous at that time. 
One may easily derive the result. For example, put one molecule of an ideal gas in a vessel, learn in which half of the vessel the molecule is (one bit of information), and put a barrier in the middle.
You will be able to allow the molecule to do the work and expand from $V/2$ to the original volume $V$. The molecule will do the work
$$ W = \int_{V/2}^V p\,dV = \int_{V/2}^V \frac{kT}{V}dV = kT \ln \frac{V}{V/2} = kT\ln 2 $$ 
where I used $pV = NkT$ for $N=1$ molecule of an ideal gas. More generally, you don't have to consider ideal gas. Just recall how work is related to the free energy, $E-TS$. To reduce the entropy of a subsystem by one bit, i.e. by $k\ln 2$ (look at Boltzmann's tomb formula to know why it's this value), we have to do change $E-TS$ by $kT\ln 2$.
