How does ATP transfer energy to a reaction? This is a question for which I've found it surprisingly hard to find a good answer. Biology texts talk mystically about the ATP->ADP reaction providing energy to power other reactions. I'd like to know some more details. Is this following roughly right ?


*

*Each reaction in a cell has a specific enzyme.

*Each enzyme has binding sites for, say, two molecular species AND for an ATP molecule.

*When a reaction takes places, the two species bind to the enzyme, and a little later, an ATP molecule binds.

*For some reason (why ?), the ATP->ADP reaction is now energetically favourable, so the high-energy bond breaks.

*This releases electromagnetic energy at some characteristic frequency.

*Certain bonds in the enzyme have a resonant frequency that allow them to absorb this electromagnetic energy (the EM energy disturbs molecular dipoles ?).

*The 3D structure of the enzyme is disturbed (i.e. it bends) in such a way that the 2 molecular species are mechanically forced together, providing sufficient activation energy for the reaction in question.

*The newly formed species no longer binds nicely to the enzyme (why ?) so it detaches, as does the ADP, which also doesn't bind as nicely as ATP.

*The End.


So, is that an accurate summary ? Anyone care to add some more physical details ? (like all the thermodynamics and quantum chemistry I have no idea about) 
 A: This is rather a strange model; I'll took a chance to explain it from scratch. Basically all the chemical reactions look like this:
$$S_1 + S_2 + \cdots + S_n  \leftrightarrows P_1 + P_2 + \cdots + P_n$$
what means that a bunch of atoms forming few substrate molecules may rearrange into few product molecules. Both states have some energies and one postulates that the reaction involves some intermediate state when everything is mixed up (this is a simple view; since there may be few such states, few hidden subreactions, stuff -- I will not discuss it).  This can be plotted as:

where S are the substrates state, with energy $E_S$, than we heave an intermediate state of a higher energy $E_I$ and finally the products state of energy $E_P$. Now, the core of chemistry:


*

*The balance between atoms in substrate state and product state after infinite time is only governed by the energy difference $E_S-E_P$ (and temperature, pressure, and other stuff that we assume constant)

*The speed of the reaction depends on the intermediate state energy $E_I$, because we assume that the substrates must gather $E_I-E_S$ from the thermal fluctuations and collide do form intermediate state.


An enzyme forms a proper environment and usually improves spatial organisation of substrates to decrease $E_I$ and greatly improve the reaction speed. 
Now, in a cell, we have dozens of different molecules and thus huge amount of possible reactions. Moreover, the cell is under a constant matter exchange with environment, so the situation is dynamic and far from static equilibrium defined by $E_S-E_P$ differences. Because of that, the enzyme ability to change reaction speed is enough to form a dynamic equilibrium which we know as a metabolism.
So I can finally go back to the ATP role. Let's consider a situation where we have a reaction $S  \leftrightarrows P$ but $E_S<E_P$; this reaction will of course rather go the opposite way we want and the equilibrium will be shifted into substrates. Now, let's say we have also a reaction $S_X  \leftrightarrows P_X$ for which $E_{S_X}>>E_{P_X}$. One can expect that it is possible that the both reactions may co-occur creating reaction $S+S_X  \leftrightarrows P+P_X$ which would have positive $E_{S}-E_{P}$ difference, however in most cases this would be very improbable (corresponding to large $E_I$). Yet, this obstacle could be easy omitted -- with a help of an appropriate enzyme. 
Reassuming, enzyme role is to temporarily shift the balance of endoenergetical reaction by enlarging the probability of its co-occurrence with exoenergetic ATP to ADP+P transition. 
This is of course a trivial and very wrong picture =) You can start learning more here. 
A: I always pictured it in the following way, even though this is not what you read on Wikipedia, or in chemistry textbooks: 
ATP has three negatively charged phosphate groups, which are highly repulsive, but stuck together by a chemical bond which is too strong for the repulsion to overcome by itself. When ATP binds to protein, the protein breaks one of the phosphates off, and this phosphate pushes hard on the other two as it zooms away, doing mechanical work. This mechanically forces parts of the protein to close or open, providing energy for the reaction. There is nothing more magical than electrostatic forces between phosphate groups, and the forces are mostly electrostatic, not entropic.
The theoretical values for ATP energy content are always given in terms of the maximum energy that could be theoretically extracted, given the temperature and concentrations, if you run a perfect adiabatic heat engine. But I think this is misleading, because the proteins aren't that sophisticated. They waste a lot of the free energy, because they are just using mechanical forces, they let the outgoing phosphate heat up the water, they aren't doing things particularly adiabatically, and I believe the actual amount of useful work done for a protein by ATP is a fraction of the free energy content, essentially only the electrostatic repulsion potential energy between the $\gamma$ phosphate and the other two, and so it doesn't depend on the temperature or on the concentration.
A: It's usually more indirect than that.   You will need to know biochemical statistical thermodynamics of irreversible processes rather better.  I'll look for some references.
To go to your example of an Enzyme E catalyzing the exothermic reaction summarized by X + Y + E + ATP -> X-Y + E + ADP + P(i), I agree with the answer above that (4) the ATP -> ADP + P(i) reaction in water at standard lab conditions is always energetically favorable, i.e. free energy is always released, and entropy increased as a result of this reaction.  It is this increase in entropy which will drive the reaction in the forward direction.   
Eventually, the free energy released will be in the form of many far away infrared photons or phonons, which is much more probable (many more such states) than having it all localized in the single bond in ATP, so the reaction will go statistically in one direction.   
Your points 5 and 6 are overly specific.  The specific mechanism you refer to, resonant energy transfer can happen in some rather special cases, but it is unusual. Instead, more often enzymes form a temporary labile phosphorylated intermediate E-P which changes the charge in the active site and therefore the shape of the active site via 
a)E + ATP -> E-P + ADP  b)H2O + E-P ->E + P(i) as part of the overall reaction summarized above.  No resonance is necessary, since charge effects do all the work.
Other different enzymes work by forming labile intermediates with one of the substrates as part of the overall reaction summary.   E + X -> E-X  .  Many other mechanisms are also possible. 
The main point is that everything does not need to happen at once, since the statistics will determine a net direction for the coupled reactions which comprise the overall reaction to go, based upon the overall free energy change.
Another point is that the enzyme is always changing shape somewhat from the ordinary room temperature fluctuations it undergoes, according to the Boltzmann formula.   
To reiterate, reactions can be driven forward overall by statistics even if the early steps are energetically unfavorable.   So the whole situation is much more diverse than you described (which is why it is chemistry).  A good example of this last mechanism is in the synthesis of nucleic acids from nucleotide triphosphates, the overall reaction being driven forward by the much later release of free energy associated with the hydrolytic splitting of the previously released pyrophosphate molecule
H2O + P-P -> 2 P(i).    
A: there are some good answers here and I'll agree it is important to understand the thermodynamical aspects of enzyme function, but I sympathize with the OP on this point: down at the lowest level you still want to describe what happens with the bonds and energy even if the macroscopics of the system (temperature, pressure, configuration) controls the reaction speeds. After all, chemistry is low-energy physics, and thermodynamics is big-number physics.
Below is linked a paper (from 2005) which summarizes the then-current knowledge about ATP hydrolysis in the F1-ATPase, one of the most strikingly mechanical enzyme complexes in nature (see the linked YouTube-video!). It's basically a stator and a rotor, which can convert between ATP<->ADP+P and mechanical rotation of the rotor. It is one of the crucial proteins that power respiration in the mitochondrias in human cells in a very mechanical way - protons fall down a gradient through the mitochondrial membrane, pulling the rotor around, which turns ADP+P into ATP. In the bacterial flagelli (tails) it is used the other way, ATP powers it as a motor which enables the bacteria to swim. 
The 3D-structure of the F1-ATPase has been known for some time and there are some beautiful images of it on the web but the process is not completely understood. The paper discusses results from experiments and atomic simulations (one of my research interests) which give clues to exactly how the ATP bond energy is transferred. 
As in most protein functions, you have a very mechanical function (as the OP suggests in pt. 7) in combination with redistribution of Coulumbic charge as well as hydrogenic and covalent bonds. Some speculate that elastic storage of mechanical energy can participate (and hence the possibility of phonons I guess) - but I have never heard of any suggestions of intra-protein photon exchange as the OP suggests in pt. 5-6.
I tried posting the link to the researcher's homepage of the paper linked below but this site didn't allow me posting more than 2 links in a post :) Sorry..
Zooming in on ATP hydrolysis in F1
ATP F1 ATPase youtube video
A: The double welled potential diagram mbq posted is an operative diagram for the energetic state of a molecule, call it M, which is phosphorylated ATP + M --> ADP + M-P.  The phosphorous atom binds onto an amino acid residue group in a polypeptide chain, often threonine or tyrosine in the case of kinases.  Kinases are enzymes which initiate biochemical pathways.  The amino acid residue which is phosphorylated has dihedral bond angles with adjacent amino acid residues in the chain.  The unphosphorylated and phosphorylated bond angles are substantially different, reflecting the two minima in the double welled potential.  The result is the enzyme exhibits two conformational shapes.  These shapes can act as binary conditions, literally on and off, for turning on and off a biochemical pathway.  
This may be compared to Landauer’s principle which relates a binary information condition to thermodynamics.  ATP contains 76j/mole of energy, which we can identify as the E.  The work or useful energy extracted is the change in the energy from the lower portion of the well to the higher.  Also there is energy in the entropy due to a chemical change, which for the two states is $k~ln(2)$.  Therefore, the free energy per mole $F~=~ 76j/mole$ must be greater than the entropy for a change in the configuration or shape of the kinase or other molecule.
