We first present a pedantic construction, then we introduce a simpler way of making the same construction.
Constructing \mathbb{P}^n by gluing.
Classical Construction
Before proceeding to the scheme case, let us look at the the classical \mathbb{P}^n(k) = \{[X_0: \ldots: X_n] \mid X_i \text{ not all $0$} \}/ \sim.
We see that \mathbb{P}^n is covered by pieces U_i = \{ X_i \neq 0\}. We can identify each U_i with affine space k^n (i.e. put affine coordinates on U_i) via the map \phi_i: U_i \to k^n given by
[X_0: \ldots: X_n] \mapsto (X_{0}/X_i, \ldots, \widehat{X_i/X_i}, \ldots, X_n/X_i).
If x \in U_i \cap U_j, what is the relationship between the coordinate \phi_i(x) and \phi_j(x)?
Suppose (x_{0/i}, \ldots, \widehat{x_{i/i}}, \ldots, x_{n/i}) denotes the coordinate of \phi_i(x). Then we must have \phi_j(x) = (\frac{x_{0/i}}{x_{j/i}}, \ldots, \widehat{x_{i/i}}, \ldots, \frac{x_{n/i}}{x_{j/i}}).
In other words, we have a map between coordinate rings of U_i and U_j, viewed as affine varieties under identification \phi_i and \phi_j given by
A(U_j) = k[x_{0/j}, \ldots, \widehat{x_{j/j}}, \ldots, x_{n/j}] \to A(U_i) = k[x_{0/i}, \ldots, \widehat{x_{i/i}}, \ldots, x_{n/i}]
x_{0/j} \mapsto \frac{x_{0/i}}{x_{j/i}} and so on.
Gluing Schemes
Motivated by the above discussion, for i = 0 to n, we let $X_i = \mathbb{A}^n = \text{Spec} k[x_{0/i}, \ldots, \widehat{x_{i/i}}, \ldots, x_{n/i}] = \text{Spec} \left(k[x_{0/i}, \ldots, x_{n/i}] /(x_{i/i} - 1)\right).$
Let X_{ij} = \text{Spec} \left(k\left[\frac{x_{0/i}}{x_{j/i}}, \ldots, \frac{x_{n/i}}{x_{j/i}} \right]/(x_{i/i}- 1)\right) = \text{Spec}\left(k\left[x_{0/i}, \ldots, x_{n/i}, \frac{1}{x_{j/i}}\right]/(x_{i/i} - 1) \right) =: \text{Spec} A_{ij}
be an open subset of X_i.
We have the following identification \varphi_{ij}: X_{ij} \to X_{ji} corresponding to the ring homomorphism A_{ji} \to A_{ij} given by
x_{0/j} \mapsto \frac{x_{0/i}}{x_{j/i}}
and so on.
Note that this map is well defined since x_{j/j} - 1 is mapped to \frac{x_{j/i}}{x_{j/i}} - 1 = 0.
It is easy to see that \varphi_{ij} and X_{ij} satisfies the cocycle condition. This is because
X_{ij} \cap X_{ik} is affine and equal to \text{Spec}(A_{ijk}) where
A_{ijk} = k\left[x_{0/i}, \ldots, x_{n/i}, \frac{1}{x_{j/i}}, \frac{1}{x_{k/i}}\right]/(x_{i/i} - 1).
As before, the map \varphi_{ij}|_{X_{ijk}}: X_{ijk} \to X_{jik} (restriction of \varphi_{ij}) corresponds to the map A_{jik} \to A_{ijk} (extension of the previous map) given by
x_{0/i} \mapsto \frac{x_{0/i}}{x_{j/i}}.
In particular \frac{1}{x_{j/i}} \mapsto 1 and \frac{1}{x_{k/i}} \mapsto \frac{x_{k_i}}{x_{j/i}}.
On the other hand the map \varphi_{kj} \circ \varphi_{ik}|_{X_{ijk}} corresponds to the map
A_{ijk} \to A_{kji} \to A_{jik} given by
x_{0/i} \mapsto \frac{x_{0/i}}{x_{k/i}} \mapsto \frac{\frac{x_{0/i}}{x_{k/i}}}{\frac{x_{j/i}}{x_{k_i}}} = \frac{x_{0/i}}{x_{j/i}}.
Thus we can glue X_i along X_{ij} via \varphi_{ij} to get a unique scheme \mathbb{P}^n.
In the interpretation of Bosch, the second equalitz concerning A_i is not true. The first one is the homogeneous localization at x_i but the second one is the unconditional localization.
ReplyDelete