I am dealing with some nice rings that are always isomorphic to some fairly nice quotient ring of a polynomial ring. A typical example is:

$$ \mathbb{C}[X,XY,XY^2] \cong \frac{\mathbb{C}[U,V,W]}{\langle V^2 - UW \rangle}. $$

I would like a nice way to write the Kahler differentials of such rings. For example when we have the following ring:

$$ \mathbb{C}[X^{ \pm 1}] \cong A := \frac{\mathbb{C}[U,V]}{\langle UV \rangle} $$ There is already a nice way of writing the differential - $d(f(X)) = \frac{\partial f}{\partial X} dX $

but also all $\ f(U,V) + \langle UV \rangle \ \in A \ $ can be written uniquely as $h(U) + g(V) + \langle UV \rangle $ for polynomials $h$ and $g$.

Then we can write something like: $d( f(U,V) + \langle UV \rangle ) = (\frac{\partial h}{\partial U} +\langle UV \rangle)dU + (\frac{\partial g}{\partial V} + \langle UV \rangle)dV + \langle d(UV)\rangle$

Note this is equivalent to the standard way to write the differential.

Can this be generalized? For example can I do this with the first example I gave?

really this is me trying to just get a nice way to write these maps, as they behave a lot like the standard way of writing the Kahler differential but the notation means i can't write $\frac{\partial f}{\partial X} $ for example.


I think the general answer to your question is given by the second fundamental sequence $$\mathfrak{m}/\mathfrak{m}^2 \rightarrow \Omega_{A/k}\otimes_k B \rightarrow \Omega_{B/k} \rightarrow 0$$ where $B=A/\mathfrak{m}$.

In practice this means that for a ring like $B=\mathbb{C}[u,v,w]/(v^2-uw)$,

$\Omega_B$ is generated by $du,dv,dw$ subject to the relation,


  • $\begingroup$ Does this give me a nice way to write the differential map $A \rightarrow \Omega^1 $ (hopefully in terms of the partial derivatives)? I feel like the quotient will mess things up. $\endgroup$ – JC574 Aug 12 '14 at 17:37
  • 1
    $\begingroup$ Isnt the differential map just $v\mapsto dv$ etc. $\endgroup$ – Rene Schipperus Aug 12 '14 at 17:41
  • $\begingroup$ Sorry, I guess I mean in the case $C[x,y]$ I can write $d: f \mapsto \frac{\partial f}{\partial x} dx + \frac{\partial f}{\partial y} dy $, and then it's easy to work out closed and exact forms and kernels etc. I'd like something like that here $\endgroup$ – JC574 Aug 12 '14 at 17:44
  • 1
    $\begingroup$ yeah doesnt that formula follow automatically, since $d$ is a derivation. I havent thought about how you define the isomorphism between your two expression, do you mean what happens to $d$ under this iso ? $\endgroup$ – Rene Schipperus Aug 12 '14 at 17:48
  • 1
    $\begingroup$ I think the point is that the differentials $dx, dy$ etc are no longer indedependent. $\endgroup$ – Rene Schipperus Aug 12 '14 at 17:52

In differential geometry, if you have a differential form on a manifold $X$ and a smooth submanifold $Y \subseteq X$, then you can always "pull back" any differential form on $X$ to a differential form on $Y$ -- all of the algebra of differential forms, scalar fields, exterior derivatives, and such remains the same: all that pulling back does is add new relations between things. (tangent vectors go in the opposite direction, from $Y$ to $X$, so things become complicated if you are also looking at partial derivatives and such)

That's how I think of it here. I understand differentials on $\mathbb{C}[U,V,W]$. The process of passing to the quotient ring by the relation $V^2 - UW = 0$ means that on the module of differentials, you take the quotient by the relation

$$ 0 = d(V^2 - UW) = 2V dV - U dW - W dU$$

And in general, I just mod out the module of differentials by the submodule generated by the differentials of the new relations I'm putting to my ring.

Disclaimer: I tend to work with relatively nice rings, so this line of thought may have some problems in full generality.

  • $\begingroup$ this is awesome, thanks. My rings are all very nice, so it should be fine. $\endgroup$ – JC574 Aug 12 '14 at 17:31
  • $\begingroup$ good answer but i gave it to rene. nothing personal, thanks for all the help $\endgroup$ – JC574 Aug 12 '14 at 17:54

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.