Let $A_0$ be a finite dimensional $k$-algebra where $k$ is an algebraically closed field of characteristic $p$ not necessarily with unit. We wish to show there is a $k$-algebra $A$ with a unit $1$ containing $A_0$ as a $k$-subalgebra of $A$ of co-dimension $1$.


Let $\{b_1,\dots,b_n\}$ be a $k$-basis of $A_0$. Formally define $A$ to be the $k$-span of $$\{b_1,\dots,b_n\} \cup \{1\}$$ where we define $$a_0 \cdot 1 = 1 \cdot a_0 = a_0 \tag{$\forall a_0 \in A_0$}$$ and $$k \cdot 1 = \underbrace{1+\dots+1}_{k \text{ times}}$$ Then it is clear that $A_0$ is a subalgebra.


There is a mapping of $k$ into the center of $A_0$ and so I can think of elements of $k$ in $A$, although not every element of $k$ is uniquely embedded in $A_0$. So, any $s \in k$ should be in the $k$-span of $\{b_1,\dots,b_n\}$ so whether or not my algebra $A$ is actually well-defined and whether or not $A_0$ would be of codimension $1$ seems unclear as the expression of every element should be unique. Clearly, I am losing some inuition somewhere in the definition of an algebra. Clarification would be greatly appreciated.

Follow Up Question:

I didn't use anything but finite dimensionality so I presume this holds in the much more general case of an algebra over a commutative ring $R$ where our algebra is finite dimensional over $R$?


I think the most natural idea is to take the unitization $A^1=k\times A_0$ where addition is defined to be $(\alpha, a)+(\beta, b)=(\alpha+\beta, a+b)$ and multiplication is defined $(\alpha, a)(\beta, b)=(\alpha\beta, \alpha b+\beta a+ab)$, where the identity will be $(1,0)$, and $A_0$ is embedded as the subalgebra $\{(0,a)\mid a\in A_0\}$.

I don't find your description of the algebra in the attempt very clear, but what I am describing is much like that.

It has nothing to do with finite dimensionality of $A_0$ or algebraic closedness or characteristic of $k$.

  • $\begingroup$ I imagine this is likely what Alperin wanted me to come up with, but it doesn't quite address my question. Do the elements of $k$ not already live in $A_0$ in some way? Doesn't that mess up the uniqueness of our linear combos? Or are we thinking about $k$ essentially being given an action on $A_0$? $\endgroup$ – Aaron Zolotor Apr 2 at 20:17
  • 1
    $\begingroup$ @AaronZolotor no, in an algebra without identity, there is not necessarily a copy of k . For example, any vector space with the zero product. The construction above furnishes at least one copy of $k$ as $k\times \{0\}$ $\endgroup$ – rschwieb Apr 2 at 23:12
  • $\begingroup$ @AaronZolotor I don't really see the necessity of talking about a basis. The Dorroh extension (the unitization I'm talking about) essentially does this: suppose I have a $k$-algebra $A_0$ and I want to lump it into a bigger ring with $k$. It would have to have things that look like $k+a$ for $k\in K$ and $a\in A_0$. What would $k+a+k'+a'$ look like? What would $(k+a)(k'+a')$ look like? After you look at that, the construction above follows naturally. $\endgroup$ – rschwieb Apr 3 at 15:19

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.