Torsion Motivic Complex at Lower Weights

Posted on November 27, 2023 by Peilin Lee

Here are some random thoughts when going through Lecture notes on motivic cohomology.

\(\mu_n\) and \(\mathbb{Z}/n(q)\)

Recall that \[ \mathbb{Z}(1)=\mathbb{Z}_{tr}(\mathbb{G}_m)\cong\mathcal{O}^*[-1] \] as presheaves with transfers. Étale sheafication gives that \[ \mathbb{Z}/l(1)_{ét}\cong\mathcal{O}_{ét}^*[-1]\otimes\mathbb{Z}/l\cong H_0(\mathcal{O}_{ét}^*[-1]\otimes^L\mathbb{Z}/l)=\mu_l. \] Note that \(\mathcal{O}_{ét}^*[-1]\otimes\mathbb{Z}/l\) is indeed quasi-isomorphic to \(\mathcal{O}_{ét}^*[-1]\otimes^L\mathbb{Z}/l\) when \(l\) is prime to the characteristic of our base field \(k\).

Moreover, in proposition 10.6 we see that \(\mu_n^{\otimes q}\rightarrow\mathbb{Z}/n(q)\) is an \(\mathbb{A}^1\)-weak equivalance in \(D^-(Sh_{ét}(Cor_k,\mathbb{Z}/n))\) when \(n\in k^\times\). Here we need some lemmas from the lecture.

[Lemma 8.13] If \(F\) and \(F'\) are étale sheaves sheaves of \(R\)-modules with transfers, and \(F\) is locally constant, then the map\[ F\otimes_{ét}F'\rightarrow F\otimes_{ét}^{tr}F' \] is an isomorphism in \(D^-(Sh_{ét}(Cor_k, R))\).

[Lemma 8.18] Let \(F\) be a locally constant étale sheaf with \(R\)-flat sections. Then the object \(E\otimes_{L,ét}^{tr}F\) is acyclic at all non-zero degrees.

These two lemmas suggest that locally constant étale sheaves with \(R\)-flat sections form a model to compute \(\otimes_{ét}^{tr}\) in \(D^-(Sh_{ét}(Cor_k, R))\). We may formalute this strictly later on.

Now we easily see that

[Proposition 10.5] The morphism \[ \mathbb{Z}/n(1)^{\otimes_L^{tr}q}\rightarrow\mathbb{Z}/n(q) \] is an \(\mathbb{A}^1\)-weak equivalance in \(D^-(Sh_{ét}(Cor_k,\mathbb{Z}/n))\).

Combining the facts above we see that \(\mu_n^{\otimes q}\rightarrow\mathbb{Z}/n(q)\) is indeed an \(\mathbb{A}^1\)-weak equivalance.

Another important fact is that when \(n\in k^\times\), we can invert the Tate twist operation to identify the localized category \(DM_{ét}^{eff,-}(k,\mathbb{Z}/n)\) with \(DM_{ét}^-(k,\mathbb{Z}/n)\).

[Corollary 8.19] Tate twist \(M\mapsto M(1)=M\otimes_{L,ét}^{tr}\mathbb{Z}/n(1)\) is invertible in \(D^-(Sh_{ét}(Cor_k,\mathbb{Z}/n))\) when \(n\in k^\times\).

Proof  Let \(\mu_n^*\) be the Cartier dual of \(\mu_n\) then we have \[ \begin{aligned} \mu_n^*\otimes_{L,ét}^{tr}\mathbb{Z}/n(1)&\cong \mu_n^*\otimes_{ét}^{tr}\mathbb{Z}/n(1)&(8.18)\\ &\cong\mu_n^*\otimes_{ét}\mathbb{Z}/n(1)&(8.13)\\ &\cong\mu_n^*\otimes_{ét}\mu_n&\\ &\cong\mu_n^*\otimes_{\mathbb{Z},ét}\mu_n\otimes_{\mathbb{Z},ét}\mathbb{Z}/n&\\ &\cong\mathbb{Z}/n&(duality) \end{aligned} \]

\(\mathbb{Z}(1)\) and \(\mathcal{O}^*\)

We now have a closer look at the motivic complex \(\mathbb{Z}(1)\). Following lecture 7, for a closed immersion \(i:Y\hookrightarrow\bar{X}\), define the presheaf (a sheaf indeed) of relative units as \[ \mathbb{G}_{\bar{X},Y}=Ker(\mathbb{G}_\bar{X}\rightarrow i_*\mathbb{G}_Y). \] The bar over \(X\) suggests a good model for this should be a pole \(Y\) of \(1\)-dimensional projective geometric variety \(\bar{X}\), which involves the perspection of a relative curve over a noetherian base. We can go quite deep into this, enabling us to easily extend our theory of motives and singular homology to schemes with singulars. This is interesting since Grothendieck’s theory of pure motives is based on the intersection theory of Chow groups, which only behaves well at smooth case, when moving lemma holds. We will discuss this later on.

When \(\bar{X}\) is geometrically connected, denote \(\mathcal{M}^*(\bar{X},Y)\) be the presheaf which associates to a smooth connected scheme \(U/k\) the abelian group of rational functions of \(\bar{X}\times_kU\) which vanishes at \(Y\times_kU\) and remains regular at a (Zariski) neighborhood of \(Y\times_kU\) under multiplcation.

[Remark 7.11] For any connected \(S\), the “stalk” \((i^{-1}\mathbb{G}_{\bar{X}\times S,Y\times S})(Y\times_k S)\) is equal to \(\mathcal{M}^*(\bar{X},Y)(S)\).

Proof  Since \(Y\) is a closed subscheme of \(\bar{X}\), we see that the presheaf pullback \(i_{pre}^{-1}\mathcal{F}\) of any zariski sheaf on \(\bar{X}\) is already a sheaf. Taking the global section of the pullbacked short exact sequence

we see that \(H^0(Y,i^*\mathbb{G}_{\bar{X},Y})\) is just the set \(\{(U\hookrightarrow X,f\in\mathcal{O}^*(U))|f\equiv 1 (\mathcal{O}^*(Y))\}\).

[Statement] \(\mathcal{M}^*(\bar{X},Y)\) is a sheaf for the Zariski topology.

Here we only need the case where \(\bar{X}=\mathbb{P}_k^1\) and \(Y=\{0,\infty\}\).

[Lemma 2.18] Let \(F\) be a presheaf. Then the maps \[ i_0^*,i_1^*:C_*F(X\times\mathbb{A}^1)\rightarrow C_*F(X) \] are chain homotopic.

[Statement] Let \(u:P_*\rightarrow Q_*\) be a morphism of zero homotopy between complex \(P\) and \(Q\). Then the homotopy class \([u(f)]\) defined by any cycle \(f\in P_i\) represents a zero homology class in \(H_i(Q)\).

Proof  Since \(u\cong 0\) in the homotopy category \(K\), the triangle \[ P\rightarrow Q\rightarrow P\oplus Q\rightarrow P[-1] \] is distinguished. Hence we get a long exact sequence of homology groups\[ \cdots\rightarrow H_i(P)\rightarrow H_i(Q)\rightarrow H_i(P)\oplus H_i(Q)\rightarrow\cdots \] the class \([f]\in H_i(P)\) maps to 0 in \(H_i(P)\oplus H_i(Q)\), hence maps to \([u(f)]=0\) in \(H_i(Q)\).

[Lemma 4.6] For \(S/k\) smooth, \(C_*(\mathcal{M}^*(\mathbb{P}^1,\{0,\infty\}))(S)\) is acyclic, hence \(C_*(\mathcal{M}^*(\mathbb{P}^1,\{0,\infty\}))\) is acyclic as complex of sheaves.

Proof  Here we work over the normalized TSP 017V \(\mathcal{M}^*(\mathbb{P}^1,\{0,\infty\})\) chain complex \[ C_*^{DK}(S)=C_*^{DK}(\mathcal{M}^*(\mathbb{P}^1,\{0,\infty\}))(S). \]

For a cycle \(f\in C_i^{DK}(S)\), that is, an element vanishing in \(C_{i-1}^{DK}(S)\), we construct a path \(h_S(f,t)\in C_i^{DK}(S\times\mathbb{A}^1)\subset C_i(S\times\mathbb{A}^1)\) such that the restriction along \(i_0\) and \(i_1\) retracts \(f\) to \(0\). Since \(f\) is a cycle, it represents a regular function on some neightborhood \(U\) of \(Z=S\times\Delta^i\times\{0,\infty\}\) which vanishes on every face \(S\times\Delta^{i-1}\times\{0,\infty\}\) and \(Z\). The path we construct would be \(h_S(f)=1-t(1-f)\), which is obviously regular on the neighborhood \(\mathbb{A}^1\times U\) of \(\mathbb{A}^1\times Z\). Moreover, \(h_S(f)\) is a cycle in \(C_i(S\times\mathbb{A}^1)\) since it equals to \(1\) when \(f\) equals to \(1\). Hence we get the result.

[Lemma 4.3] For all \(f\) in \(\mathcal{M}^*(\mathbb{P}^1,\{0,\infty\})\), denote the Weil divisor associated to \(f\) by \(D(f)\), then \(D(f)\) belongs to the subgroup \(Cor_k(X,\mathbb{A}^1\backslash 0)\) of \(Z^1(\mathbb{P}^1\times_k X)\).