Imaginary zeros of f(z)= z^(1/z) (real valued solutions f(z)>e^(1/e))

[Ivars]

The initial spider graph of Gottfried looks very much like Caley transform, except that it is has 2 conjugate angled maps of slightly elongated circles, and the lines that should be inside Unit circle in Cayleys transform have moved to the negative part of real axis.

http://mathworld.wolfram.com/CayleyTransform.html

What transform would produce such a map from real axis as Gottfrieds spider?

By the way, the result h(e^pi/2) = +-i also is a result of transform (e^pi/2)^ ((i-1/i+1)) and (e^pi/2)^((-i+1)/(-i-1))

Here the angle by which reals are turned is +-pi/2. As the values go above e^pi/2, angles increase but later decrease again. So the transform is definitely more complex than just:

e^Real^((i-Real)/(i+Real) which is true in case Real=e^pi/2.