Section II.3.1

Euclidean vector spaces

[8/100%]
Show
LaTeX
Problem II.3.1.1

In the space P3P_{3} of real polynomials f(t)f(t) of degree <3<3, the vectors 1,t1, t are orthogonal with respect to the inner product given by the formula (f∣g)=∫−11f(t)g(t)dt(f \mid g)=\int_{-1}^{1} f(t) g(t) d t (a special case of formula (2)). Find:

?
(a)

the subspace ⟨1,t⟩⊥\langle 1, t\rangle^{\perp };

(b)

an orthonormal basis in P3P_{3}.

Problem II.3.1.2

Let (V,(∗∣∗))(V,(* \mid *)) be a three-dimensional Euclidean vector space such that ∥x∥2=(x∣x)=3x12+2x22+x32−4x1x2−2x1x3+2x2x3\left\| \mathbf{x}\right\|^{2}=(\mathbf{x} \mid \mathbf{x})=3 x_{1}^{2}+2 x_{2}^{2}+x_{3}^{2}-4 x_{1} x_{2}-2 x_{1} x_{3}+2 x_{2} x_{3} (verify that this form is positive definite). Find:

?
(a)

the angle α\alpha between the vectors x=[1,1,1],y=[2,2,1]\mathbf{x}=[1,1,1], \mathbf{y}=[2,2,1];

(b)

all vectors orthogonal to x\mathbf{x}.

Problem II.3.1.3

Using the Gram--Schmidt orthogonalization process, prove that any nonsingular matrix A=(aij)∈Mn(R)A=\left(a_{i j}\right) \in M_{n}(\mathbb {R}) can be written as a product A=BCA=B C of an orthogonal matrix BB and an upper triangular matrix CC with det⁡B=±det⁡A\operatorname {det} B= \pm \operatorname {det} A.

?
Problem II.3.1.4

The set Mn(R)M_{n}(\mathbb {R}), regarded as an n2n^{2}-dimensional Euclidean space with the standard inner product, contains the group O(n)O(n) of orthogonal matrices, defined by n(n+1)/2n(n+1) / 2 relations (15) or (15') (O(n)O(n) being the orthogonal group). Thus O(n)O(n) can be regarded as an "algebraic variety" of dimension n2−n(n+1)/2=n(n−1)/2=dim⁡o(n)n^{2}-n(n+1) / 2=n(n-1) / 2=\operatorname {dim} \mathfrak {o}(n), where o(n)\mathfrak {o}(n) is the vector space of skew-symmetric matrices of order nn. It is natural to expect some good correspondence between the sets O(n)O(n) and o(n)\mathfrak {o}(n). An example of such a correspondence is the Cayley transform

K=(E−A)−1(E+A),A=(E−K)−1(E+K) K=(E-A)^{-1}(E+A), \quad A=(E-K)^{-1}(E+K)

It is required to prove that if A∈O(n),det⁡(E−A)≠0A \in O(n), \operatorname {det}(E-A) \neq 0, then K∈o(n)K \in \mathfrak {o}(n). The converse also holds: to each matrix K∈o(n)K \in \mathfrak {o}(n) there corresponds an A∈O(n)A \in O(n) with 1∉Spec⁡(A)1 \notin \operatorname {Spec}(A). Under this correspondence, one must discard from the "algebraic variety" O(n)O(n) the "hypersurface" defined by the equation det⁡(E−A)=0\operatorname {det}(E-A)=0.

One can propose another Cayley transform

K=(E+A)−1(E−A),A=(E+K)−1(E−K) K=(E+A)^{-1}(E-A), \quad A=(E+K)^{-1}(E-K)

where from O(n)O(n) one must remove the "hypersurface" with equation det⁡(A+E)=0\operatorname {det}(A+E)=0, or, equivalently, the orthogonal matrices with −1∉Spec⁡(A)-1 \notin \operatorname {Spec}(A).

It is required to verify the validity of the correspondence (19).

?
Problem II.3.1.5

Verify that the orthogonal matrices obtained by the Cayley transform from skew-symmetric ones (in relations (18) and (19)) have determinant 1 (the set of all such matrices is denoted by the symbol SO(n)S O(n)).

?
Problem II.3.1.6

Prove that the characteristic polynomial χA(t)\chi_{A}(t) of an orthogonal n×nn \times n matrix AA has the property

tnχA(1/t)=±χA(t) t^{n} \chi _{A}(1 / t)= \pm \chi _{A}(t)
?
Problem II.3.1.7

Let A=[A(1),…,A(n)]A=\left[A_{(1)}, \ldots , A_{(n)}\right] be an arbitrary matrix with pairwise orthogonal rows. Prove that

∣det⁡A∣=∥A(1)∥⋅∥A(2)∥…∥A(n)∥ \left|\operatorname {det} A\right|=\left\| A_{(1)}\right\| \cdot \left\| A_{(2)}\right\| \ldots \left\| A_{(n)}\right\|

(the standard norm of vectors in Rn\mathbb {R}^{n}).

?
Problem II.3.1.8

Let X=[X(1),…,X(n)]X=\left[X_{(1)}, \ldots , X_{(n)}\right] be an arbitrary matrix from Mn(R)M_{n}(\mathbb {R}). Prove that

∣det⁡X∣⩽∥X(1)∥⋅∥X(2)∥…∥X(n)∥ \left|\operatorname {det} X\right| \leqslant \left\| X_{(1)}\right\| \cdot \left\| X_{(2)}\right\| \ldots \left\| X_{(n)}\right\|

(Hadamard's inequality).

?