Applications to potential theory 169
After the conformal mapping w = f (z), the flow still obeys the same
boundary conditions, since f
′
(z) ∼ 1 as |z| tends to infinity in the upper
half-plane. The desired flow in the upper half-plane without a wall is easy to
find: it a a uniform flow (the wind over the plain...), which can be described
by the complex potential
Θ(w) = v
∞
× w.
Coming back to the variable w:
Φ(z) = Θ(w) = Θ
p
z
2
+ a
2
= v
∞
×
p
z
2
+ a
2
.
Two special points appear: the velocity of the flow is infinite on the top of
the wall, at z = ia (this is why the wind “whistles” when going over a pointed
obstacle: locally supersonic velocities are reached); it is zero “at the foot of
the wall” (around z = 0). These two points are of course the two points
where the transformation is no longer conformal
7
; it is therefore possible that
singularities appear in this manner, which did not appear in the flow on the
upper half-plane.
6.2.c Potential theory, lightning rods, and percolation
From the previous example we can understand the principle of th e lightning
rod. If we denote by ϕ( r) the electrostatic potential, then close to a pointed
object (like the wall above, or an acute angle in the definition of th e domain),
we can expect that ϕ
′
will diverge. Identifying C and R
2
, it is the gradient
of ϕ tha t will diverge, namely, the electric field. But we know that in the air
an electric field of intensity larger than some critical value, th e “breakdown
thresh old,” will cause a spont aneous electric discharge.
8
Saint Elmo’s fire, well
known to mariners, has the same origin.
But this is not yet the most remarkable thing. Consider a problem of
random walk of a particle in an open domain Ω ⊂ C, assuming that, when the
particle meets the boundary of Ω, it will stop. (It is possible to make numerical
simulations of this problem.) Wha t will happen after many particles have
performed this process? One may think that the part icles will accumulate on
the boundary ∂Ω, which will therefore g row, more or less regularly as time
goes. I n fact, this is not what happens. If the boundary ∂Ω ha s “cusps, ”
they will att ract more particles and, hence, will grow faster than the rest of
the boundary. If, at a given time and due to t he effect of chance, a “cusp” is
created (by accumulation of some particles) on a “ smooth” part of ∂Ω, then
this cusp will, in turn, a ttract more particles. In a word: the fluctuations of
the boundary become unst able.
7
Indeed, at t he foot of the wall, we have f
′
(z) = 0, whereas around the top, f
′
(z) diverges.
8
Of course, the rough explanation given only concerns ligh tning rods in a two-dimensional
world. In three dimensions, another argument must be found, because the one given uses
specific properties of C and not of R
2
.