> For those components where `boundary` is set to "none", any value of
> `edges` or `image` must be silently ignored.
Any value of `image` then. Edges only play a role if 'image' is different from
zero.
No! That is exactly *not* true. It makes perfect sense to specify periodic boundary conditions, an edge, but no image! This is exactly the case when you use what I call absolute coordinates, where the positions may be outside the box.
To clarify, let me try to make some examples of possible cases, all of which do make sense! I am now specifying only one dimension and a single particle with a position x=20.3. In the periodic case, the edge is 10.0.
Case 1: Open boundaries
boundary = "none"
position[0] = 20.3
In this case, any value of 'edge' or 'image' has to be ignored.
Case 2: Periodic boundaries, with folded positions and image specification
boundary = "periodic"
edge[0] = 10.0
position[0] = 0.3
image[0] = 2
This is probably the standard case that most people think of.
Case 3: Periodic boundaries, with absolute positions and without image
boundary = "periodic"
edge[0] = 10.0
position[0] = 20.3
This case also makes sense when a software never explicitly computes the image the particle is in.
Case 4: Periodic boundaries, with folded positions that may lie outside the primary image ("mixed")
boundary = "periodic"
edge[0] = 10.0
position[0] = 10.3
image[0] = 1
This case might look weird, but it does make perfect sense in some simulation packages.
Olaf