Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $A=[a_{ij}]$ be a matrix of type $n\times n$ with coefficients with the field $K$, where $K=\mathbb{R}$ or $K=\mathbb{C}$.

In the case $K=\mathbb{C}$, for example by Jordan decomposition theorem, there exist an invertible matrice $B$ with coefficients from $K=\mathbb{C}$ such that $C:=B^{-1}AB$ is a upper triangular matrix, i.e $C=[c_{ij}]$ with $c_{ij}=0$ for $i>j$.

Is it true that for every matrice $A$ of type $n\times n$ with real coefficients, which characteristic polynomial has $n$ real roots not necessarily different, there exists an invertible matrix $B$ of the same type with real coefficients such that $B^{-1}AB$ is upper triangular?


share|cite|improve this question
Do you mean "$n$ not necessarily different real roots"? – user38268 Feb 8 '12 at 9:31
up vote 2 down vote accepted

The fact that the coefficients are different are different or not in $\chi_A(x)$ makes no change. The only thing that matters is whether all the roots are real or not. It is true that every real matrix with real eigenvalues (roots of $\chi_A(x)$) is triangulable. To prove this, you need the following two facts:
1) If $\lambda$ is an eigenvalue of $A$ then there exists $0\neq v\in\mathbb{R}^n$ such that $A\cdot v=\lambda v$.
2) If $W\lneqq \mathbb{R}^n$ and all roots of $\chi_A(x)$ are real, then there exists $v\in\mathbb{R}^n$ and $\lambda\in\mathbb{R}$ such that $v\notin W$ and $Av-\lambda v\in W$
Using those two claim, you can construct a basis of $\mathbb{R}^n$, $b=(v_1,...,v_n)$ such that for all $1\leq i\leq n$ $Av_i=a_{i,1}v_1+...+a_{i,i}v_i$. Now let $B$ be a matrix whose columns are $(v_1,...,v_n)$. It's easily checked that $B^{-1}AB$ is upper-triangular.

share|cite|improve this answer
How to find $v$ and $\lambda$ from the fact 2) ? – Richard Feb 8 '12 at 9:56
@Richard: Use the Cayley–Hamilton theorem. Take any $v\notin W$. Since $\chi_A(A)=0$, we have $\chi_A(A)v=0$. Since all the roots of $\chi_A(x)$ are real, we can write $\chi_A(A)=(A-\lambda_1I)\cdot\ldots\cdot(A-\lambda_nI)$. Denote $w=(A-\lambda_iI)\cdot\ldots\cdot(A-\lambda_nI)v$ such that $w\notin W$ and $(A-\lambda_{i-1}I)w\in W$ (exists since $v\notin W$ and $\chi_A(A)v=0\in W$) – Dennis Gulko Feb 8 '12 at 16:45

In the Schaum's series book on matrices by Frank Ayres jr., "Matrizen, Theorie und Anwendungen" (German) it is explicitely stated that

"We can prove: VI. each n-dimensional square matrix A is 'similar' to a triangular matrix, whose elements on the diagonal are the characteristic roots of A"
(Ed. 1987, Chap 20, my translation)

Similarity means here the multiplication $\small A=B^{-1}\cdot C \cdot B $ Also here is no restriction that the general field of complex numbers were required.
Moreover I think to remember, that B can be a rotation (unitary) matrix, but don't have a reference at hand.
(Perhaps that fact should also be included in wikipedia/similarity)

share|cite|improve this answer
Thanks for answer and references. – Richard Feb 9 '12 at 12:33

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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