Physical explanation of osmosis I want to understand the forces involved in osmosis. If I have a molecule of water and one of a salt in the left side of a semi-permeable membrane, and a water molecule at the right side, what forces make the right side molecule travels left and remains there what forces make water molecules be more time at the left side of the membrane?
I read about a chemical potential difference between the initial and final states, but I can't understand why there is less energy in solution with more solvent. Why the impurities cause the chemical potential to reduce at the left side?
I understand that osmosis happen in other scenarios, like gases, therefore I discard think in the ionic solution or polarity of water molecules, etc.
On this site, I found:

The decrease in chemical potential occurs because there is a lower
concentration of water in the solution than in the pure liquid.
Statistically, fewer water molecules escape a solution into the vapor
phase or freeze out onto the solid phase. That's why salt lowers the
chemical potential of water in the solution.

But is still unclear to me why the salt has this effect on water molecules if their kinetic energy remains the same with or without the salt.
 A: Usually the driving force for diffusion is a concentration gradient, however that is not strictly true. It is actually a gradient in chemical potential which is the driving force for diffusional processes. For some simplifying assumptions this reduces again to concentration gradient as driving force. 
For osmosis, the chemical potential is the driving force as the presence of any impurities cause the chemical potential to reduce compared to the pure side of the membrane. This is modeled by an equation of the form:
$$\mu_i = \mu_{i,0} + RT\ln x_i$$ 
where $\mu_i$ is the chemical potential of species i, $\mu_{i,0}$ some reference state of chemical potential, $RT$ is the specific molar energy and most importantly $x_i$ is the mole fraction of solute $i$ in the solvent.
Now as $0\le x_i\le 1$, it follows that $\ln x_i<0$ and $\Delta\mu_i=\mu_i-\mu_{i,0}<0$, i.e. there is decrease in chemical potential in the solvent leading to diffusion of across the membrane.
A: John Grant Watterson claims in his paper "What drives osmosis?" that a fall in free energy, which is the rigorous thermodynamic criterion for a spontaneous change, cannot be the drive in osmotic processes: 

Our models and theories require the introduction of a parameter that
  explicitly represents structure in liquids, which until now has had no
  place in the thermodynamic description of solutions. This lack is
  surprising, when one remembers that experimental results from the
  broad range of fields of colloid, clay and biological sciences have
  clearly established the marked effect of solutes on the structural
  properties of water, globally called ‘hydration phenomena’.
The introduction of such a parameter can help explain the direction in
  which energy flows during osmosis, which has been so puzzling to those
  of us interested in mechanism since the time of Pfeffer, more than a
  century ago. Further, elementary work cycles show, that changes in
  this parameter correspond to changes in the energy associated with
  solvent structure which can be used to produce useful work. The
  ability of osmotic systems to do work is familiar to all of us
  (indeed, a nuisance to many!), and is the basis of cytomechanics,
  i.e., the physical processes observed in the living cell. The fact
  that it still has no satisfactory explanation is clearly an urgent
  problem for us all.

In his article "Quantum Worlds", James Watson reviews osmotic theory of Watterson:

Osmosis is viewed as a direct result of the wave
  structure of water or, more specifically, of the structural aggregates
  of solvent molecules known as wave units or clusters. It is the
  structure wave itself, and not the solutes, that governs the molecular
  motions underlying osmosis. Since solvent can move through the
  semipermeable membrane, it can be considered as a single continuous
  medium pervading both phases. This means that the structure wave can
  pass unhindered from one phase to the other transferring structural
  energy in the process. Addition of solute breaks down the extent of
  solvent-solvent cooperative interactions because the molecules in
  contact with the solute can no longer rotate as freely as before. As a
  consequence, the wave length is shortened in the solution resulting in
  clusters smaller in size and energy but increased in number
  (concentration). In other words, the solute causes a decrease in the
  size of the pressure pixel. The increase in concentration of clusters
  in the solution phase is equal to the concentration of solute
  molecules.
At the membrane, there is a net energy flow from the energy-rich
  clusters of solvent into the smaller clusters of solution. This
  increases the tensile strength of the intermolecular bonds, so that
  the smaller clusters can pull solvent across the membrane increasing
  the pressure on the solute side. At equilibrium, the pressure in the
  solution has become high enough to counteract the pull of the smaller
  clusters and flow equalizes. At this point, the energy of the smaller
  clusters equals that of the pure solvent clusters.

Finally, Stephen Lower agrees with a quantum explanation in his textbook section "Some Applications of Entropy and Free Energy":

Dilution of a liquid creates uncountable numbers of new microstates,
  increasing the density of quantum states in the solution compared to
  that in the pure liquid. To the extent that these new states are
  thermally accessible, they will become populated at the expense of
  some of the microstates of the other phase.

A: In "Five popular misconceptions about osmosis", Kramer and Myers write:

The force is exerted on the solvent by the semi-permeable membrane. A simple thought-experiment illustrates how this happens. Consider an idealized semipermeable membrane as a force ﬁeld that repels solute but has no eﬀect on the solvent.The Brownian motion of the solute molecules bring them into occasional contact with this ﬁeld, at which time they receive some momentum directed away from the membrane. Viscousinteractions between solute and solvent then rapidly distribute this momentum to the solvent molecules in the neighborhood of the membrane. In this way, the membrane exertsa repulsive force on the solution as a whole. Since additional pure solvent can freely crossour ideali zed membrane, it ﬂows into th e solution compartment, gradually increasing thehydrostatic pressure in the solution. Thus, a pressure gradient builds up across the thickness of the membrane. This pressure gradient exerts a second force on the solution, capableof counteracting the membrane force.

