DERIVATIONS OF (8)-(10) AND (12)-(14) The following notations are used throughout this document. All boldface letters indicate vectors (lower case) or matrices (upper case). (⋅) , (⋅) , and ‖⋅‖ stand for Hermitian transpose, matrix inversion, and Frobenius norm, respectively. The notation (⋅,⋅) denotes complex Gaussian distribution with mean and covariance. E{⋅} denotes statistical expectation. ℰ () denotes the eigenvector of matrix corresponding to the largest eigenvalue λ () , i.e. ℰ () = λ ()ℰ (). 1. Derivations of (8)-(10) According to the definition of , it can be expressed as: = ! . = ! + # where #: (0, & ' () represents circularly symmetric complex additive white Gaussian noise (AWGN) vector with power & ' ; > 0 denotes the average energy per symbol; ∈ ℂ denotes the information symbol with unit energy, i.e. E{| |' } = 1. Similarly, we have Therefore, we have / 0 = − . = + #. / (! According to the definition of 0' , it can be expressed as: 0' = 2'22'2 Therefore, the objective function (7) is rewritten as: − ! ) . / ' ! 3 min =E>‖0 ‖' ? + E>‖0' ‖' ?@ !/789:;7< = min A !/789:;7< E BC / >! +E BC2'22'2 ' − ! ? C D F. ' / ' ! C D After applying first-order derivative to the above function, we have !/ = (H + H ' ) H ! , (cf. (8)) H ≜ H ' ≜ ' 2'22'2 2'22'2 , ' . Ideally, the optimal un-quantized precoding vector ! = ℰ ( 1). Then, we have = B JKL B! D . Hence, we define JKL JKL ! D JKL /N ( ). ) (cf. [6], lemma As a result, H = ! JKL JKL H ≜ ! ! (cf. (9)) Furthermore, according to (6), the optimal solution for ! can be obtained by solving the following optimization function: O max C2'22'2 JKL O ? Hence,>! ' O ' ! C O O ? ! s. t.>! =1 = ' 2'22'2 /‖ ' 2'22'2 ‖. Then, we have O O >! ? (cf. (10)). H ' ≜ ! 2. Derivations of (12)-(14) We can rewrite (11) as: min !/VWV:VWV A E BC2'22'2 / '' >!2'22'2 +E BC ' − !2'22'2?' C D ' / ' !2'22'2 ' C D F. Notice that the rewritten version of (11) has the same pattern as the rewritten version of (7). Therefore, (12)-(14) can be easily derived following the same derivation procedure shown in Section 1.