For the upper triangular matrix ring given by $\begin{pmatrix}\mathbb{Q}&\mathbb{R}\\0&\mathbb{R}\end{pmatrix}$ I want to determine its left ideals. For example, this link refers to Lam's "A first course in non-commutative rings" applied to a generic matrix ring $\begin{pmatrix}R&M\\0&S\end{pmatrix}$:

The left ideals are all of the form $I_1\oplus I_2$ where $I_2$ is a left ideal of $S$, and $I_1$ is a left $R$ submodule of $R\oplus M$ which contains $MI_2$.

I want to translate this to explicitly determine all the left-ideals of my matrix ring above. So this is just a mere translation exercise. Let's see if I can do it correctly.

$I_2$ is a left ideal of $\mathbb{R} \implies I_2 = 0$ or $I_2 = \mathbb{R}$

$I_1$ is a left $\mathbb{Q}$-submodule of $\mathbb{Q} \oplus \mathbb{R}$ which contains $\mathbb{R}I_2$

So now, I unroll this complex definition.

A $\mathbb{Q}$-module is a vector space over $\mathbb{Q}$. The underlying set for this vector space is $\mathbb{Q} \oplus \mathbb{R}$ which I identify with $\mathbb{Q} \times \mathbb{R}$. So here I'm considering operations of the kind $q_1(q,r) = (q_1q,q_1r)$ and I have to determine the subspaces of this vector space. Is it as easy as taking a subspace of $\mathbb{Q}$ (probably $\{0\}$ or $\mathbb{Q}$ and a subspace of $\mathbb{R}$ over $\mathbb{Q}$ (giving a field extension?) and then doing the product? But then, how can this contain $\{0\},\mathbb{R}$ if it is bidimensional?

Can you help me to figure out this situation?


There is some sort of identification in the statement of the theorem because $MI_2$ should be $\mathbb{R}$ not $\{0\} \times \mathbb{R}$

ok, now I see the identification, it corresponds to $\begin{bmatrix}0 & M \\ 0 & 0\end{bmatrix}\begin{bmatrix}0 & 0 \\ 0 & I_2\end{bmatrix} = \begin{bmatrix}0 & MI_2 \\ 0 & 0\end{bmatrix}$

However I still don't see very well what should be explicitely the sets in @rschwieb answer. It seems to me that for the second one the possibilities are just $\mathbb{Q} \times \mathbb{R}$ and $\{0\} \times \mathbb{R}$. What about the first one?


There are just two cases:

$I_2=\{0\}$, and $I_1$ is any $\mathbb Q$ subspace of $\mathbb Q\times \mathbb R$

$I_2=\mathbb R$ and $I_1$ is any subspace of $\mathbb Q\times \mathbb R$ containing $\{0\}\times\mathbb R$.


$\left\{\begin{bmatrix}x&y\\0&z\end{bmatrix}\,\middle|\, (x,y)\in I_1, z\in I_2\right\}$ is a left ideal.

As you can see, there are only two left ideals of the second type, and there are infinitely many left ideals of the first type.

  • $\begingroup$ @JacquesSaliba Like it says... any $\mathbb Q$ subspace of $\mathbb Q\times \mathbb R$. What else is there to say? $\mathbb R$ is an infinite dimensional $\mathbb Q$ space, so there are lots of things that could happen. $\endgroup$ – rschwieb Apr 6 '18 at 14:56
  • $\begingroup$ @JacquesSaliba I think you are making a mental mistake in the first sentence of your last comment. YES, either all entries of the left hand size will be $0$, or else you will see all of $\mathbb Q$. On the other hand, $\mathbb Q\times\{0\}$ is not necessarily contained in $I_1$. So you could have, for example, $\langle(1,\pi)\rangle$ which is not of the form $A\times B$ for $A\subseteq \mathbb Q$ and $B\subseteq \mathbb R$. There are lots and LOTS of $\mathbb Q$ subspaces of $\mathbb Q\times \mathbb R$ with no easy description. $\endgroup$ – rschwieb Apr 6 '18 at 15:04
  • $\begingroup$ @JacquesSaliba You may as well just identify $\mathbb Q\times \mathbb R$ with $\bigoplus _{i\in I}\mathbb Q$ where $I$ is an index set (necessarily uncountable.) $\endgroup$ – rschwieb Apr 6 '18 at 15:06

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.