This question was asked before on stackexachange. The answer is to use Eulers relation for plane graphs:
faces = 2 + edges - vertices
Two circles can intersect at at most 2 points. Thus 4 circles intersect at at most 2 + 4 + 6 = 12 points (you have to start with one vertex by default, to make Eulers formula hold). The version with 12 intersections should have the most faces, every time you have less intersections, you also have less faces (I hope thats right, I am tired).
max faces = 2 + edges - 12
So we only have to figure out the number of edges a graph of four circles can possibly have.
Assume you have a graph of n circles and you add one circle. This will add new edges, depending on how many new vertices we have. On the new circle itself it will add at most 2n new edges. Every time the new circle "cuts through" an existing edge, he will devide this edge, creating at most one more in total. Thus, at most 2n + 2n = 4n new edges.
Start with one circle: 1 edge
Two circles: at most 1 + 4 = 5 edges
Three circles: a most 5 + 4 * 2 = 8 + 5 = 13 edges
Four circles: at most 13 + 4 * 3 = 25 edges
thus, max faces = 2 + max edges - 12 <= 2 + 25 - 12 = 15.
It seems like I overcalculated the number of possible new edges in the step from one to two circles. This is because the two new intersections must devide the same existing edge (there is only one) and therefore creating only one new edge on the first circle (+ the two on the second circle, thus at most three new edges in total). So the number of edges four circles can have at most is actually 24. But 25 is good enough to show that the diagram is impossible.
I hope I didn t write completely nonsense. Its pretty late now. Youre welcome to correct :)
Eulers Formula is really elegant though, so you should be happy using it in one way or other.
Edit: I just realized there is a problem with Eulers Formula if you start with only one circle. So it is probably better two start with two. But the argument should look similar then, maybe +/- 1 face. gn8