Skip to main content
edited tags
Link
Martin Sleziak
  • 4.7k
  • 4
  • 35
  • 40
Formatting
Source Link
LSpice
  • 12.9k
  • 4
  • 45
  • 69

Suppose we are given the following:

An $n \times n$ matrix $A$ (with rational entries).

$\mbox{trace}(A)=0$

$B=A+I$ where $I$ is the $n \times n$ identity matrix.

  • An $n \times n$ matrix $A$ (with rational entries).

  • $\operatorname{trace}(A)=0$

  • $B=A+I$ where $I$ is the $n \times n$ identity matrix.

Question. What conditions on $A$ would guarantee the following to be true:

$\det(B) \leq 1$ and,

$\det(B)=1$ if and only if $A$ is nilpotent.

  • $\det(B) \leq 1$ and,

  • $\det(B)=1$ if and only if $A$ is nilpotent.

===============================================================

 

Example Condition: The eigenvalues of $A$ are real and at least $-1$.

To see why, let $\lambda_1,\dots,\lambda_n$$\lambda_1,\dotsc,\lambda_n$ be the eigenvalues of $A$. The trace condition on $A$ gives us $\sum_{i=1}^n \lambda_i=0$. Now the eigenvalues of $B$ are $1+\lambda_1,\ldots,1+\lambda_n$$1+\lambda_1,\dotsc,1+\lambda_n$ which are all nonnegative, so applying AM-GMAM–GM and using the trace condition on $A$, we get

$$\det(B) = \prod_i (1+\lambda_i) \leq \left(\frac{n+\sum_i \lambda_i}{n} \right)^n=1.$$

Equality holds if and only if $1+\lambda_1=\cdots=1+\lambda_n$$1+\lambda_1=\dotsb=1+\lambda_n$, which implies all the $\lambda_i$'s are equal, and subsequently by the trace condition on $A$ that all the $\lambda_i$'s are $0$. In other words, equality holds if and only if $A$ is nilpotent.

Suppose we are given the following:

An $n \times n$ matrix $A$ (with rational entries).

$\mbox{trace}(A)=0$

$B=A+I$ where $I$ is the $n \times n$ identity matrix.

Question What conditions on $A$ would guarantee the following to be true:

$\det(B) \leq 1$ and,

$\det(B)=1$ if and only if $A$ is nilpotent.

===============================================================

Example Condition: The eigenvalues of $A$ are real and at least $-1$.

To see why, let $\lambda_1,\dots,\lambda_n$ be the eigenvalues of $A$. The trace condition on $A$ gives us $\sum_{i=1}^n \lambda_i=0$. Now the eigenvalues of $B$ are $1+\lambda_1,\ldots,1+\lambda_n$ which are all nonnegative, so applying AM-GM and using the trace condition on $A$, we get

$$\det(B) = \prod_i (1+\lambda_i) \leq \left(\frac{n+\sum_i \lambda_i}{n} \right)^n=1.$$

Equality holds if and only if $1+\lambda_1=\cdots=1+\lambda_n$, which implies all the $\lambda_i$'s are equal, and subsequently by the trace condition on $A$ that all the $\lambda_i$'s are $0$. In other words, equality holds if and only if $A$ is nilpotent.

Suppose we are given the following:

  • An $n \times n$ matrix $A$ (with rational entries).

  • $\operatorname{trace}(A)=0$

  • $B=A+I$ where $I$ is the $n \times n$ identity matrix.

Question. What conditions on $A$ would guarantee the following to be true:

  • $\det(B) \leq 1$ and,

  • $\det(B)=1$ if and only if $A$ is nilpotent.

 

Example Condition: The eigenvalues of $A$ are real and at least $-1$.

To see why, let $\lambda_1,\dotsc,\lambda_n$ be the eigenvalues of $A$. The trace condition on $A$ gives us $\sum_{i=1}^n \lambda_i=0$. Now the eigenvalues of $B$ are $1+\lambda_1,\dotsc,1+\lambda_n$ which are all nonnegative, so applying AM–GM and using the trace condition on $A$, we get

$$\det(B) = \prod_i (1+\lambda_i) \leq \left(\frac{n+\sum_i \lambda_i}{n} \right)^n=1.$$

Equality holds if and only if $1+\lambda_1=\dotsb=1+\lambda_n$, which implies all the $\lambda_i$'s are equal, and subsequently by the trace condition on $A$ that all the $\lambda_i$'s are $0$. In other words, equality holds if and only if $A$ is nilpotent.

edited title
Link

For what matrices $A$ is the following true: $\det(I+A) \leq 1$ with equality if and only if $A$ is nilpotent

edited title
Link
Loading
added 2 characters in body
Source Link
Loading
deleted 1 character in body
Source Link
Loading
Source Link
Loading