11

The definition of a smooth morphism of schemes $f:X \rightarrow S \space$ given here on stacks project is so abstract that it is intractable to me.

$f$ is called smooth if for every point $x$ of $X$, there exists an affine open neighborhood $U$ of $x$, and an affine open set $V$ of $S$, with $f(U) \subset V$, such that the induced ring homomorphism $\mathcal O_S(V) \rightarrow \mathcal O_X(U)$ is a smooth ring homomorphism. A smooth ring homomorphism is defined in terms of the "naive cotangent complex."

I just have no idea how to think about or work with this definition. So let me consider a more special case: let $R$ be a discrete valuation ring with quotient field $K$ and residue field $k$. Let $X = \operatorname{Spec} A$ be a scheme of finite type over $R$, say $X \subset \mathbb A_R^n$.

What does it mean to say that $X$ is smooth over $R$? Can we give a simpler criterion than the one given on stacks project? I'm sure a necessary condition is that $A$ is a flat $R$-algebra. Can we talk about the smoothness of $X$ over $R$ in terms of the smoothness of the generic and special fibres? That is, in terms of $X_K = \operatorname{Spec} A \otimes_R K$ and $X_k = \operatorname{Spec} A \otimes_R k$ being smooth varieties over $K$ and $k$?

D_S
  • 33,891
  • 12
    Anyone attempting to study algebraic geometry nowadays has all of my sympathy. – Randall Sep 21 '18 at 15:58
  • 3
    The way I like to think about smooth morphisms is via Theorem 25.2.2 in Vakil - which says that $\pi:X\rightarrow Y$ is smooth (of relative dimension $n$) if it is locally finitely presented, flat and geometric fibers are smooth $k$-schemes of dimension $n$.

    This tells us that a smooth morphism is one where 1) your fibers are smooth varieties and 2) your fibers are varying "nicely" (flat).

    – loch Sep 21 '18 at 17:19
  • 2
    It should be pointed out that the first sentence is just saying that a map is smooth if there exists an affine cover so that on each affine piece, the map is $\operatorname{Spec}$ of a smooth ring map. This is what you should expect: when defining a concept for schemes, it's fairly common to define what should happen for affine schemes and then patch everything together in this fashion, just in the same way that schemes are patched together out of affine schemes. – KReiser Sep 21 '18 at 18:11

1 Answers1

5

Okay, now that I am reading a bit more of that page, I see that smoothness can be characterized in more concrete way:

Let $\phi: A \rightarrow B$ be a ring homomorphism making $B$ into a finitely presented $A$-algebra. We say that $\phi$ is standard smooth is $B$ can be realized in the form $B = A[x_1, ... , x_n]/(f_1, ... , f_c)$ for $c \leq n$, such that the image of the polynomial

$$\operatorname{det}\begin{pmatrix} \frac{\partial f_i}{\partial x_j}\end{pmatrix}_{1 \leq i, j \leq c} \in A[x_1, ... , x_n]$$

is a unit in $B$.

Then $f: X \rightarrow S$ is smooth if and only if for every $x \in X$, there exists an affine open neighborhood $U$ of $x$ and an affine open set $V$ of $S$ such that $f(U) \subset V$ and $U \rightarrow V$ corresponds to a standard smooth ring homomorphism. (Lemma 28.32.11).

Also, in the example I asked about, with $R$ a DVR and $X= \operatorname{Spec} A$ a scheme of finite type over $R$, (28.32.3) says that in order to say that $X$ is smooth over $R$, it is sufficient that $X$ be over $R$ and that the generic fibre $X \times_R \operatorname{Spec}K$ and special fibre $X \times_R \operatorname{Spec}k$ be smooth varieties in the usual sense.

Another definition of smoothness (given here): $f: X \rightarrow Y$ is smooth if it is locally of finite presentation, flat, and if for all $y \in Y$, $X \times_Y \operatorname{Spec} \kappa(y)$ is smooth as a scheme over the field $\kappa(y)$. Equivalently, this last condition can be restated as saying $X \times_Y \operatorname{Spec} \overline{\kappa(y)}$ is regular over $\overline{\kappa(y)}$.

D_S
  • 33,891
  • In your last but one paragraph, did you miss out "$X$ be flat over $R$"? As $A$ is fg $R-$algebra and $R$ is Noetherian, so $A$ is an $R-$algebra of finite presentation. – Doug Feb 29 '20 at 10:08