Singular Values Calculator (2×2 Matrix)

What Are Singular Values?

Singular values are non-negative scalar descriptors of a matrix that quantify how much a linear transformation stretches or compresses space along its principal axes. For any real 2×2 matrix A, the singular values σ₁ and σ₂ are defined as the square roots of the eigenvalues of the symmetric positive semi-definite matrix A^TA. The Singular Value Decomposition (SVD) — written A = UΣV^T — decomposes any matrix into three interpretable components, making singular values foundational in numerical linear algebra, machine learning, image compression, signal processing, and control theory.

Core Formula and Derivation

For a 2×2 real matrix A with top-left entry a, top-right entry b, bottom-left entry c, and bottom-right entry d, the singular values are computed as the square roots of the eigenvalues of A^TA:

σᵢ = √( λᵢ(A^TA) )

The eigenvalues of A^TA are found in closed form using the trace and determinant of A^TA:

λ₁,₂ = tr(A^TA) / 2 ± √( (tr(A^TA) / 2)² − det(A^TA) )

This quadratic formula approach avoids numerical iteration and yields exact results for any 2×2 real matrix.

Step 1: Form A^TA

The product A^TA is a 2×2 symmetric matrix with three distinct entries: the top-left entry equals a² + c², the bottom-right entry equals b² + d², and both off-diagonal entries equal ab + cd. Symmetry of A^TA guarantees real, non-negative eigenvalues.

Step 2: Compute the Trace and Determinant

The trace of A^TA equals a² + b² + c² + d², which is the squared Frobenius norm ‖A‖²_F — the sum of squares of all matrix entries. The determinant of A^TA equals (a² + c²)(b² + d²) − (ab + cd)², which simplifies to (ad − bc)² = (det A)². This identity shows that σ₁ · σ₂ = |det(A)|, linking singular values directly to the determinant.

Step 3: Extract Singular Values

Substituting into the quadratic formula yields eigenvalues λ₁ ≥ λ₂ ≥ 0. The singular values follow as σ₁ = √λ₁ and σ₂ = √λ₂, always ordered from largest to smallest by convention. If the discriminant (tr/2)² − det equals zero, both singular values are equal and the matrix is called an isotropic scaling.

Variable Reference

• a — Top-left entry A[1,1]: contributes to the squared norm of column 1
• b — Top-right entry A[1,2]: contributes to the squared norm of column 2
• c — Bottom-left entry A[2,1]: contributes to the squared norm of column 1
• d — Bottom-right entry A[2,2]: contributes to the squared norm of column 2
• σ₁ — Largest singular value: maximum stretch factor of the linear map
• σ₂ — Smallest singular value: minimum stretch factor; zero if and only if A is singular

Worked Example

Let A = [[3, 1], [1, 3]]. Computing step by step:

• A^TA = [[3² + 1², 3·1 + 1·3], [3·1 + 1·3, 1² + 3²]] = [[10, 6], [6, 10]]
• tr(A^TA) = 10 + 10 = 20; det(A^TA) = 10·10 − 6·6 = 100 − 36 = 64
• λ₁,₂ = 20/2 ± √( (20/2)² − 64 ) = 10 ± √(100 − 64) = 10 ± 6
• λ₁ = 16, λ₂ = 4; therefore σ₁ = 4 and σ₂ = 2

The unit circle maps to an ellipse with semi-axes of length 4 and 2 under this transformation. The condition number is κ = σ₁/σ₂ = 4/2 = 2, confirming the matrix is well-conditioned. As a check: σ₁ · σ₂ = 8 = |3·3 − 1·1| = 8 ✓, and σ₁² + σ₂² = 16 + 4 = 20 = ‖A‖²_F ✓.

Applications of Singular Values

• Dimensionality reduction: Principal Component Analysis (PCA) uses SVD to identify the directions of greatest variance, reducing high-dimensional data to manageable representations.
• Image compression: Retaining only the top-k singular values and their associated vectors approximates an image matrix with a fraction of the original storage.
• Least-squares solutions: The Moore-Penrose pseudoinverse A⁺ = VΣ⁺U^T uses singular values to solve overdetermined and underdetermined systems with minimum residual.
• Numerical rank detection: Singular values near zero reveal near-linear dependence among rows or columns, exposing the effective numerical rank of a matrix.
• Robust control: The largest singular value bounds the maximum gain of a transfer matrix, a key quantity in H-infinity control design.

Methodology and Sources

The eigenvalue approach implemented in this calculator follows the rigorous derivation in the UC Berkeley Math 54 SVD Notes by Michael Hutchings, which establishes the formal relationship between A^TA eigenvalues and singular values for real matrices of any size. Computational verification follows the step-by-step numerical procedure detailed in the SVD computation example from the University of Minnesota Duluth. Geometric interpretation of singular values as semi-axis lengths of the image ellipsoid draws from the MIT 18.095 Chapter 7 SVD Notes, which provides a thorough treatment of the SVD as a generalization of spectral decomposition.