04.03.21 · algebraic-geometry / cohomology

Hochschild-Kostant-Rosenberg theorem

shipped3 tiersLean: none

Anchor (Master): Hochschild-Kostant-Rosenberg *Trans. AMS* 102 (1962); Loday *Cyclic Homology* (Springer Grundlehren 301, 1992; 2nd ed. 1998); Swan *J. Pure Appl. Algebra* 110 (1996); Yekutieli *Math. Res. Lett.* 9 (2002); Caldararu *Adv. Math.* 194 (2003); Kontsevich *Lett. Math. Phys.* 66 (2003) — formality theorem; Tamarkin 1998 / Hinich 1999 — alternative proofs of formality; Avramov-Iyengar on the non-smooth case

Intuition Beginner

A commutative algebra $A = k [x_{1}, \dots, x_{d}] / I$ over a field $k$ models a geometric object: the variety $Spec (A)$ . When this variety is smooth, the local picture at each point looks like a small piece of affine space $k^{d}$ , with a well-behaved tangent space and well-behaved differential forms $d x_{i}$ . The Kähler differentials $Ω_{A / k}^{n}$ are the algebraic version of these differential forms: $Ω_{A / k}^{1}$ is generated by symbols $d a$ for $a \in A$ subject to the Leibniz rule $d (ab) = a d b + b d a$ , and $Ω_{A / k}^{n}$ is its $n$ -th exterior power. This is the classical differential-forms calculus translated into pure algebra, and for a smooth variety it gives exactly the right notion of differential $n$ -forms.

The Hochschild homology $H H_{n} (A, A)$ of an associative algebra $A$ is a more general invariant — it makes sense even when $A$ is noncommutative. It is computed from the bar complex with the alternating-sum differential, and it measures how the multiplication of $A$ interacts with itself in a derived-categorical sense. For a noncommutative algebra, Hochschild homology is a genuine noncommutative invariant; for a commutative algebra, you might wonder whether it reduces to something more classical. The Hochschild-Kostant-Rosenberg theorem (HKR, 1962) gives the precise answer: for a smooth commutative algebra in characteristic zero, $H H_{n} (A, A)$ is canonically the $n$ -th module of Kähler differentials, $H H_{n} (A, A) ≅ Ω_{A / k}^{n}$ , with the identification given by an explicit antisymmetrisation map.

The everyday analogy is a recipe-book metaphor. The Kähler differentials are like a short recipe for the variety using the standard ingredients $d x_{1}, \dots, d x_{d}$ and the wedge-product rules — the classical differential-forms calculus. The Hochschild homology is a much longer audit of every tensor expression one can build out of the algebra's multiplication. The HKR theorem is the statement that for a smooth recipe in characteristic zero, the long audit collapses to the short recipe: the antisymmetrisation map sends a differential form to a signed average of tensor reorderings, and this map identifies the two pictures on homology.

Visual Beginner

A diagram showing on the left the Kähler differentials $Ω_{A / k}^{n}$ for $A$ a smooth commutative $k$ -algebra, with the symbols $a_{0} d a_{1} \land \dots \land d a_{n}$ , and on the right the Hochschild chain complex $C_{n} (A, A)$ (the tensor power of $A$ with itself over $k$ , $n + 1$ factors) with the antisymmetrisation map $ε_{n}$ between them. The HKR theorem is the assertion that this map induces an isomorphism in homology $H H_{n} (A, A) ≅ Ω_{A / k}^{n}$ in characteristic zero with $A$ smooth, with annotations indicating that the proof reduces to the polynomial-algebra case via étale localisation and uses the Koszul resolution.

The picture captures the structural shape: smoothness produces a Koszul resolution for $A$ as a module over its enveloping algebra, and the Koszul resolution makes the antisymmetrisation map a quasi-isomorphism, identifying Hochschild homology with Kähler differentials. A reader who internalises this picture will recognise the template every time HKR appears — smoothness produces a Koszul resolution, Koszul gives antisymmetrisation, antisymmetrisation gives the HKR identification.

Worked example Beginner

Compute the Hochschild homology of the polynomial algebra $A = k [x_{1}, x_{2}]$ in two variables over a field $k$ of characteristic zero and match it with the Kähler differentials $Ω_{A / k}^{*}$ .

Step 1. Write down the Kähler differentials. For $A = k [x_{1}, x_{2}]$ , the Kähler differentials are free $A$ -modules: $Ω_{A / k}^{0} = A = k [x_{1}, x_{2}]$ in degree zero, $Ω_{A / k}^{1} = A \cdot d x_{1} \oplus A \cdot d x_{2}$ free of rank $2$ in degree one, $Ω_{A / k}^{2} = A \cdot d x_{1} \land d x_{2}$ free of rank $1$ in degree two, and $Ω_{A / k}^{n} = 0$ for $n \geq 3$ (since the wedge powers of a rank- $2$ module vanish for $n > 2$ ).

Step 2. Identify the Koszul resolution of $A$ as a module over the enveloping algebra $A^{e}$ (the tensor product of $A$ with itself over $k$ ). Write $A^{e} = k [x_{1}, x_{2}, y_{1}, y_{2}]$ , where each $y_{i}$ stands for the second-factor copy of $x_{i}$ . The kernel of the multiplication map from $A^{e}$ to $A$ is generated by the regular sequence $x_{1} - y_{1}, x_{2} - y_{2}$ . The Koszul resolution is built from the exterior algebra on generators $e_{1}, e_{2}$ over $A^{e}$ , with the Koszul differential sending each $e_{i}$ to $x_{i} - y_{i}$ .

Step 3. Tensor with $A$ over $A^{e}$ to compute Hochschild homology. Under this tensor identification, $A^{e}$ becomes $A$ and the wedge generators $e_{i_{1}} \land \dots \land e_{i_{n}}$ become free generators over $A$ . The Koszul differential sends $e_{i}$ to $x_{i} - y_{i}$ , which becomes multiplication by $x_{i} - x_{i} = 0$ on the commutative algebra $A$ . The tensored complex has zero differential, so $H H_{n} (A, A) = A \cdot \land^{n} (k^{2})$ : namely $H H_{0} (A, A) = A$ , $H H_{1} (A, A) = A \oplus A$ (rank $2$ ), $H H_{2} (A, A) = A$ (rank $1$ ), and $H H_{n} (A, A) = 0$ for $n \geq 3$ .

Step 4. Match with the Kähler differentials via antisymmetrisation. The antisymmetrisation map $ε_{n} : Ω_{A / k}^{n} \to C_{n} (A, A)$ sends each basis differential form to its signed-average tensor expression: the form $d x_{1}$ maps to the tensor $1 \cdot x_{1}$ inside $C_{1}$ and $d x_{2}$ maps to $1 \cdot x_{2}$ , both viewed in the degree-one Hochschild chains; the form $d x_{1} \land d x_{2}$ maps to $\frac{1}{2}$ of the antisymmetric combination of the two ways of writing $x_{1}, x_{2}$ in the degree-two Hochschild chains. The ranks match: $Ω^{0} = A$ matches $H H_{0} = A$ ; $Ω^{1}$ rank $2$ matches $H H_{1}$ rank $2$ ; $Ω^{2}$ rank $1$ matches $H H_{2}$ rank $1$ ; $Ω^{n} = 0$ matches $H H_{n} = 0$ for $n \geq 3$ .

What this tells us. For the polynomial algebra $k [x_{1}, x_{2}]$ , the Hochschild homology recovers exactly the Kähler differentials in every degree, with the antisymmetrisation map providing the explicit isomorphism. This is the HKR theorem in action: the Koszul resolution of $A$ as an $A^{e}$ -module collapses the Hochschild complex to the Kähler differentials, and the smoothness of the polynomial algebra is what makes this collapse exact. The recovery formula $H H_{n} (A, A) = Ω_{A / k}^{n}$ holds for every smooth commutative $k$ -algebra in characteristic zero, with the polynomial case being the local model and the general case following by étale descent.

Check your understanding Beginner

Exercise (easy, multiple choice).

The Hochschild-Kostant-Rosenberg theorem states that for a smooth commutative $k$ -algebra $A$ in characteristic zero, the Hochschild homology $H H_{n} (A, A)$ is canonically isomorphic to which classical invariant?

A. The de Rham cohomology $H_{dR}^{n} (A / k)$ B. The $n$ -th Kähler differential module $Ω_{A / k}^{n}$ C. The $n$ -fold tensor power of $A$ over $k$ as a $k$ -vector space D. The $n$ -th algebraic $K$ -group $K_{n} (A)$

Hint

The HKR isomorphism is given explicitly by the antisymmetrisation map $Ω_{A / k}^{n} \to C_{n} (A, A)$ sending differential forms to signed averages of tensor products. The right side of HKR is the algebraic differential-forms module of the algebra.

Answer

B. HKR identifies $H H_{n} (A, A) ≅ Ω_{A / k}^{n}$ , the $n$ -th module of Kähler differentials, for smooth commutative $A$ in characteristic zero. The antisymmetrisation map sends each basic form $a_{0} d a_{1} \land \dots \land d a_{n}$ to the signed average (over the symmetric group, with normalising factor $1/ n!$ ) of the corresponding tensor expression, and induces the HKR isomorphism on homology. Option A names de Rham cohomology, which is the cohomology of the de Rham complex of forms — related to but distinct from the Kähler differentials. Option C names the bar complex itself, not its homology. Option D names algebraic $K$ -theory, which is related to Hochschild homology via the cyclotomic trace but is a different invariant.

Exercise (easy, true-false).

True or false: the HKR theorem holds for every commutative $k$ -algebra $A$ , with no hypothesis on $A$ or on the characteristic of $k$ .

Hint

The HKR theorem requires two hypotheses: smoothness of $A$ as a $k$ -algebra, and characteristic zero (or at least characteristic large enough relative to the degrees involved). The antisymmetrisation map uses a factor of $1/ n!$ .

Answer

False. The HKR theorem requires smoothness of $A$ as a $k$ -algebra (so that the cotangent complex is concentrated in degree zero and equals the Kähler differentials) and characteristic zero (so that the antisymmetrisation factor $1/ n!$ is well-defined). In positive characteristic $p$ , the antisymmetrisation fails when $p \leq n$ since $1/ n!$ is not defined in $k$ ; the correct replacement involves the derived de Rham complex of Illusie (1971/72). For non-smooth $A$ , the antisymmetrisation map is no longer a quasi-isomorphism, and the cotangent complex (Quillen, Illusie) replaces the Kähler differentials as the correct measurement of the differential-forms structure; Avramov and Iyengar have systematic results on the André-Quillen-cohomology obstruction to HKR in the non-smooth case.

Exercise (easy, multiple choice).

The antisymmetrisation map $ε_{n}$ in the HKR theorem sends the basic differential form $1 \cdot d x_{1} \land d x_{2}$ (for $A = k [x_{1}, x_{2}]$ ) to which tensor expression in $C_{2} (A, A)$ ? (Write the degree-two Hochschild chain $a_{0}, a_{1}, a_{2}$ in terms of the three slots.)

A. The single tensor with slots $1, x_{1}, x_{2}$ B. The unsigned average of the two orderings of $x_{1}, x_{2}$ in the second and third slots (with $1$ in the first slot) C. Half the antisymmetric combination of the two orderings of $x_{1}, x_{2}$ — namely $\frac{1}{2}$ times the difference between the ordering $(1, x_{1}, x_{2})$ and the ordering $(1, x_{2}, x_{1})$ D. Half the symmetric combination of the two orderings of $x_{1}, x_{2}$ — namely $\frac{1}{2}$ times the sum of the ordering $(1, x_{1}, x_{2})$ and the ordering $(1, x_{2}, x_{1})$

Hint

The formula for $ε_{n}$ involves a signed average over the symmetric group $S_{n}$ with the normalising factor $1/ n!$ . For $n = 2$ the symmetric group $S_{2}$ has two elements: the identity (sign $+ 1$ ) and the transposition (sign $- 1$ ).

Answer

C. The antisymmetrisation map sends $1 \cdot d x_{1} \land d x_{2}$ to $\frac{1}{2}$ times the antisymmetric combination of the two orderings, namely $\frac{1}{2}$ times the difference between the ordering $(1, x_{1}, x_{2})$ and the ordering $(1, x_{2}, x_{1})$ in the degree-two Hochschild chain. The formula uses $1/ n! = 1/2$ as the normalising factor, and the signed average runs over $S_{2} = {id, (12)}$ with signs $+ 1$ and $- 1$ respectively, producing the alternating combination. Option A misses the antisymmetrisation entirely. Option B has the wrong sign (it is the symmetrisation, not the antisymmetrisation). Option D combines the wrong sign with the wrong intent (symmetric average scaled by $1/2$ ). The antisymmetric combination is essential because the wedge product on the left is antisymmetric.

Formal definition Intermediate+

Let $k$ be a field of characteristic zero and let $A$ be a commutative $k$ -algebra (associative, with unit). Let $A^{e} := A \otimes_{k} A^{op} = A \otimes_{k} A$ (the second equality uses commutativity). Let $H H_{n} (A, A) := Tor_{n}^{A^{e}} (A, A)$ denote the Hochschild homology of $A$ with coefficients in itself, computed via the bar resolution $B_{∙} (A) \to A$ as in 04.03.20.

Definition (Kähler differentials). The module of Kähler differentials $Ω_{A / k}^{1}$ is the $A$ -module representing $k$ -linear derivations: there is a universal $k$ -linear derivation $d : A \to Ω_{A / k}^{1}$ such that for every $A$ -module $M$ , the map $Hom_{A} (Ω_{A / k}^{1}, M) \to Der_{k} (A, M)$ given by $ϕ \mapsto ϕ \circ d$ is an isomorphism. Concretely, $Ω_{A / k}^{1}$ is the $A$ -module generated by symbols $d a$ for $a \in A$ subject to the relations $d (a + b) = d a + d b$ , $d (λa) = λ d a$ for $λ \in k$ , and the Leibniz rule $d (ab) = a d b + b d a$ . The higher Kähler differential modules are the exterior powers $Ω_{A / k}^{n} := \land_{A}^{n} Ω_{A / k}^{1}$ , equipped with the de Rham differential $d : Ω_{A / k}^{n} \to Ω_{A / k}^{n + 1}$ extending $d : A \to Ω_{A / k}^{1}$ as the unique antiderivation of degree $+ 1$ with $d \circ d = 0$ (see 04.08.01 for the sheaf-theoretic version).

Definition (smooth $k$ -algebra). A commutative $k$ -algebra $A$ is smooth of relative dimension $d$ if $A$ is finitely presented and the Kähler differential module $Ω_{A / k}^{1}$ is a projective $A$ -module of constant rank $d$ . Equivalently, the structure morphism $Spec (A) \to Spec (k)$ is smooth in the sense of schemes (formally smooth and locally of finite presentation), and $Spec (A)$ is locally étale over affine space $A_{k}^{d}$ . Standard examples: $A = k [x_{1}, \dots, x_{d}]$ (the polynomial algebra of dimension $d$ ); any localisation $S^{- 1} A$ of a smooth algebra; any étale extension $A \to B$ .

Definition (antisymmetrisation map). For each $n \geq 0$ , the antisymmetrisation map $ε_{n} : Ω_{A / k}^{n} \to C_{n} (A, A)$ is the $A$ -linear map defined on the standard generators by $$ \varepsilon_n(a_0 , da_1 \wedge \cdots \wedge da_n) := \frac{1}{n!} \sum_{\sigma \in S_n} \mathrm{sgn}(\sigma) , a_0 \otimes a_{\sigma(1)} \otimes \cdots \otimes a_{\sigma(n)}, $$ where $S_{n}$ is the symmetric group on $n$ letters and $C_{n} (A, A) = A \otimes_{k} A^{\otimes n}$ is the degree- $n$ component of the Hochschild chain complex. The factor $1/ n!$ requires characteristic zero (or at least $char k > n$ ); the antisymmetric sum requires that the input be an antisymmetric expression (the wedge $d a_{1} \land \dots \land d a_{n}$ in $Ω_{A / k}^{n}$ ).

Theorem (Hochschild-Kostant-Rosenberg; HKR 1962 Trans. AMS 102). Let $k$ be a field of characteristic zero and let $A$ be a smooth commutative $k$ -algebra of finite type. Then:

(i) The antisymmetrisation map $ε_{n} : Ω_{A / k}^{n} \to C_{n} (A, A)$ is a chain map from the complex $(Ω_{A / k}^{*}, 0)$ (with zero differential, viewed as a chain complex) to the Hochschild complex $(C_{*} (A, A), b)$ .

(ii) The induced map on homology $$ \varepsilon_* : \Omega^n_{A/k} \to HH_n(A, A) $$ is an isomorphism for every $n \geq 0$ .

(iii) Dually, on the cochain side, there is a canonical isomorphism $$ HH^n(A, A) \cong \wedge^n_A \mathrm{Der}_k(A, A), $$ identifying Hochschild cohomology with the algebraic polyvector field module of $A$ .

Definition (polyvector fields and Schouten-Nijenhuis bracket). For $A$ a smooth commutative $k$ -algebra, the polyvector field module is $T_{poly}^{n} (A) := \land_{A}^{n} Der_{k} (A, A)$ in degree $n$ , with total module $T_{poly}^{*} (A) := ⨁_{n} T_{poly}^{n} (A)$ . The Schouten-Nijenhuis bracket $[-, -]_{S N} : T_{poly}^{p} \otimes T_{poly}^{q} \to T_{poly}^{p + q - 1}$ is the unique extension of the Lie bracket of derivations (on $T_{poly}^{1} = Der_{k} (A, A)$ ) to a graded Lie bracket on $T_{poly}^{*}$ satisfying the graded Leibniz rule with respect to the wedge product. The polyvector field algebra $(T_{poly}^{*}, \land, [-, -]_{S N})$ is a Gerstenhaber algebra with zero differential.

Definition (formality theorem; Kontsevich 2003). The Hochschild cochain complex $C^{*} (A, A)$ of a smooth commutative $k$ -algebra $A$ in characteristic zero carries a differential graded Lie algebra structure under the Gerstenhaber bracket. Kontsevich's formality theorem asserts the existence of an $L_{\infty}$ -quasi-isomorphism $U : T_{poly}^{*} (A) \to C^{*} (A, A)$ from the polyvector field algebra (with Schouten-Nijenhuis bracket and zero differential) to the Hochschild cochain complex (with Gerstenhaber bracket and Hochschild differential), lifting the HKR isomorphism on cohomology to a homotopy-coherent equivalence of $L_{\infty}$ -algebras. The theorem solves the deformation-quantisation problem: every Poisson structure on a smooth manifold deforms to an associative star product.

Counterexamples to common slips

The HKR theorem requires characteristic zero. In characteristic $p$ the antisymmetrisation involves $1/ n!$ , which fails when $p \leq n$ ; the correct replacement is the derived de Rham complex of Illusie (1971/72).
The HKR theorem requires $A$ to be smooth. For singular $A$ , the cotangent complex $L_{A / k}$ replaces the Kähler differentials, and HKR is replaced by the Adams decomposition in terms of derived symmetric powers of $L_{A / k}$ (Loday, Quillen, Pirashvili).
The antisymmetrisation map sends the wedge product $d a_{1} \land \dots \land d a_{n}$ to a signed average over $S_{n}$ . Using a tensor product in place of the wedge gives a different map that does not give the HKR isomorphism.
HKR gives an isomorphism $H H_{n} (A, A) ≅ Ω_{A / k}^{n}$ in homology, not as chain complexes: the Hochschild complex has a nonzero differential $b$ , the Kähler differentials have zero differential, and the antisymmetrisation map is a quasi-isomorphism rather than an isomorphism of complexes.
The scheme-theoretic HKR formula $H H_{n} (X) ≅ ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ is a Hodge decomposition indexed by $(p, q)$ with $p - q = n$ . Forgetting the sum and writing $H H_{n} (X) ≅ H^{*} (X, Ω_{X / k}^{n})$ loses the cross-Hodge-bidegree contributions and is incorrect.

Key theorem with proof Intermediate+

Theorem (Hochschild-Kostant-Rosenberg; HKR 1962 Trans. AMS 102, 383-408). Let $k$ be a field of characteristic zero and let $A$ be a smooth commutative $k$ -algebra of finite type. The antisymmetrisation map $$ \varepsilon_n : \Omega^n_{A/k} \to C_n(A, A), \qquad a_0 , da_1 \wedge \cdots \wedge da_n \mapsto \frac{1}{n!} \sum_{\sigma \in S_n} \mathrm{sgn}(\sigma) , a_0 \otimes a_{\sigma(1)} \otimes \cdots \otimes a_{\sigma(n)} $$ is a quasi-isomorphism of chain complexes $(\Omega^{A/k}, 0) \to (C(A, A), b)$, inducing canonical isomorphisms $$ HH_n(A, A) \xrightarrow{\sim} \Omega^n_{A/k}, \qquad HH^n(A, A) \xrightarrow{\sim} \wedge^n_A \mathrm{Der}_k(A, A) $$ for all $n \geq 0$ .

Proof. The proof proceeds in four steps. Step 1 verifies that the antisymmetrisation map is a well-defined chain map. Step 2 reduces the general smooth case to the polynomial-algebra case via étale localisation. Step 3 establishes the polynomial-algebra case by explicit Koszul computation. Step 4 derives the cohomology statement by duality.

Step 1 — Antisymmetrisation is a chain map. The complex $Ω_{A / k}^{*}$ on the left has zero differential (regarded as a chain complex). The Hochschild differential $b$ on $C_{*} (A, A)$ is the alternating sum $b = \sum_{i = 0}^{n} (- 1)^{i} d_{i}$ of face maps. Direct computation: applying $b$ to $ε_{n} (a_{0} d a_{1} \land \dots \land d a_{n}) = (1/ n!) \sum_{σ} sgn (σ) a_{0} \otimes a_{σ (1)} \otimes \dots \otimes a_{σ (n)}$ produces terms of three types: (a) the leading term $a_{0} a_{σ (1)} \otimes a_{σ (2)} \otimes \dots$ , (b) interior multiplications $a_{0} \otimes \dots \otimes a_{σ (i)} a_{σ (i + 1)} \otimes \dots$ with sign $(- 1)^{i}$ , and (c) the trailing cyclic term $(- 1)^{n} a_{σ (n)} a_{0} \otimes a_{σ (1)} \otimes \dots \otimes a_{σ (n - 1)}$ . After summing over $σ \in S_{n}$ with signs $sgn (σ)$ and using commutativity of $A$ , each interior multiplication $a_{σ (i)} a_{σ (i + 1)}$ pairs with the transposed permutation $σ^{'} = σ \cdot (i i + 1)$ to give a cancellation: the two permutations $σ, σ^{'}$ contribute terms with opposite signs that differ only in the order of $a_{σ (i)}$ and $a_{σ (i + 1)}$ ; commutativity makes the products equal, and the opposite signs make them cancel. The leading and trailing terms similarly cancel via the cyclic action on $S_{n}$ . Hence $b (ε_{n} (ω)) = 0$ for all $ω \in Ω_{A / k}^{n}$ , and $ε$ is a chain map.

Step 2 — Reduction to the polynomial algebra via étale descent. Both Hochschild homology and Kähler differentials commute with localisation at a multiplicative set $S \subset A$ : $H H_{n} (S^{- 1} A, S^{- 1} A) = S^{- 1} H H_{n} (A, A)$ (because Tor commutes with flat localisation and $S^{- 1} A^{e} = (S^{- 1} A)^{e}$ ) and $Ω_{S^{- 1} A / k}^{n} = S^{- 1} Ω_{A / k}^{n}$ (a standard universal-property computation). More generally, both Hochschild homology and Kähler differentials behave well under étale base change. By smoothness of $A$ , every closed point $m \subset A$ has an étale neighbourhood $A \to A^{'}$ such that $A^{'}$ is étale over a polynomial algebra $k [x_{1}, \dots, x_{d}]$ (the standard local-structure theorem for smooth schemes; see Stacks Project tag 039P for the affine version). Hence if HKR holds for the polynomial algebra $k [x_{1}, \dots, x_{d}]$ , it holds for the étale neighbourhood $A^{'}$ by étale base change, and then for the original $A$ by gluing over a Zariski cover.

Step 3 — Polynomial case via Koszul resolution. Let $A = k [x_{1}, \dots, x_{d}]$ and $A^{e} = A \otimes_{k} A = k [x_{1}, \dots, x_{d}, y_{1}, \dots, y_{d}]$ where $y_{i} := 1 \otimes x_{i}$ . The kernel of the multiplication map $μ : A^{e} \to A$ is the ideal $I = (x_{1} - y_{1}, \dots, x_{d} - y_{d})$ , generated by the regular sequence ${x_{i} - y_{i}}$ . The Koszul resolution of $A$ as an $A^{e}$ -module is the complex $$ K_\bullet = \wedge^*_{A^e}\left(\bigoplus_{i=1}^{d} A^e \cdot e_i\right), \qquad \partial(e_i) = x_i - y_i, $$ with the Koszul differential extending $\partial (e_{i}) = x_{i} - y_{i}$ as an antiderivation. By the regularity of the sequence ${x_{i} - y_{i}}$ , the Koszul complex $K_{∙} \to A$ is a free $A^{e}$ -resolution of $A$ of length $d$ . Each $K_{n} = \land_{A^{e}}^{n} (A^{e} \cdot e_{1} \oplus \dots \oplus A^{e} \cdot e_{d})$ is free of rank $(n d)$ over $A^{e}$ .

Tensoring $K_{∙}$ with $A$ over $A^{e}$ : under the identification $A \otimes_{A^{e}} (A^{e} \cdot e_{i_{1}} \land \dots \land e_{i_{n}}) = A \cdot e_{i_{1}} \land \dots \land e_{i_{n}}$ , the Koszul differential $\partial (e_{i}) = x_{i} - y_{i}$ acts on $A = A \otimes_{A^{e}} A^{e}$ as multiplication by $x_{i} - x_{i} = 0$ (since $y_{i}$ acts as right multiplication by $x_{i}$ , which equals left multiplication on the commutative algebra $A$ ). The differential vanishes after the tensor, so $A \otimes_{A^{e}} K_{∙}$ is the complex with $A \otimes_{k} \land^{n} (k^{d})$ in degree $n$ and zero differential. Reading off homology: $$ HH_n(A, A) = \mathrm{Tor}n^{A^e}(A, A) = A \otimes_k \wedge^n(k^d) = \Omega^n{A/k}, $$ where the final equality uses that $Ω_{A / k}^{1} = A \cdot d x_{1} \oplus \dots \oplus A \cdot d x_{d}$ is the free $A$ -module on the symbols $d x_{i}$ and $Ω_{A / k}^{n} = \land_{A}^{n} Ω_{A / k}^{1}$ .

The antisymmetrisation map $ε_{n}$ realises this isomorphism explicitly: the standard generator $d x_{i_{1}} \land \dots \land d x_{i_{n}} \in Ω_{A / k}^{n}$ corresponds to the basis element $e_{i_{1}} \land \dots \land e_{i_{n}}$ in $A \otimes_{A^{e}} K_{n}$ , and a direct comparison of the Koszul resolution with the bar resolution (the comparison map between two projective resolutions, unique up to chain homotopy) shows that $ε$ is precisely this comparison map at the level of cycles. The signed average over $S_{n}$ in the formula for $ε$ is the explicit description of the comparison map in terms of the bar-complex generators.

Step 4 — Cohomology statement by duality. For the cohomology side $H H^{n} (A, A) ≅ \land_{A}^{n} Der_{k} (A, A)$ , observe that for smooth $A$ , the universal property of Kähler differentials gives $Der_{k} (A, A) = Hom_{A} (Ω_{A / k}^{1}, A)$ , and since $Ω_{A / k}^{1}$ is projective (in fact locally free) of finite rank, the natural map $\land_{A}^{n} Hom_{A} (Ω_{A / k}^{1}, A) \to Hom_{A} (\land_{A}^{n} Ω_{A / k}^{1}, A) = Hom_{A} (Ω_{A / k}^{n}, A)$ is an isomorphism. The Hochschild cohomology $H H^{n} (A, A) = Ext_{A^{e}}^{n} (A, A)$ is computed by dualising the Koszul resolution: $Hom_{A^{e}} (K_{n}, A)$ identifies with $Hom_{A} (\land_{A}^{n} (k^{d}), A) = Hom_{A} (Ω_{A / k}^{n}, A) = \land_{A}^{n} Der_{k} (A, A)$ after the same calculation, and the differential vanishes by the same Koszul-style argument. Reading off cohomology: $$ HH^n(A, A) = \wedge^n_A \mathrm{Der}k(A, A) = T^n{\mathrm{poly}}(A), $$ the polyvector field module of $A$ . $□$

Bridge. The HKR theorem builds the bridge from the noncommutative Hochschild framework back to the classical commutative-algebra differential-forms picture, and the foundational reason the construction works is the Koszul resolution of $A$ as a module over its enveloping algebra $A^{e} = A \otimes_{k} A$ : for a smooth commutative algebra, the diagonal ideal $ker (μ : A^{e} \to A)$ is locally generated by a regular sequence, so the Koszul complex provides a finite free resolution of $A$ as an $A^{e}$ -module, and tensoring with $A$ collapses the differential to zero because of commutativity. The bridge is the antisymmetrisation map $ε : Ω_{A / k}^{*} \to C_{*} (A, A)$ , which realises the comparison between the Koszul resolution and the bar resolution at the level of cycles, and the smoothness hypothesis is what makes this comparison a quasi-isomorphism. Putting these together, the HKR theorem identifies the Hochschild homology of a smooth commutative algebra with its Kähler differentials and the Hochschild cohomology with its polyvector fields, packaging the classical differential-forms calculus and the polyvector-field calculus as homotopy invariants of the algebra.

This pattern appears again in 04.03.20 (Hochschild homology and cohomology), where HKR is one of the central computations of the Hochschild framework; in 04.08.01 (sheaf of differentials), where the Kähler differentials are constructed via the universal-property formalism that HKR reuses; and forward in 04.03.22 (cyclic homology), where the Connes SBI long exact sequence combines with HKR to identify the cyclic homology of a smooth commutative algebra in characteristic zero with its algebraic de Rham cohomology. The central insight is that for smooth commutative algebras in characteristic zero, the Hochschild and cyclic invariants reduce to the classical Kähler-differentials and de Rham-cohomology invariants — and that the bridge between the noncommutative and commutative worlds is the antisymmetrisation map, made possible by the Koszul-style structure of smooth algebras over their enveloping algebras.

Exercises Intermediate+

Exercise 1 (easy, short-answer).

Verify the HKR computation for $A = k [x]$ in one variable over a field $k$ of characteristic zero: $H H_{0} (A, A) = k [x]$ , $H H_{1} (A, A) = k [x] \cdot d x$ , $H H_{n} (A, A) = 0$ for $n \geq 2$ . Identify each Hochschild group with the corresponding Kähler differential module.

Hint

Use the length-one Koszul resolution of $k [x]$ as a $k [x] \otimes k [x]$ -module by the regular element $x \otimes 1 - 1 \otimes x$ , then tensor with $k [x]$ over $k [x]^{e}$ .

Answer

For $A = k [x]$ , the Koszul resolution of $A$ as $A^{e} = k [x] \otimes k [x] = k [x, y]$ -module is the length-one complex $0 \to A^{e} x - y A^{e} \to A \to 0$ , with augmentation by multiplication. Tensoring with $A$ over $A^{e}$ gives the two-term complex $A 0 A$ (with zero differential, since $(x - y)$ acts as multiplication by $x - x = 0$ on the commutative algebra $A$ ). Reading off homology: $H H_{0} (k [x], k [x]) = A = k [x]$ and $H H_{1} (k [x], k [x]) = A = k [x]$ ; all higher Hochschild groups vanish since the resolution has length one.

Match with Kähler differentials: $Ω_{k [x] / k}^{0} = k [x]$ matches $H H_{0} = k [x]$ tautologically (the antisymmetrisation map in degree zero is the identity). $Ω_{k [x] / k}^{1} = k [x] \cdot d x$ (a free rank-one $A$ -module generated by $d x$ ) matches $H H_{1} = k [x]$ via the antisymmetrisation map $ε_{1} (a_{0} \cdot d a_{1}) = a_{0} \otimes a_{1}$ ; setting $a_{0} = 1, a_{1} = x$ identifies the generator $d x \in Ω^{1}$ with the generator $1 \otimes x \in H H_{1}$ . $Ω_{k [x] / k}^{n} = 0$ for $n \geq 2$ (the wedge powers of a rank-one module vanish) matches $H H_{n} = 0$ . Rubric: full credit for the Koszul resolution, the cohomology computation, and the HKR identification in every degree.

Exercise 2 (easy, short-answer).

Compute the antisymmetrisation map $ε_{2} (1 \cdot d x \land d y) \in C_{2} (k [x, y], k [x, y])$ for $A = k [x, y]$ , and verify directly that it is closed under the Hochschild differential $b$ .

Hint

The antisymmetrisation formula gives $ε_{2} (1 \cdot d x \land d y) = \frac{1}{2} (1 \otimes x \otimes y - 1 \otimes y \otimes x)$ . The Hochschild differential $b$ on a $2$ -chain $a_{0} \otimes a_{1} \otimes a_{2}$ is $a_{0} a_{1} \otimes a_{2} - a_{0} \otimes a_{1} a_{2} + a_{2} a_{0} \otimes a_{1}$ .

Answer

By the antisymmetrisation formula, $ε_{2} (1 \cdot d x \land d y) = \frac{1}{2} (1 \otimes x \otimes y - 1 \otimes y \otimes x)$ . Applying the Hochschild differential: $$ b(1 \otimes x \otimes y) = (1 \cdot x) \otimes y - 1 \otimes (x \cdot y) + (y \cdot 1) \otimes x = x \otimes y - 1 \otimes xy + y \otimes x. $$ $$ b(1 \otimes y \otimes x) = (1 \cdot y) \otimes x - 1 \otimes (y \cdot x) + (x \cdot 1) \otimes y = y \otimes x - 1 \otimes yx + x \otimes y. $$ Using commutativity of $k [x, y]$ ( $x y = y x$ ): $$ b(\varepsilon_2(1 \cdot dx \wedge dy)) = \tfrac{1}{2}\left[ (x \otimes y - 1 \otimes xy + y \otimes x) - (y \otimes x - 1 \otimes yx + x \otimes y) \right] = \tfrac{1}{2}[0] = 0. $$ The two terms cancel exactly because of commutativity of $A$ : $1 \otimes x y = 1 \otimes y x$ , and the leading and trailing terms cancel because of the antisymmetric combination. This is the step-by-step verification of Step 1 of the HKR proof in a concrete case. Rubric: full credit for the antisymmetrisation computation, the differential expansion, and the use of commutativity to obtain zero.

Exercise 3 (medium, short-answer).

For $A = k [x_{1}, \dots, x_{d}]$ in $d$ variables over a field $k$ of characteristic zero, compute $H H_{n} (A, A)$ via the HKR identification with $Ω_{A / k}^{n}$ , and determine the rank of $H H_{n} (A, A)$ as a free $A$ -module for each $n$ .

Hint

The Kähler differentials of a polynomial algebra are $Ω_{A / k}^{n} = A \otimes_{k} \land^{n} (k^{d})$ , the $n$ -th exterior power of the free $k$ -module $k \cdot d x_{1} \oplus \dots \oplus k \cdot d x_{d}$ tensored back into an $A$ -module. The rank of the $n$ -th exterior power of a free $k$ -module of rank $d$ is $(n d)$ .

Answer

By HKR, $H H_{n} (k [x_{1}, \dots, x_{d}], k [x_{1}, \dots, x_{d}]) ≅ Ω_{k [x_{1}, \dots, x_{d}] / k}^{n} = A \otimes_{k} \land^{n} (k^{d}) = A^{(n d)}$ , free of rank $(n d)$ as an $A$ -module. The standard basis consists of the wedge products $d x_{i_{1}} \land \dots \land d x_{i_{n}}$ for $1 \leq i_{1} < i_{2} < \dots < i_{n} \leq d$ , and there are $(n d)$ such indices. In particular:

$H H_{0} (A, A) = A$ of rank $(0 d) = 1$ ;
$H H_{1} (A, A) = A^{d}$ of rank $(1 d) = d$ ;
$H H_{n} (A, A) = A^{(n d)}$ for $0 \leq n \leq d$ ;
$H H_{d} (A, A) = A$ of rank $(d d) = 1$ ;
$H H_{n} (A, A) = 0$ for $n > d$ (since the wedge powers of a rank- $d$ module vanish in higher degree).

The total Poincaré polynomial $\sum_{n} rank H H_{n} (A, A) \cdot t^{n} = \sum_{n} (n d) t^{n} = (1 + t)^{d}$ records the entire dimensional information. Rubric: full credit for the HKR identification, the explicit basis, the binomial-coefficient rank, and the Poincaré polynomial.

Exercise 4 (medium, short-answer).

Compute the Hochschild cohomology $H H^{*} (k [x_{1}, x_{2}], k [x_{1}, x_{2}])$ over a field $k$ of characteristic zero using HKR, and identify the cup product and Gerstenhaber-bracket structure on $H H^{*}$ in terms of the polyvector fields $\land^{*} Der_{k} (A, A)$ .

Hint

Apply HKR dually: $H H^{n} (A, A) ≅ \land_{A}^{n} Der_{k} (A, A)$ . For $A = k [x_{1}, x_{2}]$ , $Der_{k} (A, A) = A \cdot \partial_{x_{1}} \oplus A \cdot \partial_{x_{2}}$ , free of rank two. The cup product on $H H^{*}$ corresponds to the wedge product on $\land^{*} Der$ , and the Gerstenhaber bracket corresponds to the Schouten-Nijenhuis bracket.

Answer

By HKR (cohomology side), $H H^{n} (k [x_{1}, x_{2}], k [x_{1}, x_{2}]) = \land_{A}^{n} Der_{k} (A, A)$ for $A = k [x_{1}, x_{2}]$ . Compute $Der_{k} (A, A)$ : a $k$ -linear derivation of $A$ is uniquely determined by its values on $x_{1}$ and $x_{2}$ by the Leibniz rule, so $Der_{k} (A, A) = A \cdot \partial_{x_{1}} \oplus A \cdot \partial_{x_{2}}$ , free of rank two over $A$ . The exterior powers:

$H H^{0} = A$ (the centre, all of $A$ since $A$ is commutative);
$H H^{1} = A \cdot \partial_{x_{1}} \oplus A \cdot \partial_{x_{2}}$ (vector fields = derivations);
$H H^{2} = A \cdot \partial_{x_{1}} \land \partial_{x_{2}}$ (bivector fields, rank one);
$H H^{n} = 0$ for $n \geq 3$ .

Cup product. Under HKR, the cup product on $H H^{*}$ corresponds to the wedge product on polyvector fields: $H H^{p} \otimes H H^{q} \to H H^{p + q}$ becomes $\land^{p} T \otimes \land^{q} T \to \land^{p + q} T$ where $T = Der_{k} (A, A)$ . For $A = k [x_{1}, x_{2}]$ : $\partial_{x_{1}} ⌣ \partial_{x_{2}} = \partial_{x_{1}} \land \partial_{x_{2}}$ generates $H H^{2}$ , and all higher cups vanish by the dimension constraint.

Gerstenhaber bracket. Under HKR, the Gerstenhaber bracket on $H H^{*+ 1}$ corresponds to the Schouten-Nijenhuis bracket on polyvector fields, which extends the Lie bracket of vector fields to higher polyvectors via the graded Leibniz rule. For $A = k [x_{1}, x_{2}]$ : $[\partial_{x_{i}}, \partial_{x_{j}}]_{S N} = [\partial_{x_{i}}, \partial_{x_{j}}] = 0$ (the standard partial derivatives commute), and more generally for a polynomial-coefficient vector field $f \cdot \partial_{x_{i}}$ the bracket is $[f \partial_{x_{i}}, g \partial_{x_{j}}]_{S N} = f \partial_{x_{i}} (g) \partial_{x_{j}} - g \partial_{x_{j}} (f) \partial_{x_{i}}$ .

Rubric: full credit for the HKR identification, the polyvector-field computation, the cup-product structure (wedge), and the Gerstenhaber-Schouten-Nijenhuis correspondence.

Exercise 5 (medium, short-answer).

State the scheme-theoretic HKR formula for the Hochschild homology of a smooth $k$ -scheme $X$ in characteristic zero, and use it to compute $H H_{n} (X)$ for $X = P_{k}^{1}$ , the projective line over a field $k$ of characteristic zero.

Hint

The scheme-theoretic HKR formula is $H H_{n} (X) ≅ ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ . For $X = P_{k}^{1}$ : $H^{0} (P^{1}, O) = k$ , $H^{1} (P^{1}, O) = 0$ , $H^{0} (P^{1}, Ω^{1}) = 0$ , $H^{1} (P^{1}, Ω^{1}) = k$ .

Answer

The scheme-theoretic HKR formula (Swan 1996, Yekutieli 2002) reads $$ HH_n(X) \cong \bigoplus_{p - q = n} H^p(X, \Omega^q_{X/k}) $$ for $X$ a smooth $k$ -scheme of finite type in characteristic zero, where the right side is the Hodge cohomology of $X$ . The formula identifies Hochschild homology of a smooth scheme with the Hodge cohomology, packaging all the differential-forms cohomology data into the Hochschild invariant.

For $X = P_{k}^{1}$ : the cohomology of $O_{P^{1}}$ and $Ω_{P^{1} / k}^{1}$ are $$ H^0(\mathbb{P}^1, \mathcal{O}{\mathbb{P}^1}) = k, \quad H^1(\mathbb{P}^1, \mathcal{O}{\mathbb{P}^1}) = 0, $$ $$ H^0(\mathbb{P}^1, \Omega^1_{\mathbb{P}^1/k}) = 0, \quad H^1(\mathbb{P}^1, \Omega^1_{\mathbb{P}^1/k}) = k $$ (the last by Serre duality, since $Ω_{P^{1}}^{1} = O (- 2)$ and $H^{1} (P^{1}, O (- 2)) = k$ ). Summing over $p - q = n$ : $$ HH_0(\mathbb{P}^1) = H^0(\mathcal{O}) \oplus H^1(\Omega^1) = k \oplus k = k^2, $$ $$ HH_1(\mathbb{P}^1) = H^0(\Omega^1) \oplus H^1(\mathcal{O}) = 0 \oplus 0 = 0, $$ $$ HH_{-1}(\mathbb{P}^1) = H^1(\mathcal{O}) = 0, $$ $$ HH_n(\mathbb{P}^1) = 0 \text{ for } |n| \ge 2. $$ The total Hochschild homology dimension is $\sum_{n} dim H H_{n} (P^{1}) = 2 = χ (P^{1}, O) + χ (P^{1}, Ω^{1})$ , matching the Hodge numbers of $P^{1}$ . Rubric: full credit for the scheme-theoretic HKR formula, the Hodge cohomology computation for $P^{1}$ , and the explicit Hochschild groups.

Exercise 6 (medium, short-answer).

Explain why the HKR theorem fails in positive characteristic. Give a specific counterexample using the antisymmetrisation map for the polynomial algebra $A = F_{p} [x]$ over a field of characteristic $p$ and degree $n = p$ .

Hint

The antisymmetrisation formula involves the factor $1/ n!$ . In characteristic $p$ , $n!$ is zero in the field when $p \leq n$ , so $1/ n!$ is undefined.

Answer

The antisymmetrisation map $ε_{n} (a_{0} d a_{1} \land \dots \land d a_{n}) = (1/ n!) \sum_{σ} sgn (σ) a_{0} \otimes a_{σ (1)} \otimes \dots \otimes a_{σ (n)}$ requires division by $n!$ in the base field $k$ . For $A = F_{p} [x]$ with $char k = p$ and $n = p$ , the factor $1/ n! = 1/ p! \in F_{p}$ is undefined: $p! \equiv 0 (mod p)$ by Wilson-like factorial arithmetic ( $p!$ contains the factor $p$ , hence is $0$ in $F_{p}$ ). The formula for $ε_{n}$ does not even make sense in this case.

More structurally: even after rescaling away the $1/ n!$ factor (e.g., working with the unnormalised antisymmetrisation $\tilde{ε}_{n} = n! \cdot ε_{n}$ ), the resulting map is not a quasi-isomorphism in positive characteristic. The correct replacement is the derived de Rham complex $L Ω_{A / k}^{*}$ of Illusie (Complexe cotangent et déformations, Springer LNM 239 (1971) and LNM 283 (1972)), which equals $Ω_{A / k}^{*}$ for smooth $A$ in characteristic zero but differs from it in positive characteristic. The Hochschild-Hodge identification in positive characteristic involves the conjugate filtration on $L Ω^{*}$ and the Frobenius (Bhatt 2012 Compos. Math. 148, Antieau-Bhatt-Mathew 2020, Bhatt-Morrow-Scholze 2019).

Concretely for $A = F_{p} [x]$ with $n = p$ : the Hochschild homology $H H_{p} (F_{p} [x], F_{p} [x])$ is not isomorphic to $Ω_{F_{p} [x] / F_{p}}^{p} = 0$ ; it contains additional classes coming from the Frobenius twist and the failure of the antisymmetrisation. The correct statement uses the André-Quillen homology $D_{n} (A / k; A)$ in place of $Ω_{A / k}^{n}$ , which equals $Ω^{n}$ for smooth $A$ in characteristic zero and corrects the formula in positive characteristic. Rubric: full credit for the $1/ n!$ obstruction, the derived-de-Rham replacement, and the specific failure for $n = p$ .

Exercise 7 (hard, short-answer).

Show that under the HKR isomorphism $H H^{n} (A, A) ≅ \land_{A}^{n} Der_{k} (A, A)$ for $A$ smooth commutative in characteristic zero, the Gerstenhaber bracket on $H H^{*} (A, A)$ corresponds to the Schouten-Nijenhuis bracket on polyvector fields. Use this to identify the deformation-theoretic content of $H H^{2} (A, A)$ for a smooth commutative algebra with the set of Poisson structures on $A$ .

Hint

The Schouten-Nijenhuis bracket on $T^{*} = \land^{*} Der$ is the unique extension of the Lie bracket of vector fields satisfying the graded Leibniz rule with respect to the wedge product. A bivector $π \in T^{2} = \land^{2} Der$ defines a Poisson structure iff $[π, π]_{S N} = 0$ .

Answer

By HKR, $H H^{n} (A, A) ≅ T_{poly}^{n} (A) := \land_{A}^{n} Der_{k} (A, A)$ , the polyvector field module of $A$ . The Gerstenhaber bracket on $H H^{*} (A, A)$ is the canonical bracket of degree $- 1$ that, combined with the cup product, makes $H H^{*}$ into a Gerstenhaber algebra. The Schouten-Nijenhuis bracket $[-, -]_{S N} : T_{poly}^{p} \otimes T_{poly}^{q} \to T_{poly}^{p + q - 1}$ is the unique extension of the Lie bracket of derivations (on $T^{1} = Der_{k}$ ) to a graded Lie bracket on $T_{poly}^{*}$ satisfying the graded Leibniz rule $$ [X, Y \wedge Z]{SN} = [X, Y]{SN} \wedge Z + (-1)^{(p-1)q} Y \wedge [X, Z]{SN} $$ for $X \in T^{p}$ , $Y \in T^{q}$ , $Z \in T^{r}$ . The compatibility of the Gerstenhaber bracket on $H H^{*} (A, A)$ with the Schouten-Nijenhuis bracket on $T^*{\mathrm{poly}}(A) $u n d er t h eH K R i so m or p hi s m f o l l o w s f r o m t h ee x pl i c i t f or m u l a f or b o t hb r a c k e t s in t er m so f in ser t i o n s, an df r o m t h e a g r ee m e n t o f b o t h w i t h t h eco mm u t a t or o f v ec t or f i e l d so n t h e d e g r ee -$ 1 $p i ece$ T^1 = \mathrm{Der}_k(A, A)$.

Deformation-theoretic interpretation. The general principle: a degree- $2$ class $η \in H H^{2} (A, A)$ is an infinitesimal deformation of the multiplication of $A$ , with the obstruction to extending to a second-order deformation being $[η, η] \in H H^{3} (A, A)$ (Gerstenhaber 1964; see 04.03.20).

Specialising to the smooth commutative case via HKR: $H H^{2} (A, A) = T_{poly}^{2} (A) = \land_{A}^{2} Der_{k} (A, A)$ , the module of bivector fields on $Spec (A)$ . An infinitesimal deformation is a bivector $π = \sum_{i < j} π_{ij} \partial_{x_{i}} \land \partial_{x_{j}}$ , and the obstruction $[π, π]_{S N} \in T_{poly}^{3} (A) = \land_{A}^{3} Der_{k} (A, A)$ to extending the deformation to higher order is exactly the Schouten-Nijenhuis bracket of $π$ with itself. The condition $[π, π]_{S N} = 0$ is the algebraic Jacobi identity for the bracket ${f, g} := π (df, d g) = \sum_{i, j} π_{ij} (\partial_{x_{i}} f) (\partial_{x_{j}} g)$ , which is exactly the Poisson condition on $π$ .

Hence the deformation-theoretic content of $H H^{2} (A, A)$ for a smooth commutative algebra $A$ in characteristic zero is: $H H^{2} (A, A)$ classifies infinitesimal deformations of the multiplication, and the second-order obstruction $[π, π]_{S N} = 0$ is exactly the Poisson condition on the corresponding bivector. Bivectors $π$ satisfying $[π, π]_{S N} = 0$ are Poisson structures on $Spec (A)$ , and the deformation-theoretic problem of extending the infinitesimal deformation to all orders is the deformation-quantisation problem solved by Kontsevich's formality theorem (2003 Lett. Math. Phys. 66): every Poisson bivector $π$ extends to an associative star product $⋆ : A [[ℏ]] \otimes A [[ℏ]] \to A [[ℏ]]$ deforming the original multiplication.

Rubric: full credit for the Schouten-Nijenhuis correspondence, the Poisson-condition identification with $[π, π]_{S N} = 0$ , and the link to Kontsevich's deformation quantisation.

Exercise 8 (hard, short-answer).

State the scheme-theoretic HKR theorem (Swan 1996, Yekutieli 2002) and compute $H H_{n} (X)$ for $X = E$ an elliptic curve over a field $k$ of characteristic zero. Identify the contributions from each Hodge bidegree.

Hint

For $X = E$ an elliptic curve: $dim H^{0} (E, O) = 1$ , $dim H^{1} (E, O) = 1$ , $dim H^{0} (E, Ω^{1}) = 1$ , $dim H^{1} (E, Ω^{1}) = 1$ . All four Hodge numbers equal $1$ .

Answer

The scheme-theoretic HKR theorem (Swan 1996, Yekutieli 2002) states that for a smooth quasiprojective $k$ -scheme $X$ of finite type in characteristic zero, there is a canonical isomorphism $$ HH_n(X) \cong \bigoplus_{p - q = n} H^p(X, \Omega^q_{X/k}) $$ identifying the Hochschild homology of $X$ (defined as $H H_{n} (X) = Ext_{D (X \times_{k} X)}^{- n} (Δ_{*} O_{X}, Δ_{*} O_{X})$ , with $Δ : X \to X \times_{k} X$ the diagonal) with the Hodge cohomology of $X$ . The proof reduces to the affine HKR theorem via local-to-global arguments and the standard derived-category formalism on $X \times X$ .

For $X = E$ an elliptic curve over $k$ in characteristic zero: the Hodge numbers $h^{p, q} = dim_{k} H^{p} (E, Ω_{E}^{q})$ are $$ h^{0, 0} = 1 \quad (H^0(E, \mathcal{O}) = k), \qquad h^{1, 0} = 1 \quad (H^1(E, \mathcal{O}) = k \text{ by Serre duality}), $$ $$ h^{0, 1} = 1 \quad (H^0(E, \Omega^1) = k, \text{ the holomorphic differential}), \qquad h^{1, 1} = 1 \quad (H^1(E, \Omega^1) = k \text{ by Serre duality}). $$ All four Hodge numbers equal $1$ , consistent with $E$ being a Calabi-Yau variety of dimension $1$ with canonical bundle $Ω_{E}^{1} ≅ O_{E}$ (the canonical bundle is the structure sheaf).

Applying scheme-theoretic HKR: $$ HH_0(E) = H^0(E, \mathcal{O}) \oplus H^1(E, \Omega^1) = k \oplus k = k^2, $$ $$ HH_1(E) = H^0(E, \Omega^1) \oplus H^1(E, \mathcal{O}) = k \oplus k = k^2, $$ $$ HH_{-1}(E) = H^1(E, \mathcal{O}) \oplus H^0(E, \Omega^1) \cdot (\text{wait — only } p - q = -1) = 0 \quad \text{(since } p, q \ge 0 \text{ and } p - q = -1 \text{ requires } q = 1, p = 0 \text{)}. $$

Reconciling more carefully: the formula $H H_{n} (X) = ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ allows $n$ to be negative, with $H H_{- q} (X) \supset H^{0} (X, Ω^{q})$ for $q \geq 0$ . For $X = E$ : $$ HH_n(E) = \begin{cases} H^1(E, \Omega^1) = k & n = 0 \text{ contribution from } (p, q) = (1, 1) \ H^0(E, \mathcal{O}) = k & n = 0 \text{ contribution from } (p, q) = (0, 0) \ H^1(E, \mathcal{O}) = k & n = 1 \text{ contribution from } (p, q) = (1, 0) \ H^0(E, \Omega^1) = k & n = -1 \text{ contribution from } (p, q) = (0, 1) \end{cases} $$ giving $H H_{0} (E) = k^{2}$ , $H H_{1} (E) = k$ , $H H_{- 1} (E) = k$ , and $H H_{n} (E) = 0$ for $∣ n ∣ \geq 2$ . Total dimension $\sum_{n} dim H H_{n} (E) = 4 = \sum_{p, q} h^{p, q} (E)$ , recovering the total Hodge cohomology.

Comparison with the topological Euler characteristic: $χ (E) = h^{0, 0} - h^{1, 0} - h^{0, 1} + h^{1, 1} = 1 - 1 - 1 + 1 = 0$ , matching the topological Euler characteristic of a complex torus. The Hochschild graded-trace recovers the Hodge polynomial $\sum_{p, q} h^{p, q} t^{p} u^{q}$ with the appropriate signs. Rubric: full credit for the scheme-theoretic HKR formula, the Hodge numbers of $E$ , the explicit Hochschild groups, and the comparison with the Euler characteristic.

Lean formalization Intermediate+

lean_status: none — Mathlib has the KahlerDifferential package (the universal $k$ -linear-derivation target $Ω_{A / k}^{1}$ for a commutative $k$ -algebra $A$ ) and the formal-smoothness predicate Algebra.FormallySmooth, but the full apparatus needed to state and prove the Hochschild-Kostant-Rosenberg theorem is not yet assembled. The intended formalisation reads schematically:

import Mathlib.RingTheory.Kaehler.Basic
import Mathlib.RingTheory.Smooth.Basic
import Mathlib.Algebra.Homology.HomologicalComplex
import Mathlib.LinearAlgebra.ExteriorAlgebra.Basic

namespace Codex.HomologicalAlgebra.HKR

variable (k : Type*) [Field k] [CharZero k]
variable (A : Type*) [CommRing A] [Algebra k A] [Algebra.FormallySmooth k A]

/-- The n-th Kähler differential module Ω^n_{A/k} = ⋀^n_A Ω^1_{A/k}. -/
noncomputable def kahlerDifferentials (n : ℕ) : Type* :=
  ExteriorPower A n (Ω[A⁄k])

/-- The antisymmetrisation map ε_n : Ω^n_{A/k} → C_n(A, A). -/
noncomputable def antisymmetrisation (n : ℕ) :
    kahlerDifferentials k A n →ₗ[A]
      Codex.HomologicalAlgebra.Hochschild.HH k A A n := sorry

/-- Hochschild-Kostant-Rosenberg theorem: for a smooth commutative
    k-algebra A in characteristic zero, the antisymmetrisation map
    induces an isomorphism HH_n(A, A) ≅ Ω^n_{A/k}. -/
theorem HKR (n : ℕ) :
    Codex.HomologicalAlgebra.Hochschild.HH k A A n ≃ₗ[A]
      kahlerDifferentials k A n := sorry

/-- HKR cohomology version: HH^n(A, A) ≅ ⋀^n_A Der_k(A, A). -/
theorem HKR_cohomology (n : ℕ) :
    Codex.HomologicalAlgebra.Hochschild.HHCoh k A A n ≃ₗ[A]
      ExteriorPower A n (Derivation k A A) := sorry

/-- Koszul resolution of A as an A⊗A-module by the regular sequence
    {x_i ⊗ 1 - 1 ⊗ x_i} for a polynomial algebra A = k[x_1, ..., x_d]. -/
noncomputable def koszulResolution :
    ChainComplex (ModuleCat (TensorProduct k A A)) ℕ := sorry

/-- For A = k[x_1, ..., x_d], the Koszul resolution is a free A^e-resolution
    of A; this is the polynomial-algebra base case of HKR. -/
theorem koszul_is_resolution [Fintype d] (A := MvPolynomial (Fin d) k) :
    QuasiIso (koszulResolution k A) := sorry

end Codex.HomologicalAlgebra.HKR

The proof gap is substantive at multiple levels. The Hochschild-homology API itself is not in Mathlib (as noted in 04.03.20, lean_status: none), so the HKR statement currently has nothing to compare against on the left side. The antisymmetrisation map requires the exterior-power module of Kähler differentials together with the explicit signed-average formula across the symmetric group $S_{n}$ , with the $1/ n!$ factor requiring $char k = 0$ (or at least $char k > n$ ). The polynomial-algebra base case via the Koszul resolution requires the formalisation of regular sequences in the enveloping algebra, the Koszul-complex construction, the projective-resolution property, and the comparison map between Koszul and bar resolutions; the existing Mathlib KoszulComplex package covers the basics but is not wired into the bar-resolution framework. The étale-descent reduction from general smooth $A$ to the polynomial case requires the étale-base-change property of Tor over the enveloping algebra and the local-structure theorem for smooth schemes (étale-locally a polynomial algebra), neither currently packaged. The scheme-theoretic version (Swan 1996, Yekutieli 2002) requires the diagonal $Δ : X \to X \times X$ , the six-functor formalism, and the derived category of $X \times X$ , all gated on the gaps from 04.03.16. The Gerstenhaber-bracket-to-Schouten-bracket correspondence requires the Schouten-Nijenhuis bracket on polyvector fields, currently absent from Mathlib. Each component is formalisable in principle but requires substantial coordinated infrastructure that the Mathlib homological-algebra and derived-category projects have not yet completed as of 2026.

Advanced results Master

Theorem (Hochschild-Kostant-Rosenberg; HKR 1962 Trans. AMS 102). For a smooth commutative $k$ -algebra $A$ of finite type with $k$ a field of characteristic zero, the antisymmetrisation map $\varepsilon : \Omega^{A/k} \to C(A, A) $i s a q u a s i - i so m or p hi s m, in d u c in g c an o ni c a l i so m or p hi s m s$ HH_n(A, A) \cong \Omega^n_{A/k} $o n H oc h sc hi l d h o m o l o g y an d$ HH^n(A, A) \cong \wedge^n_A \mathrm{Der}_k(A, A)$ on Hochschild cohomology.

The HKR theorem is the central calculation of Hochschild theory in the commutative smooth case. It identifies Hochschild homology of a smooth commutative algebra with the classical Kähler differentials, recovering the standard differential-forms framework from the more general Hochschild apparatus. The theorem extends to smooth schemes via localisation: Swan 1996 J. Pure Appl. Algebra 110 ^{[source pending]} proved the sheafified version, showing $H H_{n} (X) ≅ ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ for $X$ a smooth quasiprojective $k$ -scheme in characteristic zero; Yekutieli 2002 Math. Res. Lett. 9 ^{[source pending]} gave the continuous and formal-completion versions; Caldararu 2003 Adv. Math. 194 ^{[source pending]} extended the formalism to the derived category and the Mukai pairing on $H H_{*}$ . In positive characteristic the antisymmetrisation involves $1/ n!$ , which fails when $char k$ divides $n!$ ; the correct replacement is the derived de Rham complex of Illusie (1971/72) and the conjugate filtration framework of Bhatt-Morrow-Scholze 2019.

Theorem (Kontsevich formality; Kontsevich 2003 Lett. Math. Phys. 66). For a smooth commutative $k$ -algebra $A$ in characteristic zero (or more generally a smooth Poisson manifold), the Hochschild cochain complex $C^(A, A) $(r e g a r d e d a s a d i f f er e n t ia l g r a d e d L i e a l g e b r a u n d er t h e G er s t e nhab er b r a c k e t) i s$ L_\infty $- q u a s i - i so m or p hi c t o t h e p o l y v ec t or f i e l d a l g e b r a$ T^_{\mathrm{poly}}(A) = \wedge^_A \mathrm{Der}_k(A, A)$ (with zero differential and Schouten-Nijenhuis bracket).*

Kontsevich's formality theorem ^{[source pending]} is one of the deepest results in derived algebraic geometry and a landmark of mathematical physics: it lifts the HKR identification on cohomology to a homotopy-coherent equivalence of $L_{\infty}$ -algebras, and it solves the deformation-quantisation problem for arbitrary Poisson manifolds. The deformation-theoretic content: an $L_{\infty}$ -quasi-isomorphism between two differential graded Lie algebras preserves Maurer-Cartan equivalence classes, and the Maurer-Cartan elements of $T_{poly}^{*} (A)$ with the Schouten-Nijenhuis bracket are exactly Poisson bivectors $π \in T_{poly}^{2}$ satisfying $[π, π]_{S N} = 0$ ; the Maurer-Cartan elements of $C^{*} (A, A)$ with the Gerstenhaber bracket are associative star products $⋆_{ℏ} : A [[ℏ]] \otimes A [[ℏ]] \to A [[ℏ]]$ deforming the multiplication of $A$ . The formality $L_{\infty}$ -quasi-isomorphism therefore gives a bijection between Poisson structures and equivalence classes of star products — the deformation quantisation of Poisson manifolds. The proof uses the explicit Kontsevich formality morphism $U = \sum_{n} U_{n}$ , with each $U_{n}$ given by an integral over the compactified configuration space of $n$ points in the upper half plane $H \subset C$ , weighted by graphs (the "Kontsevich graphs") and integrals of hyperbolic angle-cocycles.

Theorem (Tamarkin's proof of formality; Tamarkin 1998). The Kontsevich formality theorem follows from the formality of the little disks operad $E_{2}$ over the rationals.

Tamarkin's proof (preprint 1998, arXiv math/9803025 ^{[source pending]}; expository version in Hinich 2003 Forum Math. 15 ^{[source pending]}) reorganises the deformation-quantisation problem through the operadic lens. The chain operad of the little disks $E_{2}$ acts on the Hochschild cochain complex $C^{*} (A, A)$ (Deligne conjecture; proved by McClure-Smith 2002, Kontsevich-Soibelman 2009 ^{[source pending]}), and the Tamarkin-Hinich proof exploits the rational formality of $E_{2}$ together with the Etingof-Kazhdan quantisation of Lie bialgebras to produce the formality morphism abstractly. This operadic proof clarifies the conceptual role of the $E_{2}$ -action and connects the formality theorem to the broader framework of operadic deformation theory.

Theorem (HKR for derived schemes; Toën-Vezzosi 2011). For a quasi-smooth derived scheme $X$ in characteristic zero, the Hochschild homology $HH_(X) $i d e n t i f i esc an o ni c a l l y w i t h t h e d er i v e d g l o ba l sec t i o n so f t h esy mm e t r i c a l g e b r a o n t h es hi f t e d co t an g e n t co m pl e x :$ HH_*(X) \cong \mathrm{R}\Gamma(X, \mathrm{Sym}(\mathbb{L}_X[-1]))$ in the appropriate derived-categorical sense.*

The Toën-Vezzosi derived-HKR theorem (2011 Selecta Math. 17 ^{[source pending]}) extends the classical HKR formula to derived algebraic geometry, where the cotangent complex $L_{X}$ (Quillen, Illusie) replaces the classical Kähler differentials, and the derived scheme structure allows the formula to handle non-smooth situations through the derived-stack framework. The result is the foundational input to derived deformation theory and to the modern $\infty$ -categorical reformulation of Hodge theory via the Hochschild-Kostant-Rosenberg-Toën-Vezzosi framework. For a classical smooth scheme $X$ , the formula recovers the Swan-Yekutieli formula $H H_{*} (X) = ⨁_{p, q} H^{p} (X, Ω_{X}^{q})$ ; for derived schemes the formula extends to capture the higher derived-categorical structure.

Theorem (positive-characteristic HKR; Antieau-Bhatt-Mathew 2020 / Bhatt-Morrow-Scholze 2019). In positive characteristic, the classical HKR identification of Hochschild homology with Kähler differentials fails, but the topological Hochschild homology $THH$ together with the cyclotomic structure admits a refined HKR-type description in terms of the conjugate filtration on the derived de Rham complex and the Frobenius operator.

The positive-characteristic refinement of HKR is one of the central themes of modern arithmetic algebraic geometry. Bhatt 2012 Compos. Math. 148 established the comparison between Hochschild homology and the derived de Rham complex in positive characteristic; Antieau-Bhatt-Mathew 2020 ^{[source pending]} developed the systematic theory of the conjugate filtration on $THH$ ; Bhatt-Morrow-Scholze 2019 Publ. Math. IHES 129 ^{[source pending]} introduced prismatic cohomology as the unifying framework that recovers crystalline, étale, and de Rham cohomology theories in characteristic $p$ and mixed characteristic, with HKR appearing as the characteristic-zero degeneration of the prismatic framework. The full positive-characteristic HKR statement involves the motivic filtration on $THH$ and the cyclotomic structure (Nikolaus-Scholze 2018 Acta Math. 221), and it is the input to the modern computation of algebraic $K$ -theory via trace methods.

Theorem (non-smooth case via the cotangent complex; Quillen, Illusie, Avramov-Iyengar). For a non-smooth commutative $k$ -algebra $A$ in characteristic zero, the Hochschild homology $H H_{n} (A, A)$ is not isomorphic to $Ω_{A / k}^{n}$ in general. The correct replacement uses the cotangent complex $L_{A / k}$ of Quillen-Illusie, and the HKR-type identification reads $$ HH_n(A, A) \cong \mathrm{R}\mathrm{Sym}^n(\mathbb{L}_{A/k}[1]) $$ in the derived category, the André-Quillen-HKR identification.

The non-smooth-case extension of HKR is one of the foundational results of derived algebraic geometry. The cotangent complex $L_{A / k}$ (Quillen 1970 Lecture Notes in Algebra, Illusie 1971/72 Springer LNM 239 and 283) is the derived-categorical replacement for the Kähler differentials $Ω_{A / k}^{1}$ ; it equals $Ω_{A / k}^{1}$ concentrated in degree zero precisely when $A$ is smooth over $k$ , and it has additional cohomology in negative degrees when $A$ is singular. The Avramov-Iyengar series on the André-Quillen homology obstruction (Avramov 1999 Ann. of Math. 150, Iyengar 2007 surveys) develops the obstruction theory: the failure of HKR for non-smooth $A$ is measured by the André-Quillen cohomology $D_{*} (A / k; A) = Tor_{*} (L_{A / k}, A)$ , and HKR holds iff the André-Quillen obstruction vanishes (equivalently, iff $A$ is smooth over $k$ ).

Theorem (HKR and Hodge theory; Deligne, Illusie, Toën-Vezzosi). For a smooth proper $k$ -scheme $X$ in characteristic zero, the scheme-theoretic HKR identification $H H_{n} (X) = ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ is the $(p, q)$ -bigraded Hodge decomposition of Hochschild homology, identifying $HH_(X) $w i t h t h e a l g e b r ai cH o d g eco h o m o l o g y o f$ X $. T h e d e R ham - t o - H o d g es p ec t r a l se q u e n ceco n v er g in g t o$ H^_{\mathrm{dR}}(X/k) $d e g e n er a t es a t$ E_1 $(t h eH o d g e - t o - d e R ham d e g e n er a t i o n o f D e l i g n e - I l l u s i e 1987 * I n v e n t . M a t h . * 89 in c ha r a c t er i s t i c$ p$, lifted to characteristic zero by descent), confirming that the Hochschild-Hodge identification is the algebraic shadow of the classical Hodge decomposition.

The HKR-Hodge identification is the bridge from Hochschild theory to classical Hodge theory. For a smooth proper complex variety $X$ , the algebraic Hochschild homology $H H_{*} (X) = ⨁_{p, q} H^{p} (X, Ω_{X}^{q})$ recovers the algebraic Dolbeault cohomology, which by the classical Hodge decomposition is the singular cohomology $H^{*} (X (C), C) = ⨁_{p + q = n} H^{p, q} (X)$ (Deligne-Illusie 1987 in characteristic $p$ via the Cartier isomorphism, lifted via deformation to characteristic zero). The cup product on $H H^{*} (X)$ recovers the cohomological cup product on Hodge cohomology, and the entire Hochschild theory of $X$ packages the Hodge data of $X$ into a single Hochschild-categorical invariant — the input to derived noncommutative geometry (Kontsevich-Soibelman 2009 Homological Mirror Symmetry).

Synthesis. The Hochschild-Kostant-Rosenberg theorem builds toward a unified framework identifying Hochschild homology and cohomology of smooth commutative algebras and schemes with the classical differential-forms calculus and polyvector field calculus, and the foundational reason the construction works is that for a smooth algebra $A$ , the multiplication map $μ : A \otimes_{k} A \to A$ has kernel locally generated by a regular sequence (the diagonal ideal is locally a complete intersection in $A^{e}$ ), so the Koszul complex provides a finite free resolution of $A$ as an $A^{e}$ -module, and tensoring with $A$ over $A^{e}$ collapses the Koszul differential to zero by commutativity. The package is structurally tight: for affine smooth $A$ in characteristic zero, $H H_{n} (A, A) ≅ Ω_{A / k}^{n}$ and $H H^{n} (A, A) ≅ \land_{A}^{n} Der_{k} (A, A)$ ; for smooth schemes, $H H_{n} (X) ≅ ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ (Swan, Yekutieli); for derived schemes, $H H_{*} (X) ≅ R Γ (X, Sym (L_{X} [- 1]))$ (Toën-Vezzosi); in positive characteristic, the conjugate filtration on $THH$ and the prismatic cohomology framework (Bhatt-Morrow-Scholze) repair the failure of classical HKR; and for non-smooth algebras the cotangent complex $L_{A / k}$ replaces the Kähler differentials (Quillen-Illusie). The central insight is that the Hochschild invariants of a smooth commutative algebra package the entire classical differential-geometric structure of the algebra (forms, polyvector fields, Hodge cohomology) into a single homological invariant — and that the Kontsevich formality theorem lifts this identification at the cohomological level to a homotopy-coherent equivalence of $L_{\infty}$ -algebras, solving the deformation-quantisation problem and providing the algebraic foundation of quantum mechanics on Poisson manifolds.

The framework appears again in 04.03.20 (Hochschild homology and cohomology), where HKR is one of the central computations of the Hochschild apparatus; in 04.08.01 (sheaf of differentials), where the Kähler differentials are the global geometric incarnation of the algebraic differential-forms calculus that HKR recovers; in 04.03.17 (derived tensor product and Tor), where the HKR theorem is a computation of $Tor_{*}^{A^{e}} (A, A)$ via the Koszul resolution; and forward in 04.03.22 (cyclic homology), where the Connes SBI long exact sequence combines with HKR to identify cyclic homology of a smooth commutative algebra in characteristic zero with the algebraic de Rham cohomology. The recursion stabilises: every smooth commutative algebra in characteristic zero recovers its classical differential forms via HKR, every smooth proper scheme recovers its Hodge cohomology via scheme-theoretic HKR, every Poisson structure deforms to an associative star product via Kontsevich formality, and the entire framework lifts to derived algebraic geometry and to positive characteristic via the cotangent complex and the prismatic-cohomology refinements.

Full proof set Master

Proposition (Koszul resolution of a polynomial algebra over its enveloping algebra). Let $A = k [x_{1}, \dots, x_{d}]$ and $A^{e} = A \otimes_{k} A$ . The Koszul complex $K_\bullet = \wedge^_{A^e}(\bigoplus_i A^e \cdot e_i) $w i t h d i f f er e n t ia l$ \partial(e_i) = x_i \otimes 1 - 1 \otimes x_i $an d a ug m e n t a t i o n$ \mu : K_0 = A^e \to A $i s a f r ee$ A^e $- r eso l u t i o n o f$ A $o f l e n g t h$ d$.*

Proof. The kernel of the multiplication map $μ : A^{e} = k [x_{1}, \dots, x_{d}, y_{1}, \dots, y_{d}] \to A = k [x_{1}, \dots, x_{d}]$ (where $y_{i} = 1 \otimes x_{i}$ , so $μ (y_{i}) = x_{i}$ ) is the ideal $I = ker (μ) = (x_{1} - y_{1}, \dots, x_{d} - y_{d})$ . The sequence ${x_{i} - y_{i}}_{i = 1}^{d}$ is a regular sequence in $A^{e}$ : the quotient $A^{e} / (x_{1} - y_{1}, \dots, x_{i} - y_{i})$ at each step is the polynomial algebra in the remaining $2 d - i$ variables, which is a domain (in particular has no zero divisors), so each subsequent element $x_{i + 1} - y_{i + 1}$ is a non-zero-divisor in the quotient.

By the general theory of Koszul complexes for regular sequences, the Koszul complex $K_{∙}$ of a regular sequence ${f_{1}, \dots, f_{d}}$ in a commutative ring $R$ is a free $R$ -resolution of $R / (f_{1}, \dots, f_{d})$ of length $d$ . Specialising to $R = A^{e}$ and ${f_{i}} = {x_{i} - y_{i}}$ : the Koszul complex $K_{∙}$ is a free $A^{e}$ -resolution of $A^{e} / I = A^{e} / ker (μ) = A$ of length $d$ . Each $K_{n} = \land_{A^{e}}^{n} (A^{e} \cdot e_{1} \oplus \dots \oplus A^{e} \cdot e_{d})$ is free of rank $(n d)$ over $A^{e}$ , with basis the wedges $e_{i_{1}} \land \dots \land e_{i_{n}}$ for $1 \leq i_{1} < \dots < i_{n} \leq d$ . $□$

Proposition (Hochschild homology of the polynomial algebra via Koszul). For $A = k [x_{1}, \dots, x_{d}]$ , the Hochschild homology is $H H_{n} (A, A) = A \otimes_{k} \land^{n} (k^{d}) = Ω_{A / k}^{n}$ for $0 \leq n \leq d$ , and zero for $n > d$ .

Proof. By the previous proposition, the Koszul complex $K_{∙} \to A$ is a free $A^{e}$ -resolution of $A$ . Hence $H H_{n} (A, A) = Tor_{n}^{A^{e}} (A, A) = H_{n} (A \otimes_{A^{e}} K_{∙})$ . Compute $A \otimes_{A^{e}} K_{n}$ : under the standard $A$ -bimodule identification, $A \otimes_{A^{e}} A^{e} = A$ (with $y_{i}$ acting as right multiplication by $x_{i}$ , equivalently as left multiplication on the commutative algebra $A$ ), and $A \otimes_{A^{e}} (A^{e} \cdot e_{i_{1}} \land \dots \land e_{i_{n}}) = A \cdot e_{i_{1}} \land \dots \land e_{i_{n}}$ .

The Koszul differential $\partial (e_{i}) = x_{i} - y_{i}$ acts on $A \otimes_{A^{e}} K_{∙}$ as multiplication by $x_{i} - x_{i} = 0$ (since $y_{i}$ acts as multiplication by $x_{i}$ on $A$ after the tensor identification). The tensored complex has zero differential, so $$ HH_n(A, A) = H_n(A \otimes_{A^e} K_\bullet) = A \otimes_k \wedge^n(k \cdot e_1 \oplus \cdots \oplus k \cdot e_d) = A \otimes_k \wedge^n(k^d). $$ The final identification $A \otimes_{k} \land^{n} (k^{d}) = Ω_{A / k}^{n}$ uses that the Kähler differentials of the polynomial algebra are $Ω_{A / k}^{1} = A \cdot d x_{1} \oplus \dots \oplus A \cdot d x_{d}$ , free of rank $d$ , with the symbols $d x_{i}$ corresponding to the basis elements $e_{i}$ (after relabelling). The exterior power $Ω_{A / k}^{n} = \land_{A}^{n} Ω_{A / k}^{1} = A \otimes_{k} \land^{n} (k^{d})$ . $□$

Proposition (antisymmetrisation is a chain map). The antisymmetrisation map $\varepsilon : (\Omega^{A/k}, 0) \to (C(A, A), b) $d e f in e d b y$ \varepsilon_n(a_0 , da_1 \wedge \cdots \wedge da_n) = (1/n!) \sum_\sigma \mathrm{sgn}(\sigma) a_0 \otimes a_{\sigma(1)} \otimes \cdots \otimes a_{\sigma(n)} $s a t i s f i es$ b \circ \varepsilon = 0 = \varepsilon \circ 0 $f or e v er y co mm u t a t i v e$ k $- a l g e b r a$ A$ in characteristic zero.

Proof. The right side is immediate: $ε \circ 0 = 0$ . For the left side, compute $b (ε_{n} (a_{0} d a_{1} \land \dots \land d a_{n}))$ directly. The Hochschild differential on a chain $a_{0} \otimes a_{1} \otimes \dots \otimes a_{n}$ is $$ b(a_0 \otimes a_1 \otimes \cdots \otimes a_n) = a_0 a_1 \otimes a_2 \otimes \cdots \otimes a_n + \sum_{i=1}^{n-1} (-1)^i a_0 \otimes \cdots \otimes a_i a_{i+1} \otimes \cdots \otimes a_n + (-1)^n a_n a_0 \otimes a_1 \otimes \cdots \otimes a_{n-1}. $$ Applying $b$ to the antisymmetric sum $(1/ n!) \sum_{σ} sgn (σ) a_{0} \otimes a_{σ (1)} \otimes \dots \otimes a_{σ (n)}$ and using commutativity of $A$ (so all $a_{i} a_{j}$ products commute and the cyclic term $a_{n} a_{0} = a_{0} a_{n}$ ), each interior multiplication $a_{σ (i)} a_{σ (i + 1)}$ appears paired with the transposed permutation $σ^{'} = σ \cdot (i i + 1)$ : the contribution from $σ$ is $(- 1)^{i} sgn (σ) a_{0} \otimes \dots \otimes a_{σ (i)} a_{σ (i + 1)} \otimes \dots$ , while the contribution from $σ^{'}$ is $(- 1)^{i} sgn (σ^{'}) a_{0} \otimes \dots \otimes a_{σ^{'} (i)} a_{σ^{'} (i + 1)} \otimes \dots = (- 1)^{i} (- 1) sgn (σ) a_{0} \otimes \dots \otimes a_{σ (i + 1)} a_{σ (i)} \otimes \dots$ .

By commutativity, $a_{σ (i + 1)} a_{σ (i)} = a_{σ (i)} a_{σ (i + 1)}$ , so the two terms cancel exactly. Pairing up all permutations into such transposed pairs (the cosets of the subgroup $⟨(i i + 1)⟩ \subset S_{n}$ ) shows that all interior-multiplication contributions cancel. The leading term $a_{0} a_{σ (1)} \otimes a_{σ (2)} \otimes \dots$ and the trailing cyclic term $(- 1)^{n} a_{σ (n)} a_{0} \otimes a_{σ (1)} \otimes \dots$ cancel similarly by pairing $σ$ with the cyclically shifted permutation $σ \cdot c$ where $c = (12 \dots n)$ is the cyclic shift, using $sgn (c) = (- 1)^{n - 1}$ and commutativity. Hence $b (ε_{n} (ω)) = 0$ for every $ω \in Ω_{A / k}^{n}$ . $□$

Proposition (HKR for smooth $A$ via étale descent). Let $A$ be a smooth commutative $k$ -algebra of finite type in characteristic zero. The antisymmetrisation map $\varepsilon : \Omega^{A/k} \to C(A, A)$ is a quasi-isomorphism.

Proof. Both $H H_{n} (A, A) = Tor_{n}^{A^{e}} (A, A)$ and $Ω_{A / k}^{n}$ commute with localisation at a multiplicative subset $S \subset A$ (Tor commutes with flat base change, and Kähler differentials commute with localisation by their universal property), so the claim is local on $Spec (A)$ . By the local-structure theorem for smooth schemes (Stacks Project tag 039P), every closed point $m \subset A$ has an open neighbourhood $A \to A^{'}$ (Zariski open, replacing $A$ by a localisation) such that $A^{'}$ is étale over a polynomial algebra $B = k [x_{1}, \dots, x_{d}]$ for some $d =$ relative dimension at $m$ .

Both Hochschild homology and Kähler differentials commute with étale base change (étale base change for Tor over the enveloping algebra is a standard derived-categorical fact; for Kähler differentials it is the universal-property statement that étale maps induce isomorphisms on $Ω^{1}$ ). Hence the HKR map $ε : Ω_{B / k}^{*} \to H H_{*} (B, B)$ , which is a quasi-isomorphism for $B = k [x_{1}, \dots, x_{d}]$ by the previous proposition, base-changes to a quasi-isomorphism $ε : Ω_{A^{'} / k}^{*} \to H H_{*} (A^{'}, A^{'})$ on the étale neighbourhood $A^{'}$ . Gluing over a Zariski cover of $Spec (A)$ produces the global quasi-isomorphism $ε : Ω_{A / k}^{*} \to H H_{*} (A, A)$ . $□$

Proposition (HKR cohomology side via duality). For $A$ a smooth commutative $k$ -algebra in characteristic zero, the Hochschild cohomology satisfies $H H^{n} (A, A) ≅ \land_{A}^{n} Der_{k} (A, A)$ , the $n$ -th polyvector field module of $A$ .

Proof. For smooth $A$ , $Ω_{A / k}^{1}$ is a projective $A$ -module of finite rank, so by the universal property of Kähler differentials $Der_{k} (A, A) = Hom_{A} (Ω_{A / k}^{1}, A)$ . The natural map $\land_{A}^{n} Hom_{A} (Ω^{1}, A) \to Hom_{A} (\land_{A}^{n} Ω^{1}, A) = Hom_{A} (Ω_{A / k}^{n}, A)$ is an isomorphism for finitely generated projective $Ω_{A / k}^{1}$ (standard linear-algebra-over-rings fact for projective modules of finite rank). Hence $\land_{A}^{n} Der_{k} (A, A) = Hom_{A} (Ω_{A / k}^{n}, A)$ .

For the cohomology side of HKR, compute $H H^{n} (A, A) = Ext_{A^{e}}^{n} (A, A)$ by dualising the Koszul resolution (or any other $A^{e}$ -projective resolution). Using the Koszul resolution from above for the polynomial-algebra case: $Hom_{A^{e}} (K_{n}, A) = Hom_{A^{e}} (\land_{A^{e}}^{n} (⨁ A^{e} e_{i}), A)$ , which after the standard adjunction identifies with $Hom_{A} (\land_{A}^{n} (k^{d}), A) = Hom_{A} (Ω_{A / k}^{n}, A) = \land_{A}^{n} Der_{k} (A, A)$ . The induced differential is zero by the same Koszul argument as in the homology case. Reading off cohomology: $H H^{n} (A, A) = \land_{A}^{n} Der_{k} (A, A)$ . The étale-descent reduction extends the polynomial-algebra case to general smooth $A$ as in the previous proposition. $□$

Proposition (HKR and the Schouten-Nijenhuis bracket). Under the HKR isomorphism $HH^(A, A) \cong T^_{\mathrm{poly}}(A) := \wedge^A \mathrm{Der}k(A, A) $f or$ A $s m oo t h co mm u t a t i v e in c ha r a c t er i s t i cz er o, t h e G er s t e nhab er b r a c k e t o n$ HH^* $cor r es p o n d s t o t h e S c h o u t e n - N ij e nh u i s b r a c k e t$ [-, -]{SN} $o n p o l y v ec t or f i e l d s, an d t h ec u pp r o d u c t o n$ HH^ $cor r es p o n d s t o t h e w e d g e p r o d u c t o n$ T^{\mathrm{poly}}$.*

Proof. The cup product on $H H^{*}$ (defined at the cochain level by $(f ⌣ g) (a_{1}, \dots, a_{p + q}) := f (a_{1}, \dots, a_{p}) \cdot g (a_{p + 1}, \dots, a_{p + q})$ ) corresponds under HKR to the wedge product on polyvector fields: for $f \in H H^{p} = \land^{p} Der$ and $g \in H H^{q} = \land^{q} Der$ , the cup product $f ⌣ g$ identifies with $f \land g \in \land^{p + q} Der$ . This follows from the explicit antisymmetrisation-dual formula for the HKR isomorphism on the cohomology side and the multiplicativity of the wedge product.

The Gerstenhaber bracket on $H H^{*+ 1}$ (defined at the cochain level by insertions) corresponds to the Schouten-Nijenhuis bracket on $T_{poly}^{*+ 1}$ . On the degree- $1$ piece $H H^{1} = T_{poly}^{1} = Der_{k} (A, A)$ , the Gerstenhaber bracket reduces to the commutator of derivations $[D, D^{'}]_{G} = D D^{'} - D^{'} D$ (see 04.03.20 proof of "Gerstenhaber bracket of derivations"), and the Schouten-Nijenhuis bracket on $T^{1}$ similarly reduces to the commutator. The agreement on $T^{1}$ extends to all degrees by the graded-Leibniz extension property of the Schouten-Nijenhuis bracket: $[-, -]_{S N}$ is the unique extension of the Lie bracket of vector fields to $T_{poly}^{*}$ satisfying the graded Leibniz rule with respect to the wedge product. Since the Gerstenhaber bracket on $H H^{*}$ also satisfies the graded Leibniz rule with respect to the cup product (Gerstenhaber algebra axiom), and since the cup product corresponds to the wedge product on $T_{poly}^{*}$ , the two brackets must agree on all degrees by the uniqueness of the Schouten-Nijenhuis extension. $□$

Proposition (deformation-quantisation content of HKR). For $A$ a smooth commutative $k$ -algebra in characteristic zero, an infinitesimal deformation $η \in H H^{2} (A, A) = T_{poly}^{2} (A) = \land_{A}^{2} Der_{k} (A, A)$ is a bivector field on $Spec (A)$ , and the second-order obstruction $[η, η]_{S N} \in H H^{3} (A, A) = T_{poly}^{3} (A)$ to extending $η$ to a higher-order deformation vanishes iff $η$ is a Poisson bivector.

Proof. By HKR, $H H^{2} (A, A) = T_{poly}^{2} (A)$ , and a class $η \in H H^{2} (A, A)$ corresponds under HKR to a bivector field $η = \sum_{i < j} η_{ij} \partial_{x_{i}} \land \partial_{x_{j}}$ in local coordinates. By the general Gerstenhaber theory (see 04.03.20 for $H H^{2}$ as infinitesimal deformations and $H H^{3}$ as obstructions), the second-order obstruction is $[η, η]_{G} \in H H^{3}$ . Under HKR, this corresponds to $[η, η]_{S N} \in T_{poly}^{3} (A) = \land_{A}^{3} Der_{k} (A, A)$ .

The bracket ${f, g} := η (df, d g) = \sum_{i, j} η_{ij} (\partial_{x_{i}} f) (\partial_{x_{j}} g)$ is a $k$ -bilinear, antisymmetric bracket on $A$ (by the antisymmetry of $η$ as a bivector). The Jacobi identity for ${-, -}$ — namely ${f, {g, h}} + {g, {h, f}} + {h, {f, g}} = 0$ for all $f, g, h \in A$ — translates under the bivector correspondence to the condition $[η, η]_{S N} = 0$ in $T_{poly}^{3} (A)$ . A bivector $η$ satisfying the Jacobi identity (equivalently $[η, η]_{S N} = 0$ ) is called a Poisson bivector or Poisson structure on $Spec (A)$ . Hence the vanishing of the second-order obstruction is precisely the Poisson condition.

The deformation-quantisation content: the higher-order obstructions to extending an infinitesimal deformation $η \in H H^{2} = T_{poly}^{2}$ to a full deformation (an associative star product $⋆_{ℏ} : A [[ℏ]] \otimes A [[ℏ]] \to A [[ℏ]]$ with $a ⋆_{ℏ} b = ab + ℏ \cdot η (a, b) + O (ℏ^{2})$ ) all live in $H H^{3} (A, A)$ and are captured by the higher Maurer-Cartan equation for the Gerstenhaber differential graded Lie algebra structure on $C^{*} (A, A)$ . Kontsevich's formality theorem (2003) implies that all higher obstructions reduce to $[η, η]_{S N} = 0$ at the level of the polyvector field $L_{\infty}$ -algebra, and the formality $L_{\infty}$ -quasi-isomorphism $U : T_{poly}^{*} \to C^{*} (A, A)$ transports a Poisson bivector $η$ to a Maurer-Cartan element of $C^{*} (A, A)$ — an associative star product. Hence every Poisson structure on a smooth commutative $k$ -algebra in characteristic zero admits a deformation quantisation as a star product on $A [[ℏ]]$ . $□$

Proposition (scheme-theoretic HKR; Swan 1996). For a smooth quasiprojective $k$ -scheme $X$ of finite type in characteristic zero, $H H_{n} (X) ≅ ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ .

Proof. Define $H H_{n} (X) := Ext_{D (X \times_{k} X)}^{- n} (Δ_{*} O_{X}, Δ_{*} O_{X})$ , where $Δ : X \to X \times_{k} X$ is the diagonal embedding. The local-to-global computation: the local Hochschild homology sheaf $HH_{n} (X) := E x t_{O_{X \times X}}^{- n} (Δ_{*} O_{X}, Δ_{*} O_{X})$ on $X \times X$ has stalk at the diagonal point $(Δ (x))$ equal to the affine Hochschild homology of the local ring $O_{X, x}$ (a smooth local $k$ -algebra in characteristic zero), which by the affine HKR equals $Ω_{O_{X, x} / k}^{n}$ . Hence the local Hochschild homology sheaf identifies with the sheaf of Kähler differentials: $HH_{n} (X) ≅ Δ_{*} Ω_{X / k}^{n}$ on $X \times X$ .

The global Hochschild homology is then computed via the local-to-global spectral sequence $E_{2}^{p, - n} = H^{p} (X, HH_{n} (X)) \Rightarrow H H_{n - p} (X)$ . Substituting $HH_{n} (X) = Ω_{X / k}^{n}$ : $$ E_2^{p, -n} = H^p(X, \Omega^n_{X/k}) \Rightarrow HH_{n - p}(X). $$ For $X$ smooth in characteristic zero, the spectral sequence degenerates at $E_{2}$ (the Hodge-to-de Rham degeneration of Deligne-Illusie 1987, applied to the Hochschild-to-Hodge spectral sequence by a similar Cartier-isomorphism-and-deformation argument; see Swan 1996 for the original proof and Caldararu 2003 for the derived-categorical reformulation), giving the direct sum decomposition $$ HH_n(X) = \bigoplus_{p - q = n} H^p(X, \Omega^q_{X/k}). $$ This is the scheme-theoretic HKR formula, identifying Hochschild homology of a smooth scheme with its Hodge cohomology. $□$

Connections Master

Hochschild homology and cohomology 04.03.20. HKR is the central calculation of Hochschild theory in the smooth commutative case: it identifies $H H_{n} (A, A) ≅ Ω_{A / k}^{n}$ and $H H^{n} (A, A) ≅ \land_{A}^{n} Der_{k} (A, A)$ for $A$ smooth commutative in characteristic zero, recovering the classical Kähler-differentials and polyvector-field calculus from the more general Hochschild apparatus. The Gerstenhaber-bracket-to-Schouten-bracket correspondence under HKR is the bridge from the general Hochschild deformation theory to the classical Poisson-bivector framework.
Sheaf of differentials 04.08.01. The Kähler differentials $Ω_{A / k}^{n}$ on the affine side and the sheaf of differentials $Ω_{X / k}^{n}$ on the scheme side are the geometric incarnations of the differential-forms calculus that HKR recovers from Hochschild theory. The universal-property characterisation of $Ω_{A / k}^{1}$ as representing $k$ -linear derivations is the input to the HKR identification on the cohomology side, and the smoothness hypothesis on $A$ (equivalently, projectivity of $Ω_{A / k}^{1}$ as an $A$ -module) is the key technical condition that makes HKR work.
Derived tensor product $\otimes^{L}$ and Tor 04.03.17. The HKR theorem is a computation of $Tor_{*}^{A^{e}} (A, A) = H H_{*} (A, A)$ via the Koszul resolution of $A$ as an $A^{e}$ -module. The smoothness hypothesis is what makes the diagonal ideal $ker (μ : A^{e} \to A)$ locally generated by a regular sequence, hence allows the Koszul complex to provide a finite free resolution and the Tor computation to collapse to the Kähler differentials.
Derived functors $R F$ and $L F$ via derived categories 04.03.12. The Koszul resolution and the bar resolution are two different projective $A^{e}$ -resolutions of $A$ , and the HKR antisymmetrisation map is the comparison morphism between them at the level of cycles (well-defined up to chain homotopy). The general framework of left-derived functors and their independence from resolution choices applies directly, identifying the HKR isomorphism as the canonical comparison between the Koszul-computed and bar-computed Hochschild homology.
t-Structure on a triangulated category — heart and truncations 04.03.18. The Hochschild-Hodge identification for smooth proper schemes (Swan 1996) identifies $H H_{*} (X)$ with the Hodge cohomology, which is the cohomology of the perverse heart of an appropriate t-structure on the derived category $D^{b} (Coh (X))$ . The Bondal-Orlov reconstruction theorem and the Caldararu Mukai pairing extend this identification to the derived-noncommutative setting.
Six-functor formalism 04.03.16. The scheme-theoretic HKR formula uses the diagonal $Δ : X \to X \times_{k} X$ and the derived-category structure on $X \times X$ . The local Hochschild homology sheaf $HH_{n} (X) = E x t_{O_{X \times X}}^{- n} (Δ_{*} O_{X}, Δ_{*} O_{X})$ is computed via the six-functor operations, and the local-to-global spectral sequence reduces the global Hochschild homology to the Hodge cohomology on $X$ .
Spectral sequence of a filtered complex 04.03.14. The local-to-global Hochschild-to-Hodge spectral sequence $E_{2}^{p, q} = H^{p} (X, HH_{q} (X)) \Rightarrow H H_{p + q} (X)$ for a smooth scheme $X$ is a key technical input to the scheme-theoretic HKR formula, with the degeneration at $E_{2}$ (proved via the Cartier isomorphism and Hodge-to-de Rham degeneration of Deligne-Illusie 1987) producing the direct-sum decomposition $H H_{n} (X) = ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ .
Deformation quantisation and Kontsevich formality. The Kontsevich formality theorem (2003 Lett. Math. Phys. 66) lifts the HKR identification of $H H^{*} (A, A) ≅ T_{poly}^{*} (A)$ on cohomology to an $L_{\infty}$ -quasi-isomorphism of $L_{\infty}$ -algebras, identifying the Hochschild cochain complex with the polyvector field algebra as differential graded Lie algebras. The result solves the deformation-quantisation problem for arbitrary Poisson manifolds: every Poisson structure on a smooth manifold deforms to an associative star product on the algebra of functions, giving an algebraic foundation for quantum mechanics on Poisson manifolds. Forward-links: 04.03.22 (cyclic homology and Connes SBI sequence), where HKR combined with SBI identifies cyclic homology of a smooth commutative algebra in characteristic zero with algebraic de Rham cohomology.
Cotangent complex and André-Quillen homology (forward to 01.04.40). For non-smooth $A$ , the Kähler differentials $Ω_{A / k}^{*}$ are not the correct invariant; the cotangent complex $L_{A / k}$ of Quillen 1970 / Illusie 1971/72 replaces it, and the André-Quillen homology $D_{*} (A / k; M) = Tor_{*} (L_{A / k}, M)$ measures the failure of HKR. The Avramov-Iyengar work on the André-Quillen obstruction develops the obstruction theory for HKR to fail in the non-smooth case, and the Toën-Vezzosi derived HKR (2011 Selecta Math. 17) extends the HKR formula to derived algebraic geometry using the cotangent complex in place of the Kähler differentials.

Historical & philosophical context Master

The Hochschild-Kostant-Rosenberg theorem was proved by Gerhard Hochschild, Bertram Kostant, and Alex Rosenberg in 1962 in the paper "Differential forms on regular affine algebras" (Transactions of the American Mathematical Society 102, 383-408) ^{[source pending]}. The theorem identifies the Hochschild homology of a smooth commutative algebra in characteristic zero with the Kähler differentials, providing a precise bridge from the noncommutative Hochschild framework (introduced by Hochschild himself in 1945 to study cohomology of associative algebras) back to the classical commutative-algebra differential-forms calculus going back to Kähler 1933. The proof in the original HKR paper used the explicit Koszul resolution of a smooth algebra over its enveloping algebra and the antisymmetrisation map, both essentially as presented above. The "regular affine algebra" terminology in the title reflects the 1962 vocabulary for what is now called a smooth commutative algebra of finite type over a field — "regular" in the sense of regular local rings, "affine" in the sense of finite-type over the base field.

The theorem was extended to smooth schemes by R. G. Swan in 1996 ("Hochschild cohomology of quasiprojective schemes," Journal of Pure and Applied Algebra 110, 57-80) ^{[source pending]}, who proved the local-to-global identification $H H_{n} (X) ≅ ⨁_{p - q = n} H^{p} (X, Ω_{X / k}^{q})$ for $X$ a smooth quasiprojective $k$ -scheme. The proof reduces the global statement to the affine HKR via the local-to-global Hochschild-to-Hodge spectral sequence, with the degeneration at $E_{2}$ following from the Hodge-to-de Rham degeneration of Deligne-Illusie 1987 Invent. Math. 89 (which uses the Cartier isomorphism in characteristic $p$ and lifts to characteristic zero by a deformation argument). The continuous and formal-completion versions were developed by Amnon Yekutieli (2002 "The continuous Hochschild cochain complex of a scheme," Mathematical Research Letters 9, 433-451) ^{[source pending]}; the derived-category and stacky extensions, together with the Mukai pairing on $H H_{*}$ , were developed by Andrei Caldararu (2003 "The Mukai pairing II: the Hochschild-Kostant-Rosenberg isomorphism," Advances in Mathematics 194, 34-66) ^{[source pending]}.

The Kontsevich formality theorem (2003 "Deformation quantization of Poisson manifolds," Letters in Mathematical Physics 66, 157-216 ^{[source pending]}) is one of the deepest results in derived algebraic geometry and a landmark of mathematical physics. The formality theorem lifts the HKR identification on cohomology to an $L_{\infty}$ -quasi-isomorphism of $L_{\infty}$ -algebras, showing that the Hochschild cochain complex of a smooth commutative algebra in characteristic zero is formal — quasi-isomorphic to its cohomology with zero differential. The deformation-theoretic content: the Maurer-Cartan equation for the polyvector field algebra is the Poisson condition $[π, π]_{S N} = 0$ ; the Maurer-Cartan equation for the Hochschild cochain complex is the associativity of a star product. The formality $L_{\infty}$ -quasi-isomorphism gives a bijection between these two sets of Maurer-Cartan elements, solving the deformation-quantisation problem for arbitrary Poisson manifolds. The proof uses the explicit Kontsevich formality morphism given by integrals over compactified configuration spaces of points in the upper half plane, with weights determined by Kontsevich graphs. An alternative proof using the formality of the little disks operad was given independently by Dmitry Tamarkin in 1998 (arXiv math/9803025 ^{[source pending]}) and reformulated by Vladimir Hinich (2003 Forum Math. 15 ^{[source pending]}); the operadic proof connects the formality theorem to the broader framework of operadic deformation theory and the Deligne conjecture.

The positive-characteristic and derived-algebraic-geometric refinements of HKR have been a central theme of arithmetic algebraic geometry over the past two decades. Bhargav Bhatt 2012 Compositio Mathematica 148 established the comparison between Hochschild homology and the derived de Rham complex in positive characteristic; Antieau-Bhatt-Mathew 2020 ^{[source pending]} developed the systematic theory of the conjugate filtration on topological Hochschild homology $THH$ ; and Bhatt-Morrow-Scholze 2019 Publications Mathématiques de l'IHÉS 129 ^{[source pending]} introduced prismatic cohomology as the unifying framework that recovers crystalline, étale, and de Rham cohomology theories in characteristic $p$ and mixed characteristic, with HKR appearing as the characteristic-zero degeneration. The Toën-Vezzosi derived-HKR theorem (2011 Selecta Mathematica 17 ^{[source pending]}) extends the classical HKR formula to derived algebraic geometry, where the cotangent complex $L_{X}$ of Quillen-Illusie replaces the classical Kähler differentials. These refinements show that the classical HKR theorem is the characteristic-zero, smooth, classical-scheme degeneration of a much richer derived-categorical structure that organises algebraic differential forms, Hodge theory, and prismatic cohomology into a single coherent framework.

Bibliography Master

@article{HKR1962,
  author    = {Hochschild, Gerhard and Kostant, Bertram and Rosenberg, Alex},
  title     = {Differential forms on regular affine algebras},
  journal   = {Transactions of the American Mathematical Society},
  volume    = {102},
  number    = {3},
  pages     = {383--408},
  year      = {1962}
}

@book{Loday1992,
  author    = {Loday, Jean-Louis},
  title     = {Cyclic Homology},
  publisher = {Springer-Verlag},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {301},
  year      = {1992},
  edition   = {2}
}

@book{Weibel1994,
  author    = {Weibel, Charles A.},
  title     = {An Introduction to Homological Algebra},
  publisher = {Cambridge University Press},
  series    = {Cambridge Studies in Advanced Mathematics},
  volume    = {38},
  year      = {1994}
}

@book{GelfandManinAlgebraV,
  author    = {Gel'fand, Sergei I. and Manin, Yuri I.},
  title     = {Homological Algebra},
  publisher = {Springer-Verlag},
  series    = {Encyclopaedia of Mathematical Sciences},
  volume    = {38},
  year      = {1994},
  note      = {Algebra V; English translation of the 1989 VINITI Russian original}
}

@article{Kontsevich2003,
  author    = {Kontsevich, Maxim},
  title     = {Deformation quantization of Poisson manifolds},
  journal   = {Letters in Mathematical Physics},
  volume    = {66},
  number    = {3},
  pages     = {157--216},
  year      = {2003}
}

@article{Swan1996,
  author    = {Swan, Richard G.},
  title     = {Hochschild cohomology of quasiprojective schemes},
  journal   = {Journal of Pure and Applied Algebra},
  volume    = {110},
  number    = {1},
  pages     = {57--80},
  year      = {1996}
}

@article{Yekutieli2002,
  author    = {Yekutieli, Amnon},
  title     = {The continuous {H}ochschild cochain complex of a scheme},
  journal   = {Mathematical Research Letters},
  volume    = {9},
  number    = {4},
  pages     = {433--451},
  year      = {2002}
}

@article{Caldararu2003,
  author    = {C{\u{a}}ld{\u{a}}raru, Andrei},
  title     = {The {M}ukai pairing {II}: the {H}ochschild-{K}ostant-{R}osenberg isomorphism},
  journal   = {Advances in Mathematics},
  volume    = {194},
  number    = {1},
  pages     = {34--66},
  year      = {2003}
}

@article{Tamarkin1998,
  author    = {Tamarkin, Dmitry E.},
  title     = {Another proof of {M}. {K}ontsevich formality theorem},
  journal   = {arXiv preprint math/9803025},
  year      = {1998}
}

@article{Hinich2003,
  author    = {Hinich, Vladimir},
  title     = {Tamarkin's proof of {K}ontsevich formality theorem},
  journal   = {Forum Mathematicum},
  volume    = {15},
  number    = {4},
  pages     = {591--614},
  year      = {2003}
}

@book{Illusie1971,
  author    = {Illusie, Luc},
  title     = {Complexe Cotangent et D{\'e}formations {I}},
  publisher = {Springer-Verlag},
  series    = {Lecture Notes in Mathematics},
  volume    = {239},
  year      = {1971}
}

@book{Illusie1972,
  author    = {Illusie, Luc},
  title     = {Complexe Cotangent et D{\'e}formations {II}},
  publisher = {Springer-Verlag},
  series    = {Lecture Notes in Mathematics},
  volume    = {283},
  year      = {1972}
}

@article{DeligneIllusie1987,
  author    = {Deligne, Pierre and Illusie, Luc},
  title     = {Rel{\`e}vements modulo $p^2$ et d{\'e}composition du complexe de de {R}ham},
  journal   = {Inventiones Mathematicae},
  volume    = {89},
  pages     = {247--270},
  year      = {1987}
}

@article{ToenVezzosi2011,
  author    = {To{\"e}n, Bertrand and Vezzosi, Gabriele},
  title     = {Alg{\`e}bres simpliciales {$S^1$}-{\'e}quivariantes, th{\'e}orie de de {R}ham et th{\'e}or{\`e}mes {HKR} multiplicatifs},
  journal   = {Compositio Mathematica},
  volume    = {147},
  number    = {6},
  pages     = {1979--2000},
  year      = {2011}
}

@article{AntieauBhattMathew2020,
  author    = {Antieau, Benjamin and Bhatt, Bhargav and Mathew, Akhil},
  title     = {Counterexamples to {H}ochschild-{K}ostant-{R}osenberg in characteristic $p$},
  journal   = {Forum of Mathematics, Sigma},
  volume    = {9},
  pages     = {e49},
  year      = {2021}
}

@article{BhattMorrowScholze2019,
  author    = {Bhatt, Bhargav and Morrow, Matthew and Scholze, Peter},
  title     = {Topological {H}ochschild homology and integral $p$-adic {H}odge theory},
  journal   = {Publications Math{\'e}matiques de l'IH{\'E}S},
  volume    = {129},
  pages     = {199--310},
  year      = {2019}
}

@article{Bhatt2012,
  author    = {Bhatt, Bhargav},
  title     = {$p$-adic derived de {R}ham cohomology},
  journal   = {Compositio Mathematica},
  volume    = {148},
  number    = {3},
  pages     = {633--663},
  year      = {2012}
}

@incollection{KontsevichSoibelman2009,
  author    = {Kontsevich, Maxim and Soibelman, Yan},
  title     = {Notes on $A_\infty$-algebras, $A_\infty$-categories and non-commutative geometry {I}},
  booktitle = {Homological Mirror Symmetry},
  series    = {Lecture Notes in Physics},
  volume    = {757},
  pages     = {153--219},
  year      = {2009},
  publisher = {Springer}
}

@article{McClureSmith2002,
  author    = {McClure, James E. and Smith, Jeffrey H.},
  title     = {A solution of {D}eligne's {H}ochschild cohomology conjecture},
  journal   = {Contemporary Mathematics},
  volume    = {293},
  pages     = {153--193},
  year      = {2002}
}

@article{NikolausScholze2018,
  author    = {Nikolaus, Thomas and Scholze, Peter},
  title     = {On topological cyclic homology},
  journal   = {Acta Mathematica},
  volume    = {221},
  number    = {2},
  pages     = {203--409},
  year      = {2018}
}

Prerequisites

04.03.20
04.08.01

Tier anchors

beginner: Loday *Cyclic Homology* §3.1 informal statement of HKR for the polynomial algebra; Weibel *An Introduction to Homological Algebra* §9.4 worked example
intermediate: Weibel *An Introduction to Homological Algebra* Ch. 9 §4; Loday *Cyclic Homology* Ch. 3; Gelfand-Manin *Homological Algebra* (Algebra V) Ch. V §3
master: Hochschild-Kostant-Rosenberg *Trans. AMS* 102 (1962); Loday *Cyclic Homology* (Springer Grundlehren 301, 1992; 2nd ed. 1998); Swan *J. Pure Appl. Algebra* 110 (1996); Yekutieli *Math. Res. Lett.* 9 (2002); Caldararu *Adv. Math.* 194 (2003); Kontsevich *Lett. Math. Phys.* 66 (2003) — formality theorem; Tamarkin 1998 / Hinich 1999 — alternative proofs of formality; Avramov-Iyengar on the non-smooth case

References

Hochschild-Kostant-Rosenberg — Differential forms on regular affine algebras · *Transactions of the American Mathematical Society* 102 (1962), 383--408; the original HKR theorem identifying $HH_n(A, A) \cong \Omega^n_{A/k}$ for a smooth commutative $k$-algebra $A$ of finite type in characteristic zero; the antisymmetrisation map $\varepsilon_n : \Omega^n_{A/k} \to C_n(A, A)$ sending $a_0 \, da_1 \wedge \cdots \wedge da_n \mapsto (1/n!) \sum_\sigma \mathrm{sgn}(\sigma) a_0 \otimes a_{\sigma(1)} \otimes \cdots \otimes a_{\sigma(n)}$ is a quasi-isomorphism; the smoothness hypothesis and characteristic-zero hypothesis are both essential
Loday — Cyclic Homology · Springer Grundlehren der mathematischen Wissenschaften 301 (1st ed. 1992; 2nd ed. 1998), $\approx 516$ pp.; Ch. 3 §3.4 (the HKR theorem and the antisymmetrisation quasi-isomorphism); Ch. 3 §3.2 (Koszul resolution of a smooth commutative algebra and the explicit computation for the polynomial algebra); Ch. 3 §3.5 (the failure of HKR in positive characteristic and the role of the derived de Rham complex); Ch. 5 (Gerstenhaber algebra structure and the Schouten-Nijenhuis bracket on polyvector fields)
Weibel — An Introduction to Homological Algebra · Cambridge Studies in Advanced Mathematics 38 (1994); Ch. 9 §4 (the HKR theorem statement, proof for the polynomial algebra via the Koszul resolution, and the localisation argument for general smooth algebras); Ch. 9 §5 (Connes' SBI long exact sequence and the cyclic-homology extension)
Gelfand-Manin — Homological Algebra (Algebra V) · Encyclopaedia of Mathematical Sciences Vol. 38 (Springer 1994; English translation of the 1989 VINITI Russian original); Ch. V §3 (Hochschild homology of a smooth commutative algebra; the HKR theorem; antisymmetrisation; bridge to cyclic homology and the noncommutative de Rham picture)
Kontsevich — Deformation quantization of Poisson manifolds · *Letters in Mathematical Physics* 66 (2003), 157--216; the formality theorem identifying the Hochschild cochain complex $C^*(C^\infty(M))$ of the function algebra of a smooth manifold $M$ with the polyvector field algebra $T^*_{\mathrm{poly}}(M)$ (with Schouten-Nijenhuis bracket and zero differential) as $L_\infty$-algebras; the formality theorem solves the deformation-quantisation problem for arbitrary Poisson manifolds, producing an associative star product deforming every Poisson structure; the proof uses the explicit Kontsevich formality morphism given by integrals over compactified configurations of points in the upper half plane with weights determined by Kontsevich graphs
Swan — Hochschild cohomology of quasiprojective schemes · *Journal of Pure and Applied Algebra* 110 (1996), 57--80; the sheafified HKR theorem for smooth quasiprojective $k$-schemes: $HH_n(X) \cong \bigoplus_{p - q = n} H^p(X, \Omega^q_{X/k})$ identifying Hochschild homology with the Hodge cohomology of $X$
Yekutieli — The continuous Hochschild cochain complex of a scheme · *Mathematical Research Letters* 9 (2002), 433--451; sheafified and formal-completion versions of HKR; the continuous Hochschild cochain complex appropriate for analytic and adic settings
Caldararu — The Mukai pairing II: the Hochschild-Kostant-Rosenberg isomorphism · *Advances in Mathematics* 194 (2003), 34--66; the Mukai pairing on Hochschild homology of a smooth projective variety and the HKR isomorphism in the derived-category setting; stacky and derived-noncommutative-geometric extensions
Tamarkin — Another proof of M. Kontsevich formality theorem · arXiv preprint math/9803025 (1998); alternative proof of the Kontsevich formality theorem using the formality of the little disks operad and the Etingof-Kazhdan quantisation of Lie bialgebras; foundational input to the operadic interpretation of HKR and the Deligne conjecture
Hinich — Tamarkin's proof of Kontsevich formality theorem · *Forum Mathematicum* 15 (2003), 591--614; reformulation and exposition of Tamarkin's proof of the Kontsevich formality theorem in the language of operads and homotopy Lie algebras; clarifies the role of $E_2$-operad formality in the deformation-quantisation problem
Stacks Project — Hochschild-Kostant-Rosenberg and the cotangent complex · <https://stacks.math.columbia.edu/tag/0FKE> (smoothness via the cotangent complex and Kähler differentials); <https://stacks.math.columbia.edu/tag/08RB> (Naive cotangent complex); foundational background for HKR via André-Quillen homology and the cotangent complex framework

Estimated time

beginner: 18m
intermediate: 42m
master: 75m