Definition 1
Definition 1. Consider the partitioned matrix
A = [ A 11 A 12 A 21 A 22 ] A=\left[\begin{array}{ll} A_{11} & A_{12} \\ A_{21} & A_{22} \end{array}\right] A=[A11A21A12A22]
- When A 11 A_{11} A11 is nonsingular, A 22 − A 21 A 11 − 1 A 12 A_{22}-A_{21} A_{11}^{-1} A_{12} A22−A21A11−1A12 is called the Schur complement of A 11 A_{11} A11 in A A A, denoted by S c h ( A 11 ) S_{c h}\left(A_{11}\right) Sch(A11).
- When A 22 A_{22} A22 is nonsingular, A 11 − A 12 A 22 − 1 A 21 A_{11}-A_{12} A_{22}^{-1} A_{21} A11−A12A22−1A21 is called the Schur complement of A 22 A_{22} A22 in A A A, denoted by S ch ( A 22 ) S_{\text {ch }}\left(A_{22}\right) Sch (A22).
Lemma 1
Lemma 1. Let = e q \stackrel{eq}{=} =eq represent the equivalence relation between two matrices. Then for the partitioned matrix A A A the following conclusions hold.
- When A 11 A_{11} A11 is nonsingular, A = e q [ A 11 0 0 A 22 − A 21 A 11 − 1 A 12 ] = [ A 11 0 0 S c h ( A 11 ) ] A \stackrel{eq}{=}\left[\begin{array}{cc} A_{11} & 0 \\ 0 & A_{22}-A_{21} A_{11}^{-1} A_{12} \end{array}\right]=\left[\begin{array}{cc} A_{11} & 0 \\ 0 & S_{ch}\left(A_{11}\right) \end{array}\right] A=eq[A1100A22−A21A11−1A12]=[A1100Sch(A11)] and hence A A A is nonsingular if and only if S c h ( A 11 ) S_{c h}\left(A_{11}\right) Sch(A11) is nonsingular, and det A = det A 11 det S c h ( A 11 ) \operatorname{det} A=\operatorname{det} A_{11}\operatorname{det} S_{c h}\left(A_{11}\right) detA=detA11detSch(A11)
- When A 22 A_{22} A22 is nonsingular, A = e q [ A 11 − A 12 A 22 − 1 A 21 0 0 A 22 ] = [ S c h ( A 22 ) 0 0 A 22 ] A \stackrel{eq}{=}\left[\begin{array}{cc} A_{11}-A_{12} A_{22}^{-1} A_{21} & 0 \\ 0 & A_{22} \end{array}\right]=\left[\begin{array}{cc} S_{c h}\left(A_{22}\right) & 0 \\ 0 & A_{22} \end{array}\right] A=eq[A11−A12A22−1A2100A22]=[Sch(A22)00A22] hence A A A is nonsingular if and only if S c h ( A 22 ) S_{c h}\left(A_{22}\right) Sch(A22) is nonsingular, and det A = det A 22 det S c h ( A 22 ) \operatorname{det} A=\operatorname{det} A_{22}\operatorname{det} S_{c h}\left(A_{22}\right) detA=detA22detSch(A22).
note: refer to https://blog.csdn.net/weixin_44382195/article/details/102991813 for the definition of the equivalence relation.
Lemma 2
Lemma 2. Given the matrices A 11 = A 11 T , A 22 = A 22 T A_{11}=A_{11}^{T}, A_{22}=A_{22}^{T} A11=A11T,A22=A22T and A 12 A_{12} A12 with appropriate dimensions. The following LMIs are equivalent:
- [ A 11 A 12 A 12 T A 22 ] ≻ 0 \left[\begin{array}{cc}A_{11} & A_{12} \\ A_{12}^{T} & A_{22}\end{array}\right]\succ0 [A11A12TA12A22]≻0
- A 22 = A 22 T ≻ 0 ; A 11 − A 12 A 22 − 1 A 12 T ≻ 0 A_{22}=A_{22}^{T}\succ0 ; A_{11}-A_{12} A_{22}^{-1} A_{12}^{T}\succ0 A22=A22T≻0;A11−A12A22−1A12T≻0
- A 11 = A 11 T ≻ 0 ; A 22 − A 12 T A 11 − 1 A 12 ≻ 0 A_{11}=A_{11}^{T}\succ0 ; A_{22}-A_{12}^{T} A_{11}^{-1} A_{12}\succ0 A11=A11T≻0;A22−A12TA11−1A12≻0.
Lemma 3
Lemma 3. Given the matrices A = A T , C = C T A=A^{T}, C=C^{T} A=AT,C=CT and B B B with appropriate dimensions. The following LMIs are equivalent:
- [ A 11 A 12 A 12 T A 22 ] ≺ 0 \left[\begin{array}{cc}A_{11} & A_{12} \\ A_{12}^{T} & A_{22}\end{array}\right]\prec0 [A11A12TA12A22]≺0
- A 22 = A 22 T ≺ 0 ; A 11 − A 12 A 22 − 1 A 12 T ≺ 0 A_{22}=A_{22}^{T}\prec0 ; A_{11}-A_{12} A_{22}^{-1} A_{12}^{T}\prec0 A22=A22T≺0;A11−A12A22−1A12T≺0
- A 11 = A 11 T ≺ 0 ; A 22 − A 12 T A 11 − 1 A 12 ≺ 0 A_{11}=A_{11}^{T}\prec0 ; A_{22}-A_{12}^{T} A_{11}^{-1} A_{12}\prec0 A11=A11T≺0;A22−A12TA11−1A12≺0.
具体分析过程是利用一次初等行变换和一次初等列变换(实际不需要列变换也行),然后得到块对角矩阵。正(负)定矩阵的对角块均为正(负)定矩阵。既可得到上述关系。