Why do electric field lines curve at the edges of a uniform electric field? I see a lot of images, including one in my textbook, like this one, where at the ends of a uniform field, field lines curve.

However, I know that field lines are perpendicular to the surface. The only case I see them curving is when drawing field lines to connect two points which aren't collinear (like with charged sphere or opposite charges) and each point of the rod is collinear to its opposite pair, so why are they curved here?
 A: Rather than thinking of the plates as solid line charges, think of them as lines of infinitely many point charges.
2 point charges  will have a straight field-line directly between them, and weaker curved field-lines outside of that.  If you place 2 pairs of point charges next to each other — positive with positive, and negative with negative — then their field lines can overlap.  However, much like with waves, electric fields can interfere with each other, both constructively and destructively.
This means that the overlapping curved field lines will average out as a straight field line, through the middle of the point-charge pairs.  The outermost edges of the electric field, on the other hand, will have nothing to interfere with, and remain curved.

A: I have taken your image and created a few additional field lines at one end of the plates in the first diagram below.
When you come to the ends of the plates, the field starts to resemble that associated with two point charges instead of a sheet of charge. The second diagram below shows the field lines between two point charges. Note that as you move away from the two point charges an equal distance apart, the lines look like those at the ends of your parallel plate capacitor (curved lines). Towards the center between the charges, the field lines start to look straight and evenly spaced (parallel lines).
Hope this helps.


A: These are so-called edge effects. The straight electric field lines connecting two surfaces is a solution for the infinite charged plates. In practice, no plates are infinite: they have edges. Far from the edges (close to the center of the plates) one can still think of the plates as infinite, but at the edges this is clearly not true.
Note that the same is true for an infinite charged wire or cylinder: in practice one always has a finite one, but far enough from the edges, one can assume that it is infinite and thus simplify the math.
A: This is one of those questions where you just have to see it. Here is fieldline drawing of two charges. Red is a positive charge and blue is negative.

Now for 6 charges:

and finally for 40 charges:

Here is the Mathematica code for anyone interested
range = 1.4;
nCharges = 20;
xSeparation = .5;
e[r_, r0_] := (r - r0)/Norm[r - r0]^3
chargeY[n_] := If[nCharges == 1, 0, (n - 1)/(nCharges - 1) - .5];
Show[
  StreamPlot[
    Sum[e[{x, y}, {-xSeparation, chargeY[n]}], {n, 1, nCharges}] - 
    Sum[e[{x, y}, {xSeparation, chargeY[n]}], {n, 1, nCharges}], 
    {x, -range, range}, {y, -range, range}],
  ListPlot[Table[{-xSeparation, chargeY[n]}, {n, 1, nCharges}], 
    PlotStyle -> {Red, PointSize[.03]}],
  ListPlot[Table[{xSeparation, chargeY[n]}, {n, 1, nCharges}], 
    PlotStyle -> {Blue, PointSize[.03]}]
 ]

