Python Calculate Square Root of Matrix

What is Matrix Square Root?

The square root of a matrix A is another matrix X such that X² = A. For a matrix to have a square root, it must be positive definite (for real matrices) or Hermitian positive definite (for complex matrices).

Matrix square roots are used in solving differential equations, quantum mechanics, and data analysis. The principal square root is the one where all eigenvalues have non-negative real parts.

Key Properties

• Existence: A matrix has a square root if and only if it is similar to a diagonal matrix with non-negative diagonal elements.
• Uniqueness: The principal square root is unique for positive definite matrices.
• Applications: Used in control theory, signal processing, and machine learning.

Python Implementation

Python's NumPy library provides functions to compute matrix square roots. The most common method is using the scipy.linalg.sqrtm function, which computes the principal matrix square root.

Basic Example

import numpy as np
from scipy.linalg import sqrtm

# Define a positive definite matrix
A = np.array([[4, 1], [1, 3]])

# Compute the square root
X = sqrtm(A)

print("Matrix A:")
print(A)
print("\nSquare root of A:")
print(X)

Verification

To verify the result, you can multiply the square root by itself and check if it equals the original matrix:

# Verify X² = A
X_squared = np.dot(X, X)
print("\nX²:")
print(X_squared)
print("\nIs X² equal to A?")
print(np.allclose(X_squared, A))

Handling Non-Positive Definite Matrices

For matrices that are not positive definite, you may need to use eigenvalue decomposition or other numerical methods. The following example demonstrates this approach:

def matrix_sqrt_non_positive_definite(A):
    # Compute eigenvalues and eigenvectors
    eigenvalues, eigenvectors = np.linalg.eig(A)

    # Take square root of eigenvalues (handle negative values)
    sqrt_eigenvalues = np.sqrt(np.abs(eigenvalues))

    # Reconstruct the matrix
    sqrt_A = eigenvectors @ np.diag(sqrt_eigenvalues) @ np.linalg.inv(eigenvectors)

    return sqrt_A

Examples

Here are some practical examples of calculating matrix square roots in Python:

Example 1: Simple 2x2 Matrix

import numpy as np
from scipy.linalg import sqrtm

A = np.array([[9, 0], [0, 16]])
X = sqrtm(A)

print("Square root of A:")
print(X)

The output will be:

[[3. 0.]
 [0. 4.]]

Example 2: Non-Diagonal Matrix

A = np.array([[4, -1], [-1, 4]])
X = sqrtm(A)

print("Square root of A:")
print(X)

The output will be a matrix where X² ≈ A.

Example 3: Using Custom Function

A = np.array([[1, 2], [2, 1]])
X = matrix_sqrt_non_positive_definite(A)

print("Square root of A:")
print(X)

This example uses the custom function for non-positive definite matrices.