In this section we will discuss a way of defining a “square root” of a matrix. Recall that a square root of a number (or more generally, we could take any field) is another number such that . As you know, if then its square roots are only real when , and even then they are not unique. Nevertheless we have the following theorem.
If is a non-negative number (which means ), then has a unique non-negative square root.
In this section, we will generalize the above theorem to matrices, where we replace “non-negative number” with “postive semi-definite matrix”. There are several competing ways to generalize the concept of a “square root” to matrices, but in this module we will only focus on the following one.
Given a matrix , the matrix square root of is a matrix such that
The following exercise shows that matrix square roots don’t always exist:
Prove that there is no matrix such that .
[End of Exercise]
Below we will see the following analogy: Postive real numbers are to positive definite matrices, as non-negative real numbers are to positive semi-definite matrices. A matrix is positive semi-definite if:
for any non-zero vector .
So positive definite matrices are also positive semi-definite. This concept occurs naturally in probability and statistics; for example, the covariance matrix of random variables is always positive semi-definite (see MATH230).
Let be real symmetric. The following are equivalent:
is positive semi-definite,
All of the eigenvalues of are non-negative (i.e. ).
In the above theorem Sylvester’s criterion does not appear because it is no longer valid; in other words, being real, symmetric and positive semi-definite is not equivalent to being real, symmetric and having all principal minors . The only reliable test is the eigenvalue test.
The proof is similar to the proof of Theorem 5.15. ∎
Verify that the matrix is symmetric and positive semi-definite, but not positive definite.
[End of Exercise]
Let be a real symmetric positive semi-definite matrix. Then there exists a unique real symmetric positive semi-definite matrix such that
In this case, the resulting matrix is usually called “the” matrix square root of , since it’s uniquely defined. So, in this way, “real symmetric positive semi-definite matrices” may be considered as a nice generalization of “non-negative real numbers”.
There is an orthogonal matrix and diagonal matrix such that
This is the Spectral Theorem 5.7. Since is positive semi-definite, all of the diagonal entries of are non-negative (i.e. ), so we can define as follows
,
Then , and is real symmetric positive semi-definite. Finally,
Therefore, we have proved that such a always exists.
We omit the proof of uniqueness (the proof is not obvious). ∎
Find the matrix square root of from Example 5.9.
In that example we found an orthogonal and diagonal such that . By taking the square root of the diagonal entries of , we compute:
Now it is easy to check that .