Let γ_{1} be any maximally extended, future-directed, null geodesic confined to a submanifold N whose points all have some particular z ̃ value. Let q be any point in N whose r coordinate satisfies sinh^{2}r = (√2 − 1)/2. Pick any point on γ_{1}. By virtue of the homogeneity of Gödel spacetime, we can find a (temporal orientation preserving) global isometry that maps that point to q and maps N to itself. Let γ_{2} be the image of γ_{1} under that isometry. We know that at q the vector (t ̃^{a} + kφ^{a}) is null if k = 2(1 + √2). So, by virtue of the isotropy of * Gödel spacetime*, we can find a global isometry that keeps q fixed, maps N to itself, and rotates γ

_{2}onto a new null geodesic γ

_{3}whose tangent vector at q is, at least, proportional to (t ̃

^{a}+ 2(1 + √2)φ

^{a}), with positive proportionality factor. If, finally, we reparametrize γ

_{3}so that its tangent vector at q is equal to (t ̃

^{a}+ 2(1 + √2)φ

^{a}), then the resultant curve must be a special null geodesic helix through q since (up to a uniform parameter shift) there can be only one (maximally extended) geodesic through q that has that tangent vector there.

The corresponding argument for timelike geodesics is almost the same. Let γ_{1} this time be any maximally extended, future-directed, timelike geodesic confined to a submanifold N whose points all have some particular z ̃ value. Let v be the speed of that curve relative to t ̃^{a}. (The value as determined at any point must be constant along the curve since it is a geodesic.). Further, let q be any point in N whose r coordinate satisfies √2(sinh2r)/(cosh2r) = v. (We can certainly find such a point since √2 (sinh2r)/(cosh2r) runs through all values between 0 and 1 as r ranges between 0 and r_{c}/2) Now we can proceed in three stages, as before. We map γ_{1} to a curve that runs through q. Then we rotate that curve so that its tangent vector (at q) is aligned with (t ̃^{a} + kφ^{a}) for the appropriate value of k, namely k = 2 √2/(1 − 2 sinh^{2}r). Finally, we reparametrize the rotated curve so that it has that vector itself as its tangent vector at q. That final curve must be one of our special helical geodesics by the * uniqueness theorem for geodesics*.

The special timelike and null geodesics we started with – the special helices centered on the axis A – exhibit various features. Some are exhibited by all timelike and null geodesics (confined to a z ̃ = constant submanifold); some are not. It is important to keep track of the difference. What is at issue is whether the features can or cannot be captured in terms of g_{ab}, t ̃^{a}, and z ̃^{a} (or whether they make essential reference to the coordinates t ̃, r, φ themselves). So, for example, if a curve is parametrized by s, one might take its vertical “pitch” (relative to t ̃) at any point to be given by the value of dt ̃/ds there. Understood this way, the vertical pitch of the special helices centered on A is constant, but that of other timelike and null geodesics is not. For this reason, it is not correct to think of the latter, simply, as “translated” versions of the former. On the other hand, the following is true of all timelike and null geodesics (confined to a z ̃ = constant submanifold). If we project them (via t ̃^{a}) onto a two-dimensional submanifold characterized by constant values for t ̃ as well as z ̃, the result is a circle.

Here is another way to make the point. Consider any timelike or null geodesic γ (confined to a z ̃ = constant submanifold). It certainly need not be centered on the axis A and need not have constant vertical pitch relative to t ̃. But we can always find a (new) axis A′ and a new set of cylindrical coordinates t ̃′, r′, φ′ adapted to A′ such that γ qualifies as a special helical geodesic relative to those coordinates. In particular, it will have constant vertical pitch relative to t ̃′.

Let us now consider all the timelike and null geodesics that pass through some point p (and are confined to a z ̃ = constant submanifold). It may as well be on the original axis A. We can better visualize the possibilities if we direct our attention to the circles that arise after projection (via t ̃^{a}). The figure below shows a two-dimensional submanifold through p on which t ̃ and z ̃ are both constant. The dotted circle has radius r_{c}. Once again, that is the “critical radius” at which the rotational * Killing field* φ

^{a}is null. Call this dotted circle the “critical circle.” The circles that pass through p and have radius r = r

_{c}/2 are projections of null geodesics. Each shares exactly one point with the critical circle. In contrast, the circles of smaller radius that pass through p are the projections of timelike geodesics. The figure captures one of the claims – namely, that no timelike or null geodesic that passes through a point can “escape” to a radial distance from it greater than r

_{c}.

*Figure: Projections of timelike and null geodesics in Gödel spacetime. r _{c} is the “critical radius” at which the rotational Killing field φ^{a} centered at p is null*

Gödel spacetime exhibits a “boomerang effect.” Suppose an individual is at rest with respect to the cosmic source fluid in Gödel spacetime (and so his worldline coincides with some t ̃-line). If that individual shoots a gun at some point, in any direction orthogonal to z ̃^{a}, then, no matter what the muzzle speed of the gun, the bullet will eventually come back and hit him (unless it hits something else first or disintegrates).