这一节讲的内容在国内的学校教材中常被列为“(半)正定二次型”,从form的角度理解比从矩阵的角度理解要容易一些。 f f f是positive与由 f f f的矩阵 A A A形成的函数 g ( X , Y ) = Y ∗ A X g(X,Y)=Y^{\ast}AX g(X,Y)=Y∗AX是positive这二者等价,而Theorem 5说明: g ( X , Y ) = Y ∗ A X g(X,Y)=Y^{\ast}AX g(X,Y)=Y∗AX是positive又等价于存在可逆矩阵 P P P使得 A = P ∗ P A=P^{\ast}P A=P∗P。这个定理实际用起来并不容易,为了引入对positive的判别,后面介绍了principle minors的定义,也就是 A A A的前k行和前k列形成的子矩阵的行列式,记为 Δ k ( A ) \Delta_k(A) Δk(A),而通过引理和Theorem 6, f f f是positive的另一个等价条件是: A = A ∗ A=A^{\ast} A=A∗且 A A A的所有principle minor都是正的。
在上述这些内容都讨论过后,再定义了positive matrix,即对任何 X ≠ 0 X\neq 0 X=0都有 X t A X > 0 X^tAX>0 XtAX>0,以及linear operator是positive的定义。这些都可以从form中自然的得到。另外一个有用的结论是:如果 f f f是一个positive form,那么 f f f实际是一个inner product。
本节的最后,给出了一个关于positive operator,positive forms和positive matrices的等价条件的总结。
Exercises
Exercise 1. Let V V V be C 2 C^2 C2, with the standard inner product. For which vectors α ∈ V \alpha\in V α∈V is there a positive linear operator T T T such that α = T ϵ 1 \alpha=T\epsilon_1 α=Tϵ1?
Solution: If T T T is positive, then T = T ∗ T=T^{\ast} T=T∗ and we shall have ( α ∣ ϵ 1 ) > 0 (\alpha|\epsilon_1)>0 (α∣ϵ1)>0, so write α = ( a , b ) \alpha=(a,b) α=(a,b), we know that a > 0 a>0 a>0.
Exercise 2. Let V V V be R 2 R^2 R2, with the standard inner product. If θ \theta θ is a real number, let T T T be the linear operator ‘rotation through θ \theta θ,’
T θ ( x 1 , x 2 ) = ( x 1 cos θ − x 2 sin θ , x 1 sin θ + x 2 cos θ ) T_{\theta}(x_1,x_2)=(x_1\cos\theta-x_2\sin\theta,x_1\sin\theta+x_2\cos\theta) Tθ(x1,x2)=(x1cosθ−x2sinθ,x1sinθ+x2cosθ)
For which values of θ \theta θ is T θ T_{\theta} Tθ a positive operator?
Solution: The matrix of T θ T_{\theta} Tθ with the standard basis is A = [ cos θ − sin θ sin θ cos θ ] A=\begin{bmatrix}\cos\theta&-\sin\theta\\\sin\theta&\cos\theta\end{bmatrix} A=[cosθsinθ−sinθcosθ]. If T T T is positive, then we require A = A ∗ A=A^{\ast} A=A∗ and cos θ > 0 \cos\theta>0 cosθ>0, thus we must have sin θ = 0 , cos θ = 1 \sin\theta=0,\cos\theta=1 sinθ=0,cosθ=1, which gives θ = 2 k π , k ∈ N \theta=2k\pi,k\in N θ=2kπ,k∈N.
Exercise 3. Let V V V be the space of n × 1 n\times 1 n×1 matrices over C C C, with the inner product ( X ∣ Y ) = Y ∗ G X (X|Y)=Y^{\ast}GX (X∣Y)=Y∗GX (where G G G is an n × n n\times n n×n matrix such that this is an inner product). Let A A A be an n × n n\times n n×n matrix and T T T the linear operator T ( X ) = A X T(X)=AX T(X)=AX. Find T ∗ T^{\ast} T∗. If Y Y Y is a fixed element of V V V, find the element Z Z Z of V V V which determines the linear functional X → Y ∗ X X\to Y^{\ast}X X→Y∗X. In other words, find Z Z Z such that Y ∗ X = ( X ∣ Z ) Y^{\ast}X=(X|Z) Y∗X=(X∣Z) for all X ∈ V X\in V X∈V.
Solution: Since ( X ∣ Y ) = Y ∗ G X (X|Y)=Y^{\ast}GX (X∣Y)=Y∗GX is an inner product, we must have G G G positive, thus invertible.
We have ( T ( X ) ∣ Y ) = Y ∗ G A X (T(X)|Y)=Y^{\ast}GAX (T(X)∣Y)=Y∗GAX, so if we let U ( Y ) = ( G A G − 1 ) ∗ Y U(Y)=(GAG^{-1})^{\ast}Y U(Y)=(GAG−1)∗Y, we shall have
( X ∣ U ( Y ) ) = Y ∗ G A G − 1 G X = Y ∗ G A X = ( T ( X ) ∣ Y ) ⟹ U = T ∗ (X|U(Y))=Y^{\ast}GAG^{-1}GX=Y^{\ast}GAX=(T(X)|Y)\implies U=T^{\ast} (X∣U(Y))=Y∗GAG−1GX=Y∗GAX=(T(X)∣Y)⟹U=T∗
If Y Y Y is a fixed element of V V V, then let Z = ( G − 1 ) ∗ Y Z=(G^{-1})^{\ast}Y Z=(G−1)∗Y, we have ( X ∣ Z ) = Y ∗ G − 1 G X = Y ∗ X (X|Z)=Y^{\ast}G^{-1}GX=Y^{\ast}X (X∣Z)=Y∗G−1GX=Y∗X.
Exercise 4. Let V V V be a finite-dimensional inner product space. If T T T and U U U are positive linear operators on V V V, prove that ( T + U ) (T+U) (T+U) is positive. Give an example which shows that T U TU TU need not be positive.
Solution: We directly have ( T + U ) ∗ = T ∗ + U ∗ = T + U (T+U)^{\ast}=T^{\ast}+U^{\ast}=T+U (T+U)∗=T∗+U