The Maxwell field with a general gauge fixing (GF) term is nontrivial, not only the longitudinal and temporal modes are mixed up in the field equations, but also unwanted consequences might arise from the GF term. We derive the complete set of solutions in de Sitter space, and implement the covariant canonical quantization which restricts the residual gauge transformation down to a quantum residual gauge transformation. Then, in the Gupta-Bleuler (GB) physical state, we calculate the stress tensor which is amazingly independent of the gauge fixing constant and is also invariant under the quantum residual gauge transformation. The transverse components are simply the same as those in the Minkowski spacetime, and the transverse vacuum stress tensor has only one UV divergent term ($\propto k^4$), which becomes zero by the 0th-order adiabatic regularization. The longitudinal-temporal stress tensor in the GB state is zero due to a cancelation between the longitudinal and temporal parts. More interesting is the stress tensor of the GF term. Its particle contribution is zero due to the cancelation in the GB state, and its vacuum contribution is twice that of a minimally-coupling massless scalar field, containing $k^4$ and $k^2$ divergences. After the 2nd-order adiabatic regularization, the GF vacuum stress tensor becomes zero too, so that there is no need to introduce a ghost field, and the zero GF vacuum stress tensor can not be a possible candidate for the cosmological constant. Thus, all the physics predicted by the Maxwell field with the GF term will be the same as that without the GF term. We also carry out analogous calculation in the Minkowski spacetime, and the stress tensor is similar to, but simpler than that in de Sitter space.

26 pages, accepted by PRD