The user wants to find the polar decomposition of a matrix.
I need to explain what it is, why it's useful (briefly), and how to calculate it.
1. Definition:
What is the polar decomposition?
Analogous to z=reiθ for complex numbers (r≥0).
For a square matrix A, it is A=UP, where U is unitary (orthogonal) and P is positive semi-definite Hermitian (symmetric).
If A is invertible, U is unique. If A is rectangular (m×n), it's A=UP where U is m×n with orthonormal columns (partial isometry) and P is n×n positive semi-definite. Or A=SV where S is m×m and V is m×n.
Usually, people look for the "right" polar decomposition (A=UP) or the "left" polar decomposition (A=PU′).
Let's stick to the most common request: Right Polar Decomposition A=UP.
2. Algorithm/Method:
How to compute P and U?
Key relationship: A∗A=P∗U∗UP=P2 (since U∗U=I).
So, P=A∗A. This is the unique positive semi-definite square root.
Once P is found, U=AP−1.
Wait, what if A is singular? P−1 doesn't exist.
If singular, U is found by extending the orthonormal basis of the range of P to the null space of P? Or simply, solving the system AP+=U? Or using SVD.
Actually, SVD is the most robust way to explain and compute it.
Let A=WΣV∗ be the SVD.
P=VΣV∗.
U=WV∗.
Check: UP=(WV∗)(VΣV∗)=W(V∗V)ΣV∗=WΣV∗=A. Correct.
Another method: Iterative methods (like Newton's method), but for a general explanation, the algebraic construction via A∗A or SVD is best.
3. Structure of the Answer:
Definition: State what it is clearly (A=UP or A=PU).
Conditions: Mention that U is unitary/orthogonal and P is positive semi-definite Hermitian/symmetric.
How to Calculate (Step-by-Step):
Step 1: Compute M=A∗A (or ATA for real matrices).
Step 2: Compute P=M. This involves diagonalizing M (M=QDQ−1) and taking the square root of eigenvalues (P=QDQ−1). Note that eigenvectors of ATA are the right singular vectors.
Step 3: Compute U. If invertible, U=AP−1. If not, mention singular value decomposition (SVD).
Example: Create a simple 2×2 matrix and walk through it.
Matrix A=(1011). This is simple enough.
ATA=(1101)(1011)=(1112).
Eigenvalues of M: tr=3,det=1. λ2−3λ+1=0. λ=23±5. These are messy.
Let's try a matrix with integer eigenvalues for ATA.
Let A=(3403).
ATA=(3043)(3403)=(2512129).
Eigenvalues: tr=34,det=225−144=81. (λ−9)(λ−25). This is nice.
Eigenvectors:
λ=9: 16x+12y=0⇒4x+3y=0. Vector v1=(3−4). Norm is 5. Normalized: (3/5−4/5).
Q=(3/5−4/510)? No, order usually λ1<λ2. V=(3/5−4/501). Wait, orthogonal matrix requires columns to be orthonormal. (3/5−4/5) and (10) are orthogonal? 3/5∗1+(−4/5)∗0=0. Yes.
So V=(3/5−4/54/53/5) (Rotation matrix style). Wait, the second eigenvector was (1,0), which corresponds to (4/5,3/5) in the orthonormal basis?
Let's recheck eigenvectors for ATA=(2512129).
For λ=9: (1612120)→(4030)→y=t,x=−3/4t. Normalized: 51(−34).
For λ=25: (01212−16)→y=0,x=t. Normalized: (10).
Dot product is 0. Good.
V=(10−3/54/5). (Ordering by eigenvalue magnitude usually standard, but doesn't matter for P).
This seems messy to present as a "simple" example. Let's try to find a matrix A where ATA is diagonal to begin with.
Let A=(221−1).
ATA=(212−1)(221−1)=(8002).
This is perfect!
P=ATA=(8002)=(22002).
U=AP−1=(221−1)(2210021)=21(111−1).
Check U: Is it orthogonal? UTU=21(111−1)(111−1)=21(2002)=I. Yes.
This is a very clean example.
4. Left vs Right Decomposition:
Mention both briefly.
Right: A=UP (P represents the "stretch").
Left: A=P′U (P′ represents the output stretch).
Usually P′=UPU∗.
5. Summary of Content:
Concept: Factorization of a matrix into a unitary matrix and a positive semi-definite Hermitian matrix.
Formula:A=UP (Right Polar Decomposition).
Steps:
Calculate P=(A∗A)1/2.
Diagonalize A∗A=VDV∗.
P=VD1/2V∗.
Calculate U.
If A is invertible: U=AP−1.
General (SVD approach): U is constructed from eigenvectors of AA∗ and A∗A.
Example: Use the 22 example.
Refinement:
Should I mention the SVD method as the primary calculation method?
The SVD (A=WΣV∗) gives P=VΣV∗ and U=WV∗.
This is conceptually simpler than computing (ATA)1/2 manually via diagonalization, but practically, if you are doing it by hand, computing ATA and diagonalizing is the standard linear algebra exercise approach.
I will present the manual calculation method (diagonalization of A∗A) as the primary method, as that's how students are usually tested, but mention SVD as the "modern/computational" way.
Formatting:
Use LaTeX for math. Use bold for vectors. Clear headings.
Let's double check the calculation for the example.
A=(221−1).
ATA=(212−1)(221−1)=(4+42−22−21+1)=(8002).
P=(8002).
U=AP−1=(221−1)(1/8001/2).
Col 1: (2/82/8)=(1/21/2).
Col 2: (1/2−1/2).
So U=(1/21/21/2−1/2).
Check UTU:
Row 1 dot Row 1: 1/2+1/2=1.
Row 1 dot Row 2: 1/2−1/2=0.
Row 2 dot Row 2: 1/2+1/2=1.
Correct.
Result looks solid.
Final check on "Right" vs "Left".
Right: A=UP. The symmetric part P acts on the vector first, then the rotation U.
Left: A=P′U. The rotation acts first.
Most standard texts define "the" polar decomposition as the Right one, but mention the Left one as a variant. I'll define Right as the standard.
One more thing: What if the matrix is rectangular?
The question asks for "a matrix", not specifying square.
However, the most standard context is square matrices. I should add a small note about rectangular matrices (SVD is best there).
If A is m×n (m≥n), A=Um×nPn×n.
P is square n×n. U is "tall" with orthonormal columns.
If m<n, usually one writes A=P′U (Left) where P′ is m×m.
Or sticking to A=UP, P is n×n (singular) and U is m×n.
I will focus on the square case for clarity, but mention U contains orthonormal columns for the general case.
Plan:
Definition.
The Method (3 steps).
Worked Example (The integer one).
Connection to SVD (for context/computational ease).
Let's write it out.
The polar decomposition (specifically the right polar decomposition) is a factorization of a matrix A into the product of a unitary matrix U and a positive semi-definite Hermitian matrix P.
A=UP
This is analogous to the polar form of a complex number z=reiθ, where r corresponds to P (magnitude) and eiθ corresponds to U (rotation).
Here is the step-by-step method to find the polar decomposition of a square matrix A.
The Algorithm
Given a square matrix A, follow these steps:
Step 1: Calculate the Matrix P
The matrix P represents the "stretch" component. It is defined as the unique positive semi-definite square root of the matrix A∗A (where A∗ is the conjugate transpose of A).
Compute the product M=ATA (if A is real) or M=A∗A (if A is complex).
Diagonalize M to find its square root P=M.
Find the eigenvalues λi and orthonormal eigenvectors vi of M.
Construct the diagonal matrix D=diag(λ1,λ2,…) and the orthogonal matrix V whose columns are the eigenvectors.
P is given by P=VDVT, where D is the diagonal matrix of the square roots of the eigenvalues.
Step 2: Calculate the Matrix U
The matrix U represents the "rotation" (or reflection) component.
If A is invertible:
You can simply solve for U using the inverse of P:
U=AP−1
If A is singular (not invertible):P is not invertible. You must find U by mapping the range of P to the range of A. The easiest computational method in this case is using the Singular Value Decomposition (SVD) (see below).
Method 2: Using Singular Value Decomposition (SVD)
For a computer or calculator, this is often the most robust method.
If you have the SVD of A=WΣV∗:
P is computed as VΣV∗.
U is computed as WV∗.
Worked Example
Let's find the polar decomposition of the real matrix:
A=(221−1)