Sigma bonds are largely defined by the geometry of the orbitals interacting to form the bond. The electrons in an atom are generally confined to certain regions, called orbitals, in order to satisfy as many of the various electromagnetic interactions as they can. Depending on their distance from the nucleus and the energy they represent, the orbitals can take on different shapes. The outermost orbitals tend to be the highest in energy, and therefore the ones that are most likely to participate in a bond. In iron, the outermost electrons are in the D orbital, and in nitrogen, they're in the P orbital. Both orbitals have regions shaped like dumbbells that protrude outward from the nucleus along a given axis.
In order to form the best kind of bond, two orbitals should overlap in as "linear" a fashion as possible. That is to say, if we evaluate one of those dumbbell-shaped regions as if it's on an axis traveling through the nucleus, then the best kind of bond will be formed when the other atom is aligned with its own orbitals on that same axis. Basically, if you drew a line connecting the two nuclei, it would pass straight through the middle of both orbitals. When this happens, we call that bond type "sigma." In contrast, a "pi" bond is one that takes place between orbitals that aren't aligned along the same axis.
In practice, we can't really say that "there will be a sigma bond between iron and nitrogen," because the complexity of iron makes it difficult to assign particular bonding characteristics to every possible interaction. Iron is also capable of ionic interactions, as well as "coordinate bonding," which is an alternate type of covalent bond that can form following reorganization of the iron's electrons.