The (conserved) component of the group velocity normal to the magnetization axis (i.e., the Chern vector direction) has a welldefined sign and surface states cannot back-scatter along this specific direction.In 2D, the Chern vector is always fixed along the axis of the reduced dimensionality, that is, orthogonal to the plane of the system. Therefore, it can be regarded as a scalar quantity: the Chern number C, which characterizes the bulk topology of 2D CIs. [7][8][9] In this case, a "scalar" version of the bulk-boundary correspondence (sBBC) can be defined, to connect the bulk topology to the number of boundary modes. [10,11] According to sBBC in 2D CIs, an interface between two systems with Chern numbers C 1 , C 2 has n e = |C 1 − C 2 | protected chiral edge states. This means that chiral edge states can only appear in presence of a discontinuity of Chern numbers across the interface, that is, C 1 ≠ C 2 . [12][13][14][15] On the contrary, the interface between two 3D CIs can have Chern vectors C 1 and C 2 that need not be parallel/anti-parallel to each other. [2,16] This leads to vectorial aspects of the bulk boundary correspondance (BBC) which must be taken into account in order to correctly define a "vectorial" BBC (vBBC).In the broader context of topological insulators, the literature has discussed the vectorial aspects of BBC mostly in relation to 3D quantum spin-Hall fermionic insulators. [17][18][19] Such systems are characterized by vectors of 2 invariants and rely on time-reversal symmetry (TRS). However, the photonic 3D CIs studied in this manuscript break TRS and display non-zero Chern vectors, requiring a suited vectorial BBC. While scalar bulk-boundary correspondence is a well-established concept in 2D photonic crystals with broken TRS, [12][13][14][15]20] less is known in photonic 3D CIs, recently demonstrated in ref. [6] which is what motivated our work. We believe that this limited knowledge is related to the fact that, in condensed matter systems, the 3D quantum Hall effect is usually discussed in layered systems. [21][22][23][24][25][26][27][28] Since layered systems have a preferred axis, and since the magnetic field needs to be along this axis, the vectorial nature does not come up. On the contrary, the possibility of orienting Chern vectors in space demonstrated in photonic crystals, [6] open up the possibility of constructing domain walls between different orientations, and thus a definition of a vectorial BBC is required.