Due to a processing error the last paragraph of Sect. 3.1 on p. 358 was printed with an error. In addition, in Sect. 5.2 on pp. 365-366 the presentation of the Example was processed incorrectly. The second paragraph of Sect. 5.2 (two lines) must be removed and the last sentence in the same section must be replaced with 'Nevertheless, even though σ 0 maximizes in C 1 , it does not maximize in C 2 (nor in the causal homotopy class C), in agreement with our results.' The corrected paragraphs read as follows.
p. 358:In particular, if x 0 x 1 the two points can be connected by means of a timelike geodesic (in fact, by one for each time like homotopy class in C x 0 ,x 1 , as will be apparent below). If x 1 ∈ E + (x 0 ) = J + (x 0 )\I + (x 0 ) then x 0 and x 1 can still be joined by a lightlike geodesic, but this case does not make sense for the LFE. One can also wonder for the connectedness of x 0 , x 1 by means of a geodesic even if they are not causally related, as in variational frameworks described below. Although this question has a geometrical interest (see for instance the survey [37]), it does not have a direct physical interpretation, nor equivalence for LFE.pp. 365-366: 5.2. A remarkable example. Lemma 5.1 does not forbid the existence of a lightlike geodesic σ which maximizes the functional on the closure of a timelike class C x 0 ,x 1 . However, in that case the maximizer on C x 0 ,x 1 ⊃ C x 0 ,x 1 does not coincide with σ , as the following example shows.Example. Let be a surface embedded in R 3 obtained by gluing the spherical cap2 r + z with a cylinder x 2 + y 2 = r 2 /4, z < − √ 3 2 r − z ,The online version of the original article can be found at http://dx