Linear algebra (Osnabrück 2024-2025)/Part II/Lecture 34



Diagonalizability of isometries over the complex numbers


Let be a linear isometry on a finite-dimensional -vector space , endowed with an inner product. Let denote an invariant linear subspace. Then also the orthogonal complement is invariant. In particular, can be written as a direct sum

where the restrictions and

are also isometries.

We have

For such a and an arbitrary , we have

since due to the invariance of . Therefore, .


The following statement is called spectral theorem or, more specific, spectral theorem for complex isometries. In this course, we will get to know further spectral theorems; see Fact ***** and Fact *****.


Let be a finite-dimensional -vector space, endowed with an inner product, and let

be an isometry. Then has an orthonormal basis of eigenvectors of . In particular, is

diagonalizable.

We do induction on the dimension of . The statement is clear in the one-dimensional case. Due to the Fundamental theorem of algebra and Theorem 23.2 , has an eigenvalue and an eigenvector, which we can normalize. Let be the corresponding eigenline. Since we have an isometry, the orthogonal complement is also, due to Fact *****, -invariant, and the restriction

is again an isometry. By the induction hypothesis, there exists on an orthonormal basis consisting of eigenvectors. Together with the first eigenvector, these form an orthonormal basis of eigenvectors of .



Angles


For vectors and , different from , in a euklidean vector space , the inequality of Cauchy-Schwarz implies that

holds. Using the trigonometric function cosine (as a bijective mapping ) and its inverse function, the angle between the two vectors can be defined, by setting

The angle is a real number between and . The equation above can be read as

This provides the possibility to define the inner product in this way. However, then we have to find an independent definition for the angle. This approach might look a bit more intuitive but is has computationally and in terms of the proofs many disadvantages.

For an affine space over a Euclidean vector space , and three given points (a triangle) with , the angle of the triangle at is the angle .



Plane isometries


Let

be a proper linear isometry. Then is a rotation, and the describing matrix with respect to the standard basis has the form

with a uniquely determined rotation angle

.

Let and be the images of the standard vectors and . An isometry preserves the length of a vector; therefore,

Hence, is a real number between and , and , that is, is a point on the real unit circle. The unit circle is parametrized by the trigonometric functions, that is, there exists a uniquely determined angle , , such that

Since an isometry preserves orthogonality, we have

In case , this implies (because of ) . Then , and, due to the properness assumption, the sign must be that of .

So let . Then

Since both vectors have the length , the modulus of the scalar factor is . In case , we had and the determinant were . Therefore, and , yielding the claim.


The composition of two rotations and is . This property is intuitively clear, using intuitive properties of plane rotations. Using the addition theorems for the trigonometric functions, we can proof them. Moreover, the addition theorems follow from this property, see exercise *****. Therefore, the group of plane rotations is commutative.


Let

be an improper linear isometry. Then is an axis reflection, and its describing matrix, with respect to the standard basis, has the form

with a uniquely determined angle

.

We consider

which is by the multiplication theorem for the determinant a proper isometry. Because of Fact *****, there exists a uniquely determined angle such that

Therefore,


If such an axis rotation is given, then is an eigenvector for the eigenvalue , and the axis of reflection is , see exercise *****. An axis reflection is described, with respect to the basis consisting of a vector of the reflecting axis and a corresponding orthogonal vector , by the matrix . This means that the description given in Fact ***** (with respect to the standard basis) can be improved drastically.



Isometries of space


Let

be a linear isometry. Then there exists an eigenvector with eigenvalue

or .

The characteristic polynomial of is a normed polynomial of degree three. For , we have , and for , we have . Due to the intermediate value theorem, has a real zero. Such a zero is an eigenvalue of . Because of Fact *****, this eigenvalue equals or .


A proper linear isometry of the space transforms the unit sphere into itself. One might imagine an isometry as a rotation of a ball lying in a suitable bowl.


A proper isometry

has an

eigenvector with eigenvalue , that is, there exists a line (through the origin)

that is a fixed line for .

We consider the characteristic polynomial of , that is,

This is a normed real polynomial of degree three. For , we have

As the polynomial as , there must be a positive that is a zero of . Due to Fact *****, we must have .



Let

be a proper isometry. Then is a rotation around a fixed axis. This means that is described, with respect to a suitable orthonormal basis, by a matrix of the form

Due to Fact *****, there exists an eigenvector for the eigenvalue . Let denote the corresponding line. This line is fixed and, in particular, invariant under . Because of Fact *****, also the orthogonal complement is invariant under . That is, there exists a linear isometry

that coincides on with . Here, is proper. Therefore, by Fact *****, is a plane rotation. If we choose a vector of length one in , and an orthonormal basis of , then has with respect to this basis the given matrix form.


At the beginning of a football game, the football lies on the center spot. After a goal is scored, the ball is again put on the center spot. In this situation, the following holds: There exists at least two (antipodal) points on the ball (its surface)

that are at the new kick-off exactly at the position where they were at the first kick-off. The total movement of the ball is a rotation around an axis.

The total movement is a linear isometry; therefore, the statement follows from Fact *****.




The decomposition theorem for isometries


Let be a real finite-dimensional vector space, and let

denote an endomorphism. Then there exists a -invariant linear subspace of dimension

or .

We may assume , and that is described by the matrix with respect to the standard basis. If has an eigenvalue, then we are done. Otherwise, we consider the corresponding complex mapping, that is,

which is given by the same matrix . This matrix has a complex eigenvalue , and a complex eigenvector . In particular, we have

Writing

with , this means

Comparing the real part and the imaginary part we can deduce that . Therefore, the real linear subspace is invariant.



Let

be an isometry on the euclidean vector space . Then is a orthogonal direct sum

of -invariant linear subspaces,

where the are one-dimensional, and the are two-dimensional. The restriction of to the is the identity, the restriction to is the negative identity, and the restriction to is a rotation without eigenvalue.

We do induction over the dimension of . The one-dimensional case is clear, due to Fact *****. Let . The determinant has, because of Fact ***** either the value or . In case it is , the characteristic polynomial has two distinct zeroes, and these zeroes are, due to Fact *****, and . Then we have a reflection at an axis, and

If the determinant is , then we are in the situation of Fact *****, and we have a rotation. If the angle of rotation is , then we have the identity, and we can decompose . If the angle of rotation is , then we have a point reflection ; therefore, we have a decomposition . For all other angles, there is no eigenvector.

Let now be arbitrary, and suppose that the statement is proven for smaller dimensions. Because of Fact *****, there exists a -invariant linear subspace of dimension or , and, because of Fact *****, there exists an invariant orthogonal complement, that is,

The induction hypothesis, applied to , yields the result.


In this decomposition, is the eigenspace for the eigenvalue , and is the eigenspace for the eigenvalue ; the decompositions are not unique. The Isometry is proper if and only if is even.