First of: the energy W does not increase exponentially it increases quadratic!
Second:
What do you want?
Energy density!
So to get good high energy density you need high C because you can (as you already stated) only go to a certain amount of voltage until you get a breakdown. The problem is, C does not change easily. If you take a simple Plate - Plate capacitor
$C \propto \frac{A}{d}$ where A is the Area of the plates and d the distance between them. The smaller the distance the higher the electric field gets between the plates gets relative to Voltage ($|\vec E| \propto \frac{U}{d}$) the faster you get a spark. So the only thing you can actually increase safely is the area. That of course increases the wight/size of you capacitor, reducing the Energy density. This generally applies to all capacitors
Wikipedia does have an excellent article about capacitors:
http://en.wikipedia.org/wiki/Capacitor
Third:
About the capacitance:
A spring is a very good analog to your problem:
A newtonian spring has a certain flexibility, describing the force you have to pull it relative to the expansion of the spring:
$F=k \times x$
Where F is the Force k is the flexibility and x the way. So the capacitance is actually k in this analog. But F is not the Energy W.
$E = \int_0 ^x F(x) dx$
So the more you pull the harder it gets to pull because the force from the spring increases. So to get the energy you can not just calculate $E = F \times x$ but you rather have to integrate the way.
That of course gets you to
$E = \frac{1}{2} k x^2$
Back to the Capacitor:
The more electrons you put in it, the stronger becomes the electric field between the plates of the capacitor. So you have to increase the voltage. The capacitance just tells you how high your voltage has to be (how hard you have to push) to put more electrons on the plate.
Toby, I agree that this is really counter intuitive and I was also quite surprised as well when I first saw this very demonstration. I am an undergraduate TA and this is how I explained it in my lab section. I hope this helps. I see two parts to a full explanation: (1) Why is the electric field constant and (2) why does the potential difference (or voltage) increase?
Why is the electric field constant as the plates are separated?
The reason why the electric field is a constant is the same reason why an infinite charged plate’s field is a constant. Imagine yourself as a point charge looking at the positively charge plate. Your field-of-view will enclose a fixed density of field lines. As you move away from the circular plate, your field-of-view increases in size and simultaneously there is also an increase in the number of field lines such that the density of field lines remains constant. That is, the electric field remains constant. However, as you continue moving away your field-of-view will larger than the finite size of the circular plates. That is, the density of field lines decreases and therefore, the electric field decreases as well as the potential field.
To show this mathematically, the easiest way to show this for E = constant is using the relation between the electric and potential fields:
$$E = -\frac{\Delta V}{\Delta d} \longrightarrow \Delta V =-E \Delta d$$
I would expect the voltage to increase linearly as long as the field is constant. When the electric field starts decreasing, the voltage also decreases and the fields behave as finite charged plates. Although I’ve only talked about one plate, this idea immediately applies to two plates as well.
Why does the work increase the electrical potential energy of the plates?
One way to interpret why the voltage increases is to view the electric potential (not the electrical potential energy) in a completely different manner. I think of the potential function as representing the “landscape” that the source (of the field) sets up. Let me explain what the gravitational potential acts like when a ball is thrown upwards (of course, you know what happens in terms of the force of gravity or in the conservation of energy scenario). I claim that the potential function is related to the “gravitational landscape” that the earth sets up, which is derived from the potential energy and is equal to the potential energy per mass:
$$ {\Delta U = mg\Delta y} \longrightarrow \frac{\Delta U}{m} = \Delta V = g\Delta y$$
Plotting these functions, the constant gravity field sets up a gravitational potential ramp (linear behavior) that looks like
In terms of energy, the ball moves up this gravitational ramp were the ball is converting its kinetic energy into potential energy until the ball reaches it maximum height. However, the gravitational ramp exists whether the ball is thrown up or not. That is, gravity sets up a gravitational ramp (the landscape) and this is what the ball “sees” before it is thrown up.
If we now apply the above thinking to a constant electric field between the parallel plates, the electric potential function is derived in a similar manner:
$$ {\Delta U = qE\Delta r} \longrightarrow \frac{\Delta U}{q} = \Delta V = E\Delta r$$
If we look at the electric potential of the negative plate (it’s easier than the positive plate), it has a negative electrical ramp that starts at 0V.
So as your TA pulls the plates apart, the work she does moves the positive plate up the electrical ramp and increases the potential of the positive plate. So this interpretation of the electric potential is what you intuitively already think about in terms of mechanical situations like riding your bike up a hill. There is no difference in the electrical situation.
Best Answer
The answer is that your analysis has no resistance in the circuit.
Assume that the initial charge on the capacitor is $Q$, the final charge on the capacitor is $\dfrac Q2$ and the emf of the battery is $V$.
You have worked out the work done in sending charge through the battery during the separation process is $\dfrac Q2 V = U$ and there is a balanced energy equation.
If there is a resistive component in the circuit then suppose that during the separation of the plates phase the potential difference across the resistive part is $v$.
This means that the total potential difference across the battery and the resistive part of the circuit is $V+v$.
As the plates of the capacitor are being separated $V+v$ must also be the potential difference across the plates of the capacitor.
This means that the force between the plates of the capacitor, which depends on the potential difference across the plates, is increased which in turns means more external work need to be done in separating the plates.
That extra work done by the external force in separating the plates is the source of the heat dissipated by the resistive part of the circuit.