complex (Hermitian) inner product

In Vol I the dot product on R^n let you measure lengths and angles. Over the complex numbers the naive sum sum z_k w_k fails one basic test: a vector should have a real, nonnegative length. The fix is to conjugate one factor, so <z,z> = sum z_k conj(z_k) = sum |z_k|^2, which is a genuine nonnegative real number.

Precisely, a complex inner product on a complex vector space V is a map <-,-> : V x V -> C that is linear in one slot (we take the first), conjugate-symmetric (<u,v> = conj(<v,u>)), and positive-definite (<v,v> >= 0 with equality only for v = 0). Conjugate symmetry forces <v,v> to equal its own conjugate, hence to be real, so positive-definiteness even makes sense.

Because the form is linear in the first slot but conjugate symmetry flips a conjugate onto the second, it is conjugate-linear in the second slot: <u, c v> = conj(c) <u,v>. That asymmetry is the price of real lengths, and it propagates everywhere downstream — to the adjoint, the polarization identity, and the spectral theorem.

A caveat on conventions: physicists often make the SECOND slot linear (Dirac's bra-ket), mathematicians usually the first. The math is identical; only the conjugate moves. Pin down your convention before computing or every conjugate will land on the wrong side.