How exactly does the Peltier effect work? I'm having some trouble finding details on just how the Peltier effect (also known as the Peltier-Seebeck effect or the thermoelectric effect) works on a physical level. Am I correct in thinking that it has something to do with the small energy barriers that occur at ohmic contacts? I've searched extensively and the best information I can find is either lacking in details or hidden behind a paywall I can't afford.
What I mostly want to know here is: What causes the charge carriers on the cold side to pick up energy from the crystal lattice? What phonon-electron or phonon-hole interaction occurs?
The Seebeck effect (converting temperature to a potential difference) makes more intuitive sense to me, as diffusion-driven high-energy/high-density charge carriers move across the semiconductor (or metal), but this would seem to imply that it would be something that occurs in the bulk of the material, rather than at the junctions. And I don't know if just reversing that even makes sense to explain the Peltier effect (which is the reverse, current to temperature). Since the Peltier and Seebeck effects are considered to be the same thing operating in two different directions, a diffusion-based explanation would seem not to work.
 A: Here is my understanding of the Peltier effect, obviously, grossly simplified. I'll be happy to be corrected.
When electrons cross a junction from a material with a lower electrochemical potential to a material with a higher electrochemical potential, they must lose some kinetic energy, which should lead to the cooling of the junction. 
Similarly, electrons crossing from a material with a higher electrochamical potential to the material with a lower electrochemical potential must be gaining some kinetic energy and therefore cause local heating.
Although the Seebeck effect is also caused by the difference in electochemical potentials, its explanation does not directly follow from the Peltier's. 
My understanding is that it is based on the difference in the dynamic equilibrium of the hot and cold junctions. Heating of a junction has a greater effect on the electrons in the material with the lower electrochemical potential and, correspondingly, causes a shift in the dynamic balance. 
The resulting difference in the dynamic balance point between the two junctions forces re-distribution of electrons between the junctions, and since, this process continues as long as one of the junctions is being heated, there is a continuous re-distribution of the electrons, i.e., current.     
A: No, it does not depend on junctions. The elements work with $p$- and $n$-type materials arranged like this:

The electrical current goes back and forth between the hot plate and the cold plate. But the conductors are arranged in such a way that the charge carriers always go in the same direction, carrying heat. So the heat flow in this figure is upward, for all semiconducting connections.
