Lemma 15.12.5. Let $(A, I) = \mathop{\mathrm{colim}}\nolimits (A_ i, I_ i)$ be a filtered colimit of pairs. The functor of Lemma 15.12.1 gives $A^ h = \mathop{\mathrm{colim}}\nolimits A_ i^ h$ and $I^ h = \mathop{\mathrm{colim}}\nolimits I_ i^ h$.

Proof. By Categories, Lemma 4.24.6 we see that $(A^ h, I^ h)$ is the colimit of the system $(A_ i^ h, I_ i^ h)$ in the category of henselian pairs. Thus for a henselian pair $(B, J)$ we have

$\mathop{Mor}\nolimits ((A^ h, I^ h), (B, J)) = \mathop{\mathrm{lim}}\nolimits \mathop{Mor}\nolimits ((A_ i^ h, I_ i^ h), (B, J)) = \mathop{Mor}\nolimits (\mathop{\mathrm{colim}}\nolimits (A_ i^ h, I_ i^ h), (B, J))$

Here the colimit is in the category of pairs. Since the colimit is filtered we obtain $\mathop{\mathrm{colim}}\nolimits (A_ i^ h, I_ i^ h) = (\mathop{\mathrm{colim}}\nolimits A_ i^ h, \mathop{\mathrm{colim}}\nolimits I_ i^ h)$ in the category of pairs; details omitted. Again using the colimit is filtered, this is a henselian pair (Lemma 15.11.12). Hence by the Yoneda lemma we find $(A^ h, I^ h) = (\mathop{\mathrm{colim}}\nolimits A_ i^ h, \mathop{\mathrm{colim}}\nolimits I_ i^ h)$. $\square$

Comment #4896 by Laurent Moret-Bailly on

Lemma 0038 says that $(A,I)^h$ is the colimit of the system $(A_i^h,I_i^h)$ in the category of henselian pairs. This colimit is (looking at the universal properties) the henselization of $(\operatorname{colim} A_i^h, \operatorname{colim} I_i^h)$. The latter is henselian if the colimit is filtered, but in general this is not clear to me.

Comment #4897 by Laurent Moret-Bailly on

Follow-up to Comment #4896: consider for instance, for an odd prime $p$, the coproduct of $(\mathbb{Z}_p,(p))$ with itself: in the category of pairs, this is $(A,(p))$ with $A:=\mathbb{Z}_p \otimes \mathbb{Z}_p$; however $A/(p)=\mathbb{F}_p$ and $A$ is not connected, so $(A,I)$ is not even a Zariski pair. (To see that $A$ is not connected, let $t:=\sqrt{1+p}\in 1+p\mathbb{Z}_p$. Then $\mathbb{Z}_p$ is faithfully flat over $A_0:=\mathbb{Z}[t]\left[\frac{1}{t+1}\right]$, hence $A$ is faithfully flat over $A_0\otimes_\mathbb{Z} A_0$ which is not connected.)

Comment #4901 by on

Good catch! I will fix this later today or tomorrow. Thanks very much!

Comment #5019 by 羽山籍真 on

In the comment 4897, if the prime $p=3$, then $t=2$. So $A_0=\mathbb{Z}[t]\left[\frac{1}{1+t} \right]=\mathbb{Z}[2]\left[\frac{1}{3}\right]=\mathbb{Z}\left[\frac{1}{3}\right]$. So $A_0\rightarrow \mathbb{Z}_3$ is $\mathbb{Z}\left[\frac{1}{3}\right]\rightarrow \mathbb{Z}_3$, which is problematic because the image $x$ of $\frac{1}{3}$ should satisfy $3\cdot x=1$, but such $x$ does not exist in $\mathbb{Z}_3$.

Comment #5020 by Laurent Moret-Bailly on

@ 羽山籍真: No, $t$ is the square root of $1+p$ which is $\equiv1$ mod $p$, so in this case it is $-2$, not $2$, and $A_0=\mathbb{Z}$.

Comment #5021 by 羽山籍真 on

@Laurent Moret-Bailly Thanks. But in this case, $A_0\rightarrow \mathbb{Z}_p$ is $\mathbb{Z}\rightarrow \mathbb{Z}_3$, which is not faithfully flat.

Comment #5035 by slogan_bot on

Suggested slogan: "Henselization of pairs commutes with filtered colimits"

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).