Skip to content

Commit

Permalink
Lectures 07A
Browse files Browse the repository at this point in the history
  • Loading branch information
zeramorphic committed Oct 19, 2023
1 parent f6e319a commit 15783d6
Showing 1 changed file with 262 additions and 1 deletion.
263 changes: 262 additions & 1 deletion iii/commalg/02_tensor_products.tex
Original file line number Diff line number Diff line change
Expand Up @@ -819,7 +819,7 @@ \subsection{Exactness properties of the tensor product}
We show this is well-defined.
If \( g(x) = g(y) \), then \( g(x - y) = 0 \), so \( x - y = f(a) \) for some \( a \in A \).
But then \( \varphi(f(a)) = 0 \), so \( \varphi(x) = \varphi(y) \).
As \( \varphi \) and \( g \) are ring homomorphisms, so is \( \psi \).
As \( \varphi \) and \( g \) are \( R \)-linear, so is \( \psi \).
Hence \( g^\star(\psi) = \varphi \) as required.
\end{proof}
\begin{lemma}
Expand Down Expand Up @@ -871,3 +871,264 @@ \subsection{Exactness properties of the tensor product}
\end{proof}
By the universal property of the tensor product,
\[ \Hom_R(M \otimes_R N, L) \simeq \operatorname{Bilin}_R(M \times N, L) \simeq \Hom_R(N, \Hom_R(M, L)) \]
mapping \( \varphi \mapsto n \mapsto m \mapsto \varphi(m \otimes n) \) and \( \varphi \mapsto m \otimes n \mapsto \varphi(m)(n) \).
This bijection is \emph{natural}, in the sense that many commutative diagrams involving them will commute.
\begin{proposition}
Let \( M \) be an \( R \)-module.
Then the functor \( T_M = M \otimes_R (-) \) is right exact.
\end{proposition}
\begin{proof}
Consider an exact sequence of \( R \)-modules
% https://q.uiver.app/#q=WzAsNCxbMCwwLCJBIl0sWzEsMCwiQiJdLFsyLDAsIkMiXSxbMywwLCIwIl0sWzAsMSwiZiJdLFsxLDIsImciXSxbMiwzXV0=
\[\begin{tikzcd}
A & B & C & 0
\arrow["f", from=1-1, to=1-2]
\arrow["g", from=1-2, to=1-3]
\arrow[from=1-3, to=1-4]
\end{tikzcd}\]
We must show that
% https://q.uiver.app/#q=WzAsNCxbMCwwLCJNIFxcb3RpbWVzX1IgQSJdLFsxLDAsIk0gXFxvdGltZXNfUiBCIl0sWzIsMCwiTSBcXG90aW1lc19SIEMiXSxbMywwLCIwIl0sWzAsMSwiXFxpZF9NIFxcb3RpbWVzIGYiXSxbMSwyLCJcXGlkX00gXFxvdGltZXMgZyJdLFsyLDNdXQ==
\[\begin{tikzcd}
{M \otimes_R A} & {M \otimes_R B} & {M \otimes_R C} & 0
\arrow["{\id_M \otimes f}", from=1-1, to=1-2]
\arrow["{\id_M \otimes g}", from=1-2, to=1-3]
\arrow[from=1-3, to=1-4]
\end{tikzcd}\]
is exact.
Let \( P \) be an \( R \)-module, and consider apply the functor \( \Hom(-, P) \) to this sequence.
As this is left exact, the resulting sequence will be exact.
% https://q.uiver.app/#q=WzAsNCxbMCwwLCIwIl0sWzEsMCwiXFxIb21fUihDLCBQKSJdLFsyLDAsIlxcSG9tX1IoQiwgUCkiXSxbMywwLCJcXEhvbV9SKEEsIFApIl0sWzAsMV0sWzEsMiwiZ15cXHN0YXIiXSxbMiwzLCJmXlxcc3RhciJdXQ==
\[\begin{tikzcd}
0 & {\Hom_R(C, P)} & {\Hom_R(B, P)} & {\Hom_R(A, P)}
\arrow[from=1-1, to=1-2]
\arrow["{g^\star}", from=1-2, to=1-3]
\arrow["{f^\star}", from=1-3, to=1-4]
\end{tikzcd}\]
Then, apply the functor \( \Hom(M, -) \), which is also left exact.
% https://q.uiver.app/#q=WzAsNCxbMCwwLCIwIl0sWzEsMCwiXFxIb21fUihNLCBcXEhvbV9SKEMsIFApKSJdLFsyLDAsIlxcSG9tX1IoTSwgXFxIb21fUihCLCBQKSkiXSxbMywwLCJcXEhvbV9SKE0sIFxcSG9tX1IoQSwgUCkpIl0sWzAsMV0sWzEsMiwiKGdeXFxzdGFyKV9cXHN0YXIiXSxbMiwzLCIoZl5cXHN0YXIpX1xcc3RhciJdXQ==
\[\begin{tikzcd}
0 & {\Hom_R(M, \Hom_R(C, P))} & {\Hom_R(M, \Hom_R(B, P))} & {\Hom_R(M, \Hom_R(A, P))}
\arrow[from=1-1, to=1-2]
\arrow["{(g^\star)_\star}", from=1-2, to=1-3]
\arrow["{(f^\star)_\star}", from=1-3, to=1-4]
\end{tikzcd}\]
We thus obtain
% https://q.uiver.app/#q=WzAsOCxbMCwxLCIwIl0sWzEsMSwiXFxIb21fUihNIFxcb3RpbWVzX1IgQywgUCkiXSxbMiwxLCJcXEhvbV9SKE0gXFxvdGltZXNfUiBCLCBQKSJdLFszLDEsIlxcSG9tX1IoTSBcXG90aW1lc19SIEEsIFApIl0sWzAsMCwiMCJdLFsxLDAsIlxcSG9tX1IoTSwgXFxIb21fUihDLCBQKSkiXSxbMiwwLCJcXEhvbV9SKE0sIFxcSG9tX1IoQiwgUCkpIl0sWzMsMCwiXFxIb21fUihNLCBcXEhvbV9SKEEsIFApKSJdLFswLDFdLFsxLDJdLFsyLDNdLFs0LDVdLFs1LDZdLFs2LDddLFs3LDMsIlxcc2ltZXEiXSxbNiwyLCJcXHNpbWVxIl0sWzUsMSwiXFxzaW1lcSJdLFs0LDAsIlxcc2ltZXEiXV0=
\[\begin{tikzcd}
0 & {\Hom_R(M, \Hom_R(C, P))} & {\Hom_R(M, \Hom_R(B, P))} & {\Hom_R(M, \Hom_R(A, P))} \\
0 & {\Hom_R(M \otimes_R C, P)} & {\Hom_R(M \otimes_R B, P)} & {\Hom_R(M \otimes_R A, P)}
\arrow[from=2-1, to=2-2]
\arrow[from=2-2, to=2-3]
\arrow[from=2-3, to=2-4]
\arrow[from=1-1, to=1-2]
\arrow[from=1-2, to=1-3]
\arrow[from=1-3, to=1-4]
\arrow["\simeq", from=1-4, to=2-4]
\arrow["\simeq", from=1-3, to=2-3]
\arrow["\simeq", from=1-2, to=2-2]
\arrow["\simeq", from=1-1, to=2-1]
\end{tikzcd}\]
As this diagram commutes, the bottom sequence is exact.
Since this holds for all \( P \), by the previous lemma, we can cancel \( P \) to give exact sequences
\[\begin{tikzcd}
0 & {M \otimes_R C} & {M \otimes_R B}
\arrow[from=1-2, to=1-3]
\arrow[from=1-1, to=1-2]
\end{tikzcd}
\quad\quad
\begin{tikzcd}
{M \otimes_R C} & {M \otimes_R B} & {M \otimes_R A}
\arrow[from=1-1, to=1-2]
\arrow[from=1-2, to=1-3]
\end{tikzcd}\]
which combine into the longer sequence as required.
\end{proof}
\begin{remark}
It is not the case that if
% https://q.uiver.app/#q=WzAsMyxbMCwwLCJBIl0sWzEsMCwiQiJdLFsyLDAsIkMiXSxbMCwxXSxbMSwyXV0=
\[\begin{tikzcd}
A & B & C
\arrow[from=1-1, to=1-2]
\arrow[from=1-2, to=1-3]
\end{tikzcd}\]
is exact, then
% https://q.uiver.app/#q=WzAsMyxbMCwwLCJNIFxcb3RpbWVzX1IgQSJdLFsxLDAsIk0gXFxvdGltZXNfUiBCIl0sWzIsMCwiTSBcXG90aW1lc19SIEMiXSxbMCwxXSxbMSwyXV0=
\[\begin{tikzcd}
{M \otimes_R A} & {M \otimes_R B} & {M \otimes_R C}
\arrow[from=1-1, to=1-2]
\arrow[from=1-2, to=1-3]
\end{tikzcd}\]
is also exact; the fact that the sequence has a zero on the right is important.
Consider the exact sequence
% https://q.uiver.app/#q=WzAsMyxbMCwwLCIwIl0sWzEsMCwiXFxtYXRoYmIgWiJdLFsyLDAsIlxcbWF0aGJiIFoiXSxbMCwxXSxbMSwyLCJcXHRpbWVzIDIiXV0=
\[\begin{tikzcd}
0 & {\mathbb Z} & {\mathbb Z}
\arrow[from=1-1, to=1-2]
\arrow["{\times 2}", from=1-2, to=1-3]
\end{tikzcd}\]
and tensor with \( \faktor{\mathbb Z}{2\mathbb Z} \).
We would then obtain
% https://q.uiver.app/#q=WzAsNixbMCwwLCIwIl0sWzEsMCwiXFxmYWt0b3J7XFxtYXRoYmIgWn17MlxcbWF0aGJiIFp9IFxcb3RpbWVzIFxcbWF0aGJiIFoiXSxbMiwwLCJcXGZha3RvcntcXG1hdGhiYiBafXsyXFxtYXRoYmIgWn0gXFxvdGltZXMgXFxtYXRoYmIgWiJdLFswLDEsIjAiXSxbMSwxLCJcXGZha3RvcntcXG1hdGhiYiBafXsyXFxtYXRoYmIgWn0iXSxbMiwxLCJcXGZha3RvcntcXG1hdGhiYiBafXsyXFxtYXRoYmIgWn0iXSxbMCwxXSxbMSwyXSxbMCwzLCJcXHNpbWVxIl0sWzEsNCwiXFxzaW1lcSJdLFsyLDUsIlxcc2ltZXEiXSxbMyw0XSxbNCw1XV0=
\[\begin{tikzcd}
0 & {\faktor{\mathbb Z}{2\mathbb Z} \otimes \mathbb Z} & {\faktor{\mathbb Z}{2\mathbb Z} \otimes \mathbb Z} \\
0 & {\faktor{\mathbb Z}{2\mathbb Z}} & {\faktor{\mathbb Z}{2\mathbb Z}}
\arrow[from=1-1, to=1-2]
\arrow["\times 2", from=1-2, to=1-3]
\arrow["\simeq", from=1-1, to=2-1]
\arrow["\simeq", from=1-2, to=2-2]
\arrow["\simeq", from=1-3, to=2-3]
\arrow[from=2-1, to=2-2]
\arrow["\times 2", from=2-2, to=2-3]
\end{tikzcd}\]
but this sequence is not exact.
\end{remark}

\subsection{Flat modules}
\begin{definition}
An \( R \)-module \( M \) is \emph{flat} if whenever \( f : N \to N' \) is \( R \)-linear and injective, the map
\[ \id_M \otimes f : M \otimes_R N \to M \otimes_R N' \]
is injective.
\end{definition}
\begin{example}
\begin{enumerate}
\item \( \faktor{\mathbb Z}{2\mathbb Z} \) is not a flat \( \mathbb Z \)-module.
\item Free modules are flat.
Suppose \( f : N \to N' \) is an injective \( R \)-linear map.
Then
% https://q.uiver.app/#q=WzAsNCxbMCwwLCJSXntcXG9wbHVzIEl9IFxcb3RpbWVzX1IgTiJdLFsxLDAsIlJee1xcb3BsdXMgSX0gXFxvdGltZXNfUiBOJyJdLFsxLDEsIihOJylee1xcb3BsdXMgSX0iXSxbMCwxLCJOXntcXG9wbHVzIEl9Il0sWzAsMSwiXFxpZF97Ul57XFxvcGx1cyBJfX0gXFxvdGltZXMgZiJdLFsxLDIsIlxcc2ltZXEiXSxbMCwzLCJcXHNpbWVxIiwyXSxbMywyLCJnIiwyXV0=
\[\begin{tikzcd}
{R^{\oplus I} \otimes_R N} & {R^{\oplus I} \otimes_R N'} \\
{N^{\oplus I}} & {(N')^{\oplus I}}
\arrow["{\id_{R^{\oplus I}} \otimes f}", from=1-1, to=1-2]
\arrow["\simeq", from=1-2, to=2-2]
\arrow["\simeq"', from=1-1, to=2-1]
\arrow["g"', from=2-1, to=2-2]
\end{tikzcd}\]
commutes, where
\[ g((n_i)_{i \in I}) = (f(n_i))_{i \in I} \]
But \( g \) is injective, so \( \id_{R^{\oplus I}} \otimes f \) must also be injective.
\item The base ring matters. One can see that \( \faktor{\mathbb Z}{2\mathbb Z} \) is not a flat \( \mathbb Z \)-module but it is a flat \( \faktor{\mathbb Z}{2\mathbb Z} \)-module as it is a free \( \faktor{\mathbb Z}{2\mathbb Z} \)-module.
\end{enumerate}
\end{example}
\begin{definition}
An \( R \)-module \( M \) is \emph{torsion-free} if \( rm \neq 0 \) whenever \( r \) is not a zero divisor in \( R \) and \( m \neq 0 \).
\end{definition}
\begin{proposition}
Flat modules are torsion-free.
\end{proposition}
\begin{proof}
Suppose \( M \) is not torsion-free.
Then there is \( r_0 \in R \) not a zero divisor and \( m_0 \neq 0 \), such that \( r_0 m_0 = 0 \).
Consider the \( R \)-linear map \( f : R \to R \) given by multiplication by \( r_0 \).
Its kernel is zero, as \( r_0 \) is not a zero divisor.
So \( f \) is injective.
The following diagram commutes.
% https://q.uiver.app/#q=WzAsNCxbMCwwLCJNIFxcb3RpbWVzX1IgUiJdLFsxLDAsIk0gXFxvdGltZXNfUiBSIl0sWzEsMSwiTSJdLFswLDEsIk0iXSxbMCwxLCJcXGlkX00gXFxvdGltZXMgZiJdLFsxLDIsIlxcc2ltZXEiXSxbMCwzLCJcXHNpbWVxIiwyXSxbMywyLCJtIFxcbWFwc3RvIHJfMCBtIiwyXV0=
\[\begin{tikzcd}
{M \otimes_R R} & {M \otimes_R R} \\
M & M
\arrow["{\id_M \otimes f}", from=1-1, to=1-2]
\arrow["\simeq", from=1-2, to=2-2]
\arrow["\simeq"', from=1-1, to=2-1]
\arrow["{m \mapsto r_0 m}"', from=2-1, to=2-2]
\end{tikzcd}\]
If \( M \) were flat, \( \id_M \otimes f \) would be injective, but then the map \( m \mapsto r_0 m \) would also be injective, which is a contradiction.
\end{proof}
% note: all integral domains are nonzero. we want the definition to really be: the zero ideal is prime.
\begin{example}
Let \( R \) be an integral domain, and let \( I \) be a nonzero ideal of \( R \).
Then \( \faktor{R}{I} \) is not flat.
Indeed, if \( I = R \) then \( \faktor{R}{I} = 0 \) is not flat.
Instead, suppose \( I \subsetneq R \), and let \( 0 \neq x \in I \).
Tensoring with \( \faktor{R}{I} \), the map \( \faktor{R}{I} \to \faktor{R}{I} \) given by multiplication by \( x \) is the zero map, but \( \faktor{R}{I} \) is not the zero module, so \( \faktor{R}{I} \) is not torsion-free.
\end{example}
\begin{proposition}
Let \( M \) be an \( R \)-module.
Then the following are equivalent.
\begin{enumerate}
\item \( T_M \) preserves exactness of all exact sequences;
\item \( T_M \) preserves exactness of short exact sequences;
\item \( M \) is flat;
\item if \( f : N \to N' \) is \( R \)-linear and injective, and \( N, N' \) are finitely generated \( R \)-modules, then \( \id_M \otimes f \) is injective.
\end{enumerate}
\end{proposition}
Note that a map \( f : M \to N \) is injective exactly when the sequence
% https://q.uiver.app/#q=WzAsNCxbMCwwLCJNIFxcb3RpbWVzX1IgUiJdLFsxLDAsIk0gXFxvdGltZXNfUiBSIl0sWzEsMSwiTSJdLFswLDEsIk0iXSxbMCwxLCJcXGlkX00gXFxvdGltZXMgZiJdLFsxLDIsIlxcc2ltZXEiXSxbMCwzLCJcXHNpbWVxIiwyXSxbMywyLCJtIFxcbWFwc3RvIHJfMCBtIiwyXV0=
\[\begin{tikzcd}
{M \otimes_R R} & {M \otimes_R R} \\
M & M
\arrow["{\id_M \otimes f}", from=1-1, to=1-2]
\arrow["\simeq", from=1-2, to=2-2]
\arrow["\simeq"', from=1-1, to=2-1]
\arrow["{m \mapsto r_0 m}"', from=2-1, to=2-2]
\end{tikzcd}\]
is exact, so all of these conditions relate exact sequences.
\begin{proof}
Note that (i) implies (ii) which implies (iii) which implies (iv).

\emph{(ii) implies (i).}
Suppose the sequence
% https://q.uiver.app/#q=WzAsMyxbMCwwLCJBIl0sWzEsMCwiQiJdLFsyLDAsIkMiXSxbMCwxLCJmIl0sWzEsMiwiZyJdXQ==
\[\begin{tikzcd}
A & B & C
\arrow["f", from=1-1, to=1-2]
\arrow["g", from=1-2, to=1-3]
\end{tikzcd}\]
is exact.
Then certainly
% https://q.uiver.app/#q=WzAsNSxbMCwwLCIwIl0sWzEsMCwiXFxmYWt0b3J7QX17XFxrZXIgZn0iXSxbMiwwLCJCIl0sWzMsMCwiXFxJbSBnIl0sWzQsMCwiMCJdLFswLDFdLFsxLDIsIlxcb3ZlcmxpbmUgZiJdLFsyLDMsImciXSxbMyw0XV0=
\[\begin{tikzcd}
0 & {\faktor{A}{\ker f}} & B & {\Im g} & 0
\arrow[from=1-1, to=1-2]
\arrow["{\overline f}", from=1-2, to=1-3]
\arrow["g", from=1-3, to=1-4]
\arrow[from=1-4, to=1-5]
\end{tikzcd}\]
is exact.
As the exactness of short exact sequences is preserved under \( T_M \) by hypothesis,
% https://q.uiver.app/#q=WzAsNSxbMCwwLCIwIl0sWzEsMCwiTSBcXG90aW1lc19SIFxcZmFrdG9ye0F9e1xca2VyIGZ9Il0sWzIsMCwiTSBcXG90aW1lc19SIEIiXSxbMywwLCJNIFxcb3RpbWVzX1IgXFxJbSBnIl0sWzQsMCwiMCJdLFswLDFdLFsxLDIsIlxcaWRfTSBcXG90aW1lcyBcXG92ZXJsaW5lIGYiXSxbMiwzLCJcXGlkX00gXFxvdGltZXMgZyJdLFszLDRdXQ==
\[\begin{tikzcd}
0 & {M \otimes_R \faktor{A}{\ker f}} & {M \otimes_R B} & {M \otimes_R \Im g} & 0
\arrow[from=1-1, to=1-2]
\arrow["{\id_M \otimes \overline f}", from=1-2, to=1-3]
\arrow["{\id_M \otimes g}", from=1-3, to=1-4]
\arrow[from=1-4, to=1-5]
\end{tikzcd}\]
is also exact.
Then
\[ \ker (\id_M \otimes g) = \Im (\id_M \otimes \overline f) = \Im (\id_M \otimes f) \]
Thus
% https://q.uiver.app/#q=WzAsMyxbMCwwLCJNIFxcb3RpbWVzX1IgQSJdLFsxLDAsIk0gXFxvdGltZXNfUiBCIl0sWzIsMCwiTSBcXG90aW1lc19SIEMiXSxbMCwxLCJcXGlkX00gXFxvdGltZXMgZiJdLFsxLDIsIlxcaWRfTSBcXG90aW1lcyBnIl1d
\[\begin{tikzcd}
{M \otimes_R A} & {M \otimes_R B} & {M \otimes_R C}
\arrow["{\id_M \otimes f}", from=1-1, to=1-2]
\arrow["{\id_M \otimes g}", from=1-2, to=1-3]
\end{tikzcd}\]
is exact.

\emph{(iii) implies (ii).}
\( T_M \) is right exact.
% write in more detail. need to use flatness to do the LHS of the sequence.

\emph{(iv) implies (iii).}
% todo: use prop above about being zero in finitely generated submodules
\end{proof}
\begin{proposition}
Let \( f : R \to S \) be a ring homomorphism, and let \( M \) be a flat \( R \)-module.
Then \( S \otimes_R M \) is a flat \( S \)-module.
\end{proposition}
\begin{proof}
Let \( g : N \to N' \) be an \( S \)-linear injective map.
Then
\[\begin{tikzcd}
{(S \otimes_R M) \otimes_S N} & {(S \otimes_R M) \otimes_S N'} \\
{M \otimes_R N} & {M \otimes_R N'}
\arrow["{\id_M \otimes g}"', from=2-1, to=2-2]
\arrow["\simeq"', from=1-1, to=2-1]
\arrow["\simeq", from=1-2, to=2-2]
\arrow["{\id_{S \otimes_R M} \otimes g}", from=1-1, to=1-2]
\end{tikzcd}\]
commutes.
The map \( \id_M \otimes g \) is injective as \( M \) is flat, so the map \( \id_{S \otimes_R M} \otimes g \) is also injective.
Thus \( S \otimes_R M \) is a flat \( S \)-module.
\end{proof}

0 comments on commit 15783d6

Please sign in to comment.