polar decomposition


mathematics Mathematics (from Greek: ) includes the study of such topics as numbers ( and ), formulas and related structures (), shapes and spaces in which they are contained (), and quantities and their changes ( and ). There is no general consensus abo ...
, the polar decomposition of a square
real Real may refer to: * Reality Reality is the sum or aggregate of all that is real or existent within a system, as opposed to that which is only Object of the mind, imaginary. The term is also used to refer to the ontological status of things, ind ...
complex The UCL Faculty of Mathematical and Physical Sciences is one of the 11 constituent faculties of University College London , mottoeng = Let all come who by merit deserve the most reward , established = , type = Public university, Public rese ...

matrix Matrix or MATRIX may refer to: Science and mathematics * Matrix (mathematics), a rectangular array of numbers, symbols, or expressions * Matrix (logic), part of a formula in prenex normal form * Matrix (biology), the material in between a eukaryoti ...
A is a
factorization In mathematics Mathematics (from Ancient Greek, Greek: ) includes the study of such topics as quantity (number theory), mathematical structure, structure (algebra), space (geometry), and calculus, change (mathematical analysis, analysis). I ...
of the form A = U P, where U is a
unitary matrix In linear algebra, a Complex number, complex Matrix (mathematics), square matrix is unitary if its conjugate transpose is also its Invertible matrix, inverse, that is, if U^* U = UU^* = UU^ = I, where is the identity matrix. In physics, espec ...
and P is a positive semi-definite Hermitian matrix, both square and of the same size. Intuitively, if a real n\times n matrix A is interpreted as a linear transformation of n-dimensional Cartesian space, space \mathbb^n, the polar decomposition separates it into a rotation (geometry), rotation or reflection (geometry), reflection U of \mathbb^n, and a scaling (geometry), scaling of the space along a set of n orthogonal axes. The polar decomposition of a square matrix A always exists. If A is invertible matrix, invertible, the decomposition is unique, and the factor P will be positive-definite matrix, positive-definite. In that case, A can be written uniquely in the form A = U e^X , where U is unitary and X is the unique self-adjoint logarithm of a matrix, logarithm of the matrix P. This decomposition is useful in computing the fundamental group of (matrix) Lie groups. The polar decomposition can also be defined as A = P U where P is symmetric positive-definite but is in general a different matrix, while U is the same matrix as above. The polar decomposition of a matrix can be seen as the matrix analog of the complex number#Polar form, polar form of a complex number z as z = u r, where r is its absolute value#Complex numbers, absolute value (a non-negative real number), and u is a complex number with unit norm (an element of the circle group). The definition A = UP may be extended to rectangular matrices A\in\mathbb^ by requiring U\in\mathbb^ to be a Semi-orthogonal matrix, semi-unitary matrix and P\in\mathbb^ to be a positive-semidefinite Hermitian matrix. The decomposition always exists and P is always unique. The matrix U is unique if and only if A has full rank.

Intuitive interpretation

A real square m\times m matrix A can be interpreted as the linear transformation of \mathbb^m that takes a column vector x to A x. Then, in the polar decomposition A = RP, the factor R is an m\times m real orthonormal matrix. The polar decomposition then can be seen as expressing the linear transformation defined by A into a scaling (geometry), scaling of the space \mathbb^m along each eigenvector e_i of A by a scale factor \sigma_i (the action of P), followed by a single rotation or reflection of \mathbb^m (the action of R). Alternatively, the decomposition A=P R expresses the transformation defined by A as a rotation (R) followed by a scaling (P) along certain orthogonal directions. The scale factors are the same, but the directions are different.


The polar decomposition of the complex conjugate of A is given by \overline = \overline\overline. Note that :\det A = \det U \det P = e^\cdot r gives the corresponding polar decomposition of the determinant of ''A'', since \det U = e^ and \det P = r = , \det A, . In particular, if A has determinant 1 then both U and P have determinant 1. The positive-semidefinite matrix ''P'' is always unique, even if ''A'' is Singular matrices, singular, and is denoted as :P = \left(A^* A\right)^\frac, where A^* denotes the conjugate transpose of A. The uniqueness of ''P'' ensures that this expression is well-defined. The uniqueness is guaranteed by the fact that A^* A is a positive-semidefinite Hermitian matrix and, therefore, has a unique positive-semidefinite Hermitian square root of a matrix, square root. If ''A'' is invertible, then ''P'' is positive-definite, thus also invertible and the matrix ''U'' is uniquely determined by :U = AP^.

Relation to the SVD

In terms of the singular value decomposition (SVD) of A, A = W\Sigma V^*, one has :\begin P &= V\Sigma V^* \\ U &= WV^* \end where U, V, and W are unitary matrices (called orthogonal matrices if the field is the reals \mathbb). This confirms that P is positive-definite and U is unitary. Thus, the existence of the SVD is equivalent to the existence of polar decomposition. One can also decompose A in the form :A = P'U Here U is the same as before and P' is given by :P' = UPU^ = \left(AA^*\right)^\frac = W \Sigma W^*. This is known as the left polar decomposition, whereas the previous decomposition is known as the right polar decomposition. Left polar decomposition is also known as reverse polar decomposition. The polar decomposition of a square invertible real matrix A is of the form :A = , A, R where , A, = \left(AA^\textsf\right)^\frac is a positive-semidefinite matrix, positive-definite Hermitian matrix, matrix and R = , A, ^A is an orthogonal matrix.

Construction and proofs of existence

The core idea behind the construction of the polar decomposition is similar to that used to compute the singular-value decomposition. For any A, the matrix A^* A is Hermitian and positive semi-definite, and therefore unitarily equivalent to a positive semi-definite ''diagonal'' matrix. Let then V be the unitary such that A^* A = VDV^*, with D diagonal and positive semi-definite.

Derivation for normal matrices

If A is Normal matrix, normal, then it is unitarily equivalent to a diagonal matrix: A = V\Lambda V^* for some unitary matrix V and some diagonal matrix \Lambda. This makes the derivation of its polar decomposition particularly straightforward, as we can then write :A = V\Phi_\Lambda , \Lambda, V^* = \underbrace_ \underbrace_, where \Phi_\Lambda is a diagonal matrix containing the ''phases'' of the elements of \Lambda, that is, (\Phi_\Lambda)_\equiv \Lambda_/ , \Lambda_, when \Lambda_\neq 0, and (\Phi_\Lambda)_=0 when \Lambda_=0. The polar decomposition is thus A=UP, with U and P diagonal in the eigenbasis of A and having eigenvalues equal to the phases and absolute values of those of A, respectively.

Derivation for invertible matrices

From the singular-value decomposition, it can be shown that a A is invertible if and only if A^* A (equivalently, AA^*) is. Moreover, this is true if and only if the eigenvalues of A^* A are all not zero.Note how this implies, by the positivity of A^* A, that the eigenvalues are all real and strictly positive. In this case, the polar decomposition is directly obtained by writing :A = A\left(A^* A\right)^\left(A^* A\right)^\frac, and observing that A\left(A^* A\right)^ is unitary. To see this, we can exploit the spectral decomposition of A^* A to write A\left(A^* A\right)^ = AVD^V^*. In this expression, V^* is unitary because V is. To show that also AVD^ is unitary, we can use the singular-value decomposition, SVD to write A = WD^\fracV^*, so that :AV D^ = WD^\fracV^* VD^ = W, where again W is unitary by construction. Yet another way to directly show the unitarity of A\left(A^* A\right)^ is to note that, writing the singular-value decomposition, SVD of A in terms of rank-1 matrices as A = \sum_k s_k v_k w_k^*, where s_kare the singular values of A, we have :A\left(A^* A\right)^ = \left(\sum_j \lambda_j v_j w_j^*\right)\left(\sum_k , \lambda_k, ^ w_k w_k^*\right) = \sum_k \frac v_k w_k^*, which directly implies the unitarity of A\left(A^* A\right)^ because a matrix is unitary if and only if its singular values have unitary absolute value. Note how, from the above construction, it follows that ''the unitary matrix in the polar decomposition of an invertible matrix is uniquely defined''.

General derivation

The SVD of A reads A = W D^\frac V^*, with W, V unitary matrices, and D a diagonal, positive semi-definite matrix. By simply inserting an additional pair of Ws or Vs, we obtain the two forms of the polar decomposition of A: : A = WD^\fracV^* = \underbrace_P \underbrace_U = \underbrace_U \underbrace_.

Bounded operators on Hilbert space

The polar decomposition of any bounded linear operator ''A'' between complex Hilbert spaces is a canonical factorization as the product of a partial isometry and a non-negative operator. The polar decomposition for matrices generalizes as follows: if ''A'' is a bounded linear operator then there is a unique factorization of ''A'' as a product ''A'' = ''UP'' where ''U'' is a partial isometry, ''P'' is a non-negative self-adjoint operator and the initial space of ''U'' is the closure of the range of ''P''. The operator ''U'' must be weakened to a partial isometry, rather than unitary, because of the following issues. If ''A'' is the shift operator, one-sided shift on ''l''2(N), then , ''A'', = ½ = ''I''. So if ''A'' = ''U'' , ''A'', , ''U'' must be ''A'', which is not unitary. The existence of a polar decomposition is a consequence of Douglas' lemma: :Lemma If ''A'', ''B'' are bounded operators on a Hilbert space ''H'', and ''AA'' ≤ ''BB'', then there exists a contraction ''C'' such that ''A = CB''. Furthermore, ''C'' is unique if ''Ker''(''B'') ⊂ ''Ker''(''C''). The operator ''C'' can be defined by ''C(Bh)'' := ''Ah'' for all ''h'' in ''H'', extended by continuity to the closure of ''Ran''(''B''), and by zero on the orthogonal complement to all of ''H''. The lemma then follows since ''AA'' ≤ ''BB'' implies ''Ker''(''B'') ⊂ ''Ker''(''A''). In particular. If ''AA'' = ''BB'', then ''C'' is a partial isometry, which is unique if ''Ker''(''B'') ⊂ ''Ker''(''C''). In general, for any bounded operator ''A'', :A^*A = \left(A^*A\right)^\frac \left(A^*A\right)^\frac, where (''AA'')½ is the unique positive square root of ''AA'' given by the usual functional calculus. So by the lemma, we have :A = U\left(A^*A\right)^\frac for some partial isometry ''U'', which is unique if ''Ker''(''A'') ⊂ ''Ker''(''U''). Take ''P'' to be (''AA'')½ and one obtains the polar decomposition ''A'' = ''UP''. Notice that an analogous argument can be used to show ''A = P'U'', where ''P' '' is positive and ''U'' a partial isometry. When ''H'' is finite-dimensional, ''U'' can be extended to a unitary operator; this is not true in general (see example above). Alternatively, the polar decomposition can be shown using the operator version of singular value decomposition#Bounded operators on Hilbert spaces, singular value decomposition. By property of the continuous functional calculus, '', A, '' is in the C*-algebra generated by ''A''. A similar but weaker statement holds for the partial isometry: ''U'' is in the von Neumann algebra generated by ''A''. If ''A'' is invertible, the polar part ''U'' will be in the C*-algebra as well.

Unbounded operators

If ''A'' is a closed, densely defined unbounded operator between complex Hilbert spaces then it still has a (unique) polar decomposition :A = U , A, \, where , ''A'', is a (possibly unbounded) non-negative self adjoint operator with the same domain as ''A'', and ''U'' is a partial isometry vanishing on the orthogonal complement of the range ''Ran''(, ''A'', ). The proof uses the same lemma as above, which goes through for unbounded operators in general. If ''Dom''(''AA'') = ''Dom''(''BB'') and ''AAh'' = ''BBh'' for all ''h'' ∈ ''Dom''(''AA''), then there exists a partial isometry ''U'' such that ''A'' = ''UB''. ''U'' is unique if ''Ran''(''B'') ⊂ ''Ker''(''U''). The operator ''A'' being closed and densely defined ensures that the operator ''AA'' is self-adjoint (with dense domain) and therefore allows one to define (''AA'')½. Applying the lemma gives polar decomposition. If an unbounded operator ''A'' is affiliated operator, affiliated to a von Neumann algebra M, and ''A'' = ''UP'' is its polar decomposition, then ''U'' is in M and so is the spectral projection of ''P'', 1''B''(''P''), for any Borel set ''B'' in [0, ∞).

Quaternion polar decomposition

The polar decomposition of quaternions H depends on the unit 2-dimensional sphere \lbrace x i + y j + z k \in H : x^2 + y^2 +z^2 = 1 \rbrace of quaternion#Square roots of −1, square roots of minus one. Given any ''r'' on this sphere, and an angle −π < ''a'' ≤ π, the versor e^ = \cos (a) + r\ \sin (a) is on the unit 3-sphere of H. For ''a'' = 0 and ''a'' = π, the versor is 1 or −1 regardless of which ''r'' is selected. The norm (mathematics), norm ''t'' of a quaternion ''q'' is the Euclidean distance from the origin to ''q''. When a quaternion is not just a real number, then there is a ''unique'' polar decomposition q = t e^.

Alternative planar decompositions

In the Cartesian plane, alternative planar ring (mathematics), ring decompositions arise as follows:

Numerical determination of the matrix polar decomposition

To compute an approximation of the polar decomposition ''A'' = ''UP'', usually the unitary factor ''U'' is approximated. The iteration is based on Heron's method for the square root of ''1'' and computes, starting from U_0 = A, the sequence :U_ = \frac\left(U_k + \left(U_k^*\right )^\right),\qquad k = 0, 1, 2, \ldots The combination of inversion and Hermite conjugation is chosen so that in the singular value decomposition, the unitary factors remain the same and the iteration reduces to Heron's method on the singular values. This basic iteration may be refined to speed up the process:

See also

*Cartan decomposition *Matrix decomposition#Algebraic polar decomposition, Algebraic polar decomposition *Complex measure#Variation of a complex measure and polar decomposition, Polar decomposition of a complex measure *Lie group decomposition


* John B. Conway, Conway, J.B.: A Course in Functional Analysis. Graduate Texts in Mathematics. New York: Springer 1990 * Ronald G. Douglas, Douglas, R.G.: On Majorization, Factorization, and Range Inclusion of Operators on Hilbert Space. Proceedings of the American Mathematical Society, Proc. Amer. Math. Soc. 17, 413-415 (1966) * . * {{SpectralTheory Lie groups Operator theory Matrix theory Matrix decompositions