The Stacks project

90.11 Tangent spaces of functors

Let $R$ be a ring. We write $\text{Mod}_ R$ for the category of $R$-modules and $\text{Mod}^{fg}_ R$ for the category of finitely generated $R$-modules.

Definition 90.11.1. Let $L: \text{Mod}^{fg}_ R \to \text{Mod}_ R$, resp. $L: \text{Mod}_ R \to \text{Mod}_ R$ be a functor. We say that $L$ is $R$-linear if for every pair of objects $M, N$ of $\text{Mod}^{fg}_ R$, resp. $\text{Mod}_ R$ the map

\[ L : \mathop{\mathrm{Hom}}\nolimits _ R(M, N) \longrightarrow \mathop{\mathrm{Hom}}\nolimits _ R(L(M), L(N)) \]

is a map of $R$-modules.

Remark 90.11.2. One can define the notion of an $R$-linearity for any functor between categories enriched over $\text{Mod}_ R$. We made the definition specifically for functors $L: \text{Mod}^{fg}_ R \to \text{Mod}_ R$ and $L: \text{Mod}_ R \to \text{Mod}_ R$ because these are the cases that we have needed so far.

Remark 90.11.3. If $L: \text{Mod}^{fg}_ R \to \text{Mod}_ R$ is an $R$-linear functor, then $L$ preserves finite products and sends the zero module to the zero module, see Homology, Lemma 12.3.7. On the other hand, if a functor $\text{Mod}^{fg}_ R \to \textit{Sets}$ preserves finite products and sends the zero module to a one element set, then it has a unique lift to a $R$-linear functor, see Lemma 90.11.4.

Lemma 90.11.4. Let $L: \text{Mod}^{fg}_ R \to \textit{Sets}$, resp. $L: \text{Mod}_ R \to \textit{Sets}$ be a functor. Suppose $L(0)$ is a one element set and $L$ preserves finite products. Then there exists a unique $R$-linear functor $\widetilde{L} : \text{Mod}^{fg}_ R \to \text{Mod}_ R$, resp. $\widetilde{L} : \text{Mod}^{fg}_ R \to \text{Mod}_ R$, such that

\[ \vcenter { \xymatrix{ & \text{Mod}_ R \ar[dr]^{\text{forget}} & \\ \text{Mod}^{fg}_ R \ar[ur]^{\widetilde{L}} \ar[rr]^{L} & & \textit{Sets} } } \quad \text{resp.}\quad \vcenter { \xymatrix{ & \text{Mod}_ R \ar[dr]^{\text{forget}} & \\ \text{Mod}_ R \ar[ur]^{\widetilde{L}} \ar[rr]^{L} & & \textit{Sets} } } \]


Proof. We only prove this in case $L: \text{Mod}^{fg}_ R \to \textit{Sets}$. Let $M$ be a finitely generated $R$-module. We define $\widetilde{L}(M)$ to be the set $L(M)$ with the following $R$-module structure.

Multiplication: If $r \in R$, multiplication by $r$ on $L(M)$ is defined to be the map $L(M) \to L(M)$ induced by the multiplication map $r \cdot : M \to M$.

Addition: The sum map $M \times M \to M: (m_1, m_2) \mapsto m_1 + m_2$ induces a map $L(M \times M) \to L(M)$. By assumption $L(M \times M)$ is canonically isomorphic to $L(M) \times L(M)$. Addition on $L(M)$ is defined by the map $L(M) \times L(M) \cong L(M \times M) \to L(M)$.

Zero: There is a unique map $0 \to M$. The zero element of $L(M)$ is the image of $L(0) \to L(M)$.

We omit the verification that this defines an $R$-module $\widetilde{L}(M)$, the unique such that is $R$-linearly functorial in $M$. $\square$

Lemma 90.11.5. Let $L_1, L_2: \text{Mod}^{fg}_ R \to \textit{Sets}$ be functors that take $0$ to a one element set and preserve finite products. Let $t : L_1 \to L_2$ be a morphism of functors. Then $t$ induces a morphism $\widetilde{t} : \widetilde{L}_1 \to \widetilde{L}_2$ between the functors guaranteed by Lemma 90.11.4, which is given simply by $\widetilde{t}_ M = t_ M: \widetilde{L}_1(M) \to \widetilde{L}_2(M)$ for each $M \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ R)$. In other words, $t_ M: \widetilde{L}_1(M) \to \widetilde{L}_2(M)$ is a map of $R$-modules.

Proof. Omitted. $\square$

In the case $R = K$ is a field, a $K$-linear functor $L : \text{Mod}^{fg}_ K \to \text{Mod}_ K$ is determined by its value $L(K)$.

Lemma 90.11.6. Let $K$ be a field. Let $L: \text{Mod}^{fg}_ K \to \text{Mod}_ K$ be a $K$-linear functor. Then $L$ is isomorphic to the functor $L(K) \otimes _ K - : \text{Mod}^{fg}_ K \to \text{Mod}_ K$.

Proof. For $V \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ K)$, the isomorphism $L(K) \otimes _ K V \to L(V)$ is given on pure tensors by $x \otimes v \mapsto L(f_ v)(x)$, where $f_ v: K \to V$ is the $K$-linear map sending $1 \mapsto v$. When $V = K$, this is the isomorphism $L(K) \otimes _ K K \to L(K)$ given by multiplication by $K$. For general $V$, it is an isomorphism by the case $V = K$ and the fact that $L$ commutes with finite products (Remark 90.11.3). $\square$

For a ring $R$ and an $R$-module $M$, let $R[M]$ be the $R$-algebra whose underlying $R$-module is $R \oplus M$ and whose multiplication is given by $(r, m) \cdot (r', m') = (rr', rm' + r'm)$. When $M = R$ this is the ring of dual numbers over $R$, which we denote by $R[\epsilon ]$.

Now let $S$ be a ring and assume $R$ is an $S$-algebra. Then the assignment $M \mapsto R[M]$ determines a functor $\text{Mod}_ R \to S\text{-Alg}/R$, where $S\text{-Alg}/R$ denotes the category of $S$-algebras over $R$. Note that $S\text{-Alg}/R$ admits finite products: if $A_1 \to R$ and $A_2 \to R$ are two objects, then $A_1 \times _ R A_2$ is a product.

Lemma 90.11.7. Let $R$ be an $S$-algebra. Then the functor $\text{Mod}_ R \to S\text{-Alg}/R$ described above preserves finite products.

Proof. This is merely the statement that if $M$ and $N$ are $R$-modules, then the map $R[M \times N] \to R[M] \times _ R R[N]$ is an isomorphism in $S\text{-Alg}/R$. $\square$

Lemma 90.11.8. Let $R$ be an $S$-algebra, and let $\mathcal{C}$ be a strictly full subcategory of $S\text{-Alg}/R$ containing $R[M]$ for all $M \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ R)$. Let $F: \mathcal{C} \to \textit{Sets}$ be a functor. Suppose that $F(R)$ is a one element set and that for any $M, N \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ R)$, the induced map

\[ F(R[M] \times _ R R[N]) \to F(R[M]) \times F(R[N]) \]

is a bijection. Then $F(R[M])$ has a natural $R$-module structure for any $M \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ R)$.

Proof. Note that $R \cong R[0]$ and $R[M] \times _ R R[N] \cong R[M \times N]$ hence $R$ and $R[M] \times _ R R[N]$ are objects of $\mathcal{C}$ by our assumptions on $\mathcal{C}$. Thus the conditions on $F$ make sense. The functor $\text{Mod}_ R \to S\text{-Alg}/R$ of Lemma 90.11.7 restricts to a functor $\text{Mod}^{fg}_ R \to \mathcal{C}$ by the assumption on $\mathcal{C}$. Let $L$ be the composition $\text{Mod}^{fg}_ R \to \mathcal{C} \to \textit{Sets}$, i.e., $L(M) = F(R[M])$. Then $L$ preserves finite products by Lemma 90.11.7 and the assumption on $F$. Hence Lemma 90.11.4 shows that $L(M) = F(R[M])$ has a natural $R$-module structure for any $M \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ R)$. $\square$

Definition 90.11.9. Let $\mathcal{C}$ be a category as in Lemma 90.11.8. Let $F : \mathcal{C} \to \textit{Sets}$ be a functor such that $F(R)$ is a one element set. The tangent space $TF$ of $F$ is $F(R[\epsilon ])$.

When $F : \mathcal{C} \to \textit{Sets}$ satisfies the hypotheses of Lemma 90.11.8, the tangent space $TF$ has a natural $R$-module structure.

Example 90.11.10. Since $\mathcal{C}_\Lambda $ contains all $k[V]$ for finite dimensional vector spaces $V$ we see that Definition 90.11.9 applies with $S = \Lambda $, $R = k$, $\mathcal{C} = \mathcal{C}_\Lambda $, and $F : \mathcal{C}_\Lambda \to \textit{Sets}$ a predeformation functor. The tangent space is $TF = F(k[\epsilon ])$.

Example 90.11.11. Let us work out the tangent space of Example 90.11.10 when $F : \mathcal{C}_\Lambda \to \textit{Sets}$ is a prorepresentable functor, say $F = \underline{S}|_{\mathcal{C}_\Lambda }$ for $S \in \mathop{\mathrm{Ob}}\nolimits (\widehat{\mathcal{C}}_\Lambda )$. Then $F$ commutes with arbitrary limits and thus satisfies the hypotheses of Lemma 90.11.8. We compute

\[ TF = F(k[\epsilon ]) = \mathop{\mathrm{Mor}}\nolimits _{\mathcal{C}_\Lambda }(S, k[\epsilon ]) = \text{Der}_\Lambda (S, k) \]

and more generally for a finite dimensional $k$-vector space $V$ we have

\[ F(k[V]) = \mathop{\mathrm{Mor}}\nolimits _{\mathcal{C}_\Lambda }(S, k[V]) = \text{Der}_\Lambda (S, V). \]

Explicitly, a $\Lambda $-algebra map $f : S \to k[V]$ compatible with the augmentations $q : S \to k$ and $k[V] \to k$ corresponds to the derivation $D$ defined by $s \mapsto f(s) - q(s)$. Conversely, a $\Lambda $-derivation $D : S \to V$ corresponds to $f : S \to k[V]$ in $\mathcal{C}_\Lambda $ defined by the rule $f(s) = q(s) + D(s)$. Since these identifications are functorial we see that the $k$-vector spaces structures on $TF$ and $\text{Der}_\Lambda (S, k)$ correspond (see Lemma 90.11.5). It follows that $\dim _ k TF$ is finite by Lemma 90.4.5.

Example 90.11.12. The computation of Example 90.11.11 simplifies in the classical case. Namely, in this case the tangent space of the functor $F = \underline{S}|_{\mathcal{C}_\Lambda }$ is simply the relative cotangent space of $S$ over $\Lambda $, in a formula $TF = T_{S/\Lambda }$. In fact, this works more generally when the field extension $k/k'$ is separable. See Exercises, Exercise 111.35.2.

Lemma 90.11.13. Let $F, G: \mathcal{C} \to \textit{Sets}$ be functors satisfying the hypotheses of Lemma 90.11.8. Let $t : F \to G$ be a morphism of functors. For any $M \in \mathop{\mathrm{Ob}}\nolimits (\text{Mod}^{fg}_ R)$, the map $t_{R[M]}: F(R[M]) \to G(R[M])$ is a map of $R$-modules, where $F(R[M])$ and $G(R[M])$ are given the $R$-module structure from Lemma 90.11.8. In particular, $t_{R[\epsilon ]} : TF \to TG$ is a map of $R$-modules.

Proof. Follows from Lemma 90.11.5. $\square$

Example 90.11.14. Suppose that $f : R \to S$ is a ring map in $\widehat{\mathcal{C}}_\Lambda $. Set $F = \underline{R}|_{\mathcal{C}_\Lambda }$ and $G = \underline{S}|_{\mathcal{C}_\Lambda }$. The ring map $f$ induces a transformation of functors $G \to F$. By Lemma 90.11.13 we get a $k$-linear map $TG \to TF$. This is the map

\[ TG = \text{Der}_\Lambda (S, k) \longrightarrow \text{Der}_\Lambda (R, k) = TF \]

as follows from the canonical identifications $F(k[V]) = \text{Der}_\Lambda (R, V)$ and $G(k[V]) = \text{Der}_\Lambda (S, V)$ of Example 90.11.11 and the rule for computing the map on tangent spaces.

Lemma 90.11.15. Let $F: \mathcal{C} \to \textit{Sets}$ be a functor satisfying the hypotheses of Lemma 90.11.8. Assume $R = K$ is a field. Then $F(K[V]) \cong TF \otimes _ K V$ for any finite dimensional $K$-vector space $V$.

Proof. Follows from Lemma 90.11.6. $\square$

Comments (0)

Post a comment

Your email address will not be published. Required fields are marked.

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).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.

In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 06I2. Beware of the difference between the letter 'O' and the digit '0'.