Lemma 60.6.3. Let (A, I, \gamma ) \to (B, J, \delta ) be a homomorphism of divided power rings. Let (B(1), J(1), \delta (1)) be the coproduct of (B, J, \delta ) with itself over (A, I, \gamma ), i.e., such that
\xymatrix{ (B, J, \delta ) \ar[r] & (B(1), J(1), \delta (1)) \\ (A, I, \gamma ) \ar[r] \ar[u] & (B, J, \delta ) \ar[u] }
is cocartesian. Denote K = \mathop{\mathrm{Ker}}(B(1) \to B). Then K \cap J(1) \subset J(1) is preserved by the divided power structure and
\Omega _{B/A, \delta } = K/ \left(K^2 + (K \cap J(1))^{[2]}\right)
canonically.
Proof.
The fact that K \cap J(1) \subset J(1) is preserved by the divided power structure follows from the fact that B(1) \to B is a homomorphism of divided power rings.
Recall that K/K^2 has a canonical B-module structure. Denote s_0, s_1 : B \to B(1) the two coprojections and consider the map \text{d} : B \to K/(K^2 +(K \cap J(1))^{[2]}) given by b \mapsto s_1(b) - s_0(b). It is clear that \text{d} is additive, annihilates A, and satisfies the Leibniz rule. We claim that \text{d} is a divided power A-derivation. Let x \in J. Set y = s_1(x) and z = s_0(x). We will use \delta instead of \delta (1) for the divided power structure on J(1). We have to show that \delta _ n(y) - \delta _ n(z) = \delta _{n - 1}(y)(y - z) modulo K^2 +(K \cap J(1))^{[2]} for n \geq 1. The equality holds for n = 1. Assume n > 1. Note that \delta _ i(y - z) lies in (K \cap J(1))^{[2]} for i > 1. Calculating modulo K^2 + (K \cap J(1))^{[2]} we have
\delta _ n(z) = \delta _ n(z - y + y) = \sum \nolimits _{i = 0}^ n \delta _ i(z - y)\delta _{n - i}(y) = \delta _{n - 1}(y) \delta _1(z - y) + \delta _ n(y)
This proves the desired equality.
Let M be a B-module. Let \theta : B \to M be a divided power A-derivation. Set D = B \oplus M where M is an ideal of square zero. Define a divided power structure on J \oplus M \subset D by setting \delta _ n(x + m) = \delta _ n(x) + \delta _{n - 1}(x)m for n > 1, see Lemma 60.3.1. There are two divided power algebra homomorphisms B \to D: the first is given by the inclusion and the second by the map b \mapsto b + \theta (b). Hence we get a canonical homomorphism B(1) \to D of divided power algebras over (A, I, \gamma ). This induces a map K \to M which annihilates K^2 (as M is an ideal of square zero) and (K \cap J(1))^{[2]} as M^{[2]} = 0. The composition B \to K/K^2 + (K \cap J(1))^{[2]} \to M equals \theta by construction. It follows that \text{d} is a universal divided power A-derivation and we win.
\square
Comments (4)
Comment #4222 by Dario Weißmann on
Comment #4400 by Johan on
Comment #4402 by Dario Weißmann on
Comment #4504 by Johan on
There are also: