When is the determinant of a upper triangular not the product of its diagonals?

I'm trying to dive deeper into some topics for class, I was wondering if its possible to have an upper triangular matrix where its determinant is not equal to the product of its diagonals

  • 2
    $\begingroup$ By "diagonals", you mean the elements on the diagonal, right? Then this is always true, argue by induction, expanding the determinant by the first column. $\endgroup$ – астон вілла олоф мэллбэрг Feb 28 at 16:02
  • $\begingroup$ In infinite-order matrices the quantities may not exist. When they do, a straightforward application of the "permutation based" definition renders the two quantities equal. $\endgroup$ – Oscar Lanzi Feb 28 at 16:10

Alternatively, we can also prove that for an upper triangular matrix $A$, we have $$ \det(A) = a_{11}a_{22}\dots a_{nn} $$ using just the definition of determinant.

Let $S_n$ be the set of all permutations of $\{1,2,\dots,n\}$. Recall that $$ \det(A) = \sum_{\sigma\in S_n} \text{sgn}(\sigma) a_{1\sigma(1)}a_{2\sigma(2)}\dots a_{n\sigma(n)}. $$ For any $\sigma\in S_n$ that is not the identity (i.e. $\sigma(i)\ne i$ for some $i\le n$), we let $j\in \{1,2,\dots,n\}$ be the largest number such that $\sigma(j)\ne j$. Since $j$ is largest, we must have $\sigma(j)< j$. This means that $a_{j\sigma(j)}=0$ since $A$ is upper triangular.

The above argument shows that $a_{1\sigma(1)}a_{2\sigma(2)}\dots a_{n\sigma(n)}=0$ for all other $\sigma\in S_n$ except the one that $\sigma(i)=i$ for all $i$. Since the sign of the identity permutation is $1$, this concludes what we want to prove.


No, it is not possible.

Assuming that $A$ is an $n\times n$ upper (or lower) triangular matrix whose diagonal entries are $d_1,\dots, d_n$.

We can perform a series of row operations on $A$ which consists entirely of adding a multiple of one row to another row. This operation does not change the determinant of $A$.

If we have $d_i=0$ for some $i\le n$, then the row operations would give us a row which consists entirely of zeroes, hence $\det(A) = 0 = d_1 d_2\dots d_n$.

If we $d_i\ne 0$ for all $i\le n$, then the row operations would give us a diagonal matrix whose elements on the diagonal are still $d_1,\dots,d_n$. Obviously, the determinant of the diagonal matrix is $d_1 d_2\dots d_n$, thus we have $$ \det(A) = d_1 d_2\dots d_n $$ in either cases.

  • $\begingroup$ Unless the matrices are of infinite order and the quantities do not exist. If they do exist, equality carries over to the infinite order case. $\endgroup$ – Oscar Lanzi Feb 28 at 16:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.