04.08.01 · algebraic-geometry / differentials

Sheaf of differentials

shipped3 tiersLean: partial

Anchor (Master): Grothendieck-Dieudonné EGA IV; Liu Ch. 6; Hartshorne §II.8

Intuition [Beginner]

The sheaf of differentials is the algebraic-geometry analogue of " $df$ " — the differential of a function — packaged into a sheaf that lives on a scheme. On the affine line $A^{1} = Spec k [x]$ , the differential of the function $x^{2}$ is $2 x d x$ . The sheaf of differentials records all such " $df$ " data globally, gluing local pieces consistently.

Why bother building this algebraically? Because algebraic geometry must work over any field, including those where calculus-style limits don't exist (finite fields, $p$ -adic fields). Erich Kähler's 1939 Algebra und Differentialrechnung constructed a purely algebraic version of "differentials" using only the ring structure: a Kähler differential $Ω_{R / S}^{1}$ is generated by symbols $df$ for $f \in R$ , subject to the Leibniz and additivity rules.

Grothendieck globalised this in EGA IV (1967): for any morphism of schemes $f : X \to Y$ , the relative sheaf $Ω_{X / Y}^{1}$ is the universal target for $O_{Y}$ -linear derivations on $O_{X}$ . It is the foundation for the canonical sheaf, Serre duality, and Riemann-Roch.

Visual [Beginner]

A scheme with cotangent vectors at each point — directions in which a function can change — assembled into a sheaf of $O$ -modules.

Worked example [Beginner]

On the affine line $X = Spec k [x]$ , the sheaf of differentials $Ω_{X / k}^{1}$ is the free rank-1 module generated by the symbol $d x$ . So $Ω_{X / k}^{1} = k [x] \cdot d x$ as a $k [x]$ -module. The differential of any polynomial $f (x) \in k [x]$ is computed by the formal rule

df = f^{'} (x) d x,

where $f^{'}$ is the formal derivative. For $f (x) = x^{3} + 2 x$ : $df = (3 x^{2} + 2) d x$ . The " $df$ " data is a single coefficient times the formal generator $d x$ .

On $A^{2} = Spec k [x, y]$ , the sheaf of differentials is the free rank-2 module generated by $d x, d y$ : $Ω_{A^{2} / k}^{1} = k [x, y] d x \oplus k [x, y] d y$ . The differential of $f (x, y) = x^{2} y + y^{3}$ is

df = 2 x y d x + (x^{2} + 3 y^{2}) d y .

The familiar partial-derivative formula, now framed algebraically as a $k$ -linear derivation $d : k [x, y] \to Ω_{A^{2} / k}^{1}$ .

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $f : X \to Y$ be a morphism of schemes. The sheaf of relative differentials $Ω_{X / Y}^{1}$ is the quasi-coherent $O_{X}$ -module representing $O_{Y}$ -linear derivations on $O_{X}$ .

Affine local construction. For a ring map $ϕ : S \to R$ , the module of Kähler differentials $Ω_{R / S}^{1}$ is the $R$ -module generated by symbols ${d r : r \in R}$ subject to the relations:

$d (r_{1} + r_{2}) = d r_{1} + d r_{2}$ (additivity)
$d (r_{1} r_{2}) = r_{1} \cdot d r_{2} + r_{2} \cdot d r_{1}$ (Leibniz)
$d s = 0$ for $s \in image (ϕ)$ ( $S$ -linearity)

The map $d : R \to Ω_{R / S}^{1}$ , $r \mapsto d r$ , is a derivation — an $S$ -linear map satisfying the Leibniz rule. Universally: any $S$ -linear derivation $δ : R \to M$ into an $R$ -module $M$ factors uniquely through $d$ as $δ = D \circ d$ for an $R$ -module map $D : Ω_{R / S}^{1} \to M$ .

Equivalently, $Ω_{R / S}^{1} = I / I^{2}$ where $I \subset R \otimes_{S} R$ is the kernel of the multiplication $R \otimes_{S} R \to R$ , with $d r := r \otimes 1 - 1 \otimes r (mod I^{2})$ .

Globalisation. For a morphism of schemes $f : X \to Y$ , the sheaf $Ω_{X / Y}^{1}$ is the unique quasi-coherent $O_{X}$ -module whose sections on an affine open $Spec (R) \subset X$ mapping to an affine open $Spec (S) \subset Y$ equal $Ω_{R / S}^{1}$ . Equivalently, $Ω_{X / Y}^{1} = Δ^{*} (I / I^{2})$ where $Δ : X \to X \times_{Y} X$ is the diagonal and $I$ its ideal sheaf.

The universal derivation $d : O_{X} \to Ω_{X / Y}^{1}$ is the sheafification of the affine $d$ maps; it is $f^{- 1} O_{Y}$ -linear and satisfies the Leibniz rule.

Properties.

(D1) Base change. For $g : Y^{'} \to Y$ and $X^{'} = X \times_{Y} Y^{'}$ : $Ω_{X^{'} / Y^{'}}^{1} = g^{*} Ω_{X / Y}^{1}$ .

(D2) First fundamental exact sequence. For $X f Y g Z$ :

f^{*} Ω_{Y / Z}^{1} \to Ω_{X / Z}^{1} \to Ω_{X / Y}^{1} \to 0.

The sequence is exact at the right-hand terms; the leftmost map is injective when $f$ is smooth.

(D3) Second fundamental exact sequence. For a closed immersion $i : Z ↪ X$ over $Y$ with ideal sheaf $I_{Z}$ :

I_{Z} / I_{Z}^{2} d i^{*} Ω_{X / Y}^{1} \to Ω_{Z / Y}^{1} \to 0.

The leftmost map is the conormal sheaf, and injectivity holds when $Z \subset X$ is regularly embedded.

(D4) Smooth case. If $f : X \to Y$ is smooth of relative dimension $n$ , then $Ω_{X / Y}^{1}$ is locally free of rank $n$ . The dual is the relative tangent sheaf $T_{X / Y}$ .

(D5) Affine space. $Ω_{A_{S}^{n} / S}^{1} = O_{A_{S}^{n}} \cdot d x_{1} \oplus \dots \oplus O_{A_{S}^{n}} \cdot d x_{n}$ .

(D6) Polynomial relations. For $X = V (f_{1}, \dots, f_{r}) \subset A^{n}$ , the conormal sequence reads

O_{X}^{r} [\partial f_{i} / \partial x_{j}] O_{X}^{n} \to Ω_{X / k}^{1} \to 0

with the Jacobian matrix of partial derivatives.

Key theorem with proof [Intermediate+]

Theorem (universal property of the sheaf of differentials). Let $f : X \to Y$ be a morphism of schemes. The pair $(Ω_{X / Y}^{1}, d : O_{X} \to Ω_{X / Y}^{1})$ represents the functor

Der_{Y} (O_{X}, -) : QCoh (X) \to Set, F \mapsto Der_{f^{- 1} O_{Y}} (O_{X}, F),

assigning to a quasi-coherent sheaf $F$ the set of $f^{- 1} O_{Y}$ -linear derivations $O_{X} \to F$ . That is: every derivation $δ : O_{X} \to F$ factors uniquely as $δ = D \circ d$ for a unique $O_{X}$ -linear $D : Ω_{X / Y}^{1} \to F$ .

Proof. The result is local on $X$ and $Y$ , so we reduce to the affine case $X = Spec (R) \to Y = Spec (S)$ .

Step 1 — affine universal property. For a ring map $S \to R$ and an $R$ -module $M$ , an $S$ -linear derivation $δ : R \to M$ is an $S$ -linear map satisfying $δ (r_{1} r_{2}) = r_{1} δ (r_{2}) + r_{2} δ (r_{1})$ .

By construction of $Ω_{R / S}^{1}$ as the $R$ -module with generators ${d r}$ modulo additivity, Leibniz, and $S$ -linearity relations: any $S$ -linear derivation $δ : R \to M$ defines an $R$ -linear map $D : Ω_{R / S}^{1} \to M$ by $D (d r) := δ (r)$ . The Leibniz and additivity relations on $Ω_{R / S}^{1}$ are exactly those satisfied by $δ$ , so $D$ is well-defined. Conversely, any $R$ -linear $D : Ω_{R / S}^{1} \to M$ composes with $d$ to give a derivation $δ = D \circ d$ . The two operations are inverse, so $Hom_{R} (Ω_{R / S}^{1}, M) = Der_{S} (R, M)$ .

Step 2 — gluing. The construction is functorial in $S \to R$ and respects localisation: $Ω_{R [f^{- 1}] / S}^{1} = R [f^{- 1}] \otimes_{R} Ω_{R / S}^{1}$ . So local affine pieces glue to a quasi-coherent sheaf $Ω_{X / Y}^{1}$ on $X$ , and the universal property holds at the level of sheaves: $Hom_{O_{X}} (Ω_{X / Y}^{1}, F) = Der_{f^{- 1} O_{Y}} (O_{X}, F)$ for any quasi-coherent $F$ .

Step 3 — derivation $d$ . The universal derivation $d : O_{X} \to Ω_{X / Y}^{1}$ is the sheafification of the affine $d_{R} : R \to Ω_{R / S}^{1}$ . By the universal property at the affine level and the gluing in Step 2, $d$ is itself the universal derivation. $□$

The universal property characterises $Ω_{X / Y}^{1}$ up to canonical isomorphism. It is the categorical embodiment of "the universal place where derivations land."

Bridge. The construction here builds toward later units of the strand, where the same pattern is taken up at higher structure. The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Exercise 4 (medium, proof).

State and prove the conormal sequence for a regular embedding $Z ↪ X$ of smooth varieties.

Hint

The conormal sheaf $N_{Z / X}^{\lor} = I_{Z} / I_{Z}^{2}$ enters as the kernel.

Answer

Let $i : Z ↪ X$ be a regular closed embedding of smooth $k$ -varieties of dimensions $d$ and $n$ respectively, codimension $c = n - d$ . Then the second fundamental sequence is exact:

0 \to I_{Z} / I_{Z}^{2} d i^{*} Ω_{X / k}^{1} \to Ω_{Z / k}^{1} \to 0.

Proof. Right exactness is automatic (D3). For injectivity: locally, the regular embedding means $I_{Z}$ is generated by a regular sequence $f_{1}, \dots, f_{c}$ . Then $I_{Z} / I_{Z}^{2}$ is locally free on $Z$ of rank $c$ , with basis the images of $f_{i}$ . The map $d$ sends $\overset{ˉ}{f}_{i} \mapsto d f_{i} \in i^{*} Ω_{X / k}^{1}$ .

The pullback $i^{*} Ω_{X / k}^{1}$ is locally free of rank $n$ , with basis $d x_{1}, \dots, d x_{n}$ for local parameters on $X$ . The forms $d f_{1}, \dots, d f_{c}$ are linearly independent in this rank- $n$ space (smoothness of the regular embedding implies the Jacobian has full rank $c$ ). The cokernel is a locally free quotient of rank $n - c = d$ , identified with $Ω_{Z / k}^{1}$ .

The left map is injective on the locally free conormal sheaf (no kernel since the Jacobian has rank $c$ ). Hence the sequence is exact. $□$

The dual sequence on tangent sheaves is $0 \to T_{Z} \to i^{*} T_{X} \to N_{Z / X} \to 0$ , with $N_{Z / X}$ the normal sheaf.

Exercise 6 (hard, proof).

Compute $Ω_{X / k}^{1}$ for $X = P_{k}^{n}$ and identify the canonical line bundle $det Ω^{1}$ .

Hint

Use the Euler sequence: $0 \to Ω_{P^{n}}^{1} \to O_{P^{n}} (- 1)^{n + 1} \to O_{P^{n}} \to 0$ .

Answer

The Euler sequence on $P_{k}^{n}$ :

0 \to Ω_{P^{n} / k}^{1} \to O_{P^{n}} (- 1)^{n + 1} \to O_{P^{n}} \to 0.

This identifies $Ω_{P^{n}}^{1}$ as the kernel of the map $O (- 1)^{n + 1} \to O$ given by $(s_{0}, \dots, s_{n}) \mapsto \sum x_{i} s_{i}$ — the "evaluation against the Euler vector field."

Taking determinants of the Euler sequence:

det O (- 1)^{n + 1} = O (- (n + 1)) = det Ω_{P^{n}}^{1} \otimes det O_{P^{n}} = det Ω_{P^{n}}^{1} \otimes O .

So $det Ω_{P^{n}}^{1} = ω_{P^{n}} = O_{P^{n}} (- (n + 1))$ .

This is the foundational computation: the canonical bundle of projective space is $O (- n - 1)$ , hence has degree $- (n + 1)$ along any line. In particular, $ω_{P^{1}} = O (- 2)$ , $ω_{P^{2}} = O (- 3)$ , etc. The negative-degree canonical bundle reflects the Fano (positivity) nature of projective space. $□$

Exercise 7 (hard, conceptual).

Why does the sheaf-of-differentials formalism work over arbitrary base schemes — including arithmetic bases like $Spec (Z)$ — where classical calculus has no obvious meaning?

Hint

The Kähler-differential construction uses only ring-theoretic data, no analytic structure.

Answer

The Kähler differential $Ω_{R / S}^{1}$ is constructed entirely from:

The ring structures on $R$ and $S$ .
The ring map $S \to R$ .

No metric, no topology beyond Zariski, no real or complex structure. The Leibniz rule and additivity are algebraic properties of any ring derivation.

This means $Ω^{1}$ makes sense for:

Schemes over $Spec (Z)$ — arithmetic schemes, where the base is the integers and there's no field-of-real-numbers analogue.
Schemes over $Spec (F_{p})$ — varieties in positive characteristic, where formal differentiation produces unfamiliar phenomena (e.g., $d (x^{p}) = p x^{p - 1} d x = 0$ ).
Formal schemes and adic schemes — Grothendieck and Berkovich's $p$ -adic geometry.
Derived schemes and animated rings — modern higher-categorical algebraic geometry (Lurie).

The functorial / relative point of view — $Ω_{X / Y}^{1}$ as a relative sheaf depending on the morphism, not on absolute notions — is Grothendieck's contribution in EGA IV. The relative perspective makes differentials transform correctly under base change and provides the right framework for arithmetic geometry, where one works over $Spec (Z)$ as the universal base.

This is one of the great triumphs of Grothendieck's reformulation: classical calculus, recovered from pure algebra, generalises to settings where calculus has no meaning. $□$

Lean formalization [Intermediate+]

lean_status: partial — Mathlib has KaehlerDifferential for the algebraic case (Ω[R/S] over commutative rings); the schematic globalisation as a quasi-coherent sheaf is partially formalised.

[object Promise]

Advanced results [Master]

Cotangent complex. The sheaf of differentials $Ω_{X / Y}^{1}$ is the $H^{0}$ of a more refined object — the cotangent complex $L_{X / Y}$ , an object of the derived category of $O_{X}$ -modules. André and Quillen independently introduced the cotangent complex in the 1960s. For local complete intersections, $L_{X / Y}$ has only two non-vanishing cohomologies, related to the conormal sequence. For more general morphisms, all higher $H^{- i} (L)$ can be non-zero, encoding "derived-deformation" data.

Higher differentials and de Rham cohomology. The exterior powers $Ω_{X / k}^{p} = ⋀^{p} Ω_{X / k}^{1}$ form the algebraic de Rham complex $(Ω_{X / k}^{∙}, d)$ . Algebraic de Rham cohomology is $H_{dR}^{*} (X / k) = H^{*} (X, Ω_{X / k}^{∙})$ (hypercohomology). Grothendieck's algebraic de Rham theorem (1966): for smooth complex varieties, algebraic de Rham cohomology equals singular cohomology with $C$ coefficients.

Frobenius and characteristic $p$ . In characteristic $p > 0$ , the Frobenius morphism $F : X \to X^{(p)}$ has $Ω_{X / X^{(p)}}^{1} = 0$ — Frobenius is infinitesimally inert. The Cartier operator $C : H^{p} (Ω^{∙}) \to Ω^{p}$ is an inverse to $F^{*}$ on cohomology, foundational for $p$ -adic Hodge theory.

Hodge-de-Rham spectral sequence. $E_{1}^{p, q} = H^{q} (X; Ω^{p}) \Rightarrow H_{dR}^{p + q} (X)$ . For smooth proper $X$ over a characteristic- $0$ field, the spectral sequence degenerates at $E_{1}$ (the Hodge degeneration theorem) — proved analytically by Hodge for compact Kähler, algebraically by Deligne-Illusie 1987 via reduction mod $p$ .

Crystalline cohomology. Berthelot-Grothendieck's crystalline cohomology (1968–74) is a refinement of de Rham cohomology in positive characteristic, defined via the crystalline site. It is the natural target for Frobenius cohomology operators and powers Berthelot's $p$ -adic Hodge theory.

Synthesis. This construction generalises the pattern fixed in 04.02.01 (scheme), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

The universal property is proved in the formal-definition section. The Euler sequence on $P^{n}$ is proved by direct computation using the homogeneous-coordinate description. The fundamental exact sequences are proved by the standard reduction to affine local ring data (cf. Hartshorne §II.8 Proposition 8.11–8.12, Vakil §21.4–21.5).

Algebraic de Rham theorem (Grothendieck 1966): proved via comparison with analytic de Rham + Serre's GAGA. Hodge degeneration in char 0 (Deligne-Illusie 1987): proved via reduction mod $p$ , lift to characteristic 0, and a Frobenius-twisted decomposition of $Ω^{∙}$ .

Connections [Master]

Scheme 04.02.01 — the sheaf of differentials is the foundational construction internal to scheme theory.
Coherent sheaf 04.06.02 — $Ω_{X / k}^{1}$ is coherent on locally Noetherian schemes; finitely generated module on each affine.
Morphism of schemes 04.02.04 — relativisation: $Ω_{X / Y}^{1}$ depends on the morphism, base-changes correctly.
Canonical sheaf 04.08.02 — $ω_{X} = det Ω_{X}^{1}$ , the top exterior power.
Serre duality 04.08.03 — uses the canonical sheaf $ω_{X}$ as the dualising sheaf.
Riemann-Roch theorem for curves 04.04.01 — the canonical divisor $K_{C}$ is the divisor class of $ω_{C} = Ω_{C}^{1}$ .
De Rham cohomology 03.04.06 — algebraic de Rham cohomology is $H^{*} (X; Ω_{X}^{∙})$ , mirrored by smooth de Rham theory.
Hodge decomposition 04.09.01 — relies on the spectral sequence $H^{q} (Ω^{p}) \Rightarrow H_{dR}^{p + q}$ .

Historical & philosophical context [Master]

The notion of differential in algebra goes back to Erich Kähler's 1939 monograph Algebra und Differentialrechnung. Kähler — the same Kähler whose name graces Kähler manifolds in differential geometry — introduced an algebraic version of " $df$ " generated by abstract symbols modulo Leibniz and linearity rules. His motivation was to extend differential calculus to commutative rings and, more specifically, to ideal-theoretic contexts where classical analysis was inapplicable.

The name Kähler differentials and the systematic construction $Ω_{R / S}^{1}$ as a quotient of $R \otimes_{S} R$ became standard through the work of Cartier, Grothendieck, and Bourbaki in the late 1950s. Grothendieck's EGA IV (1967, Publ. Math. IHES 32, the longest single volume of EGA at over 800 pages) gave the definitive scheme-theoretic treatment: the sheaf $Ω_{X / Y}^{1}$ , the cotangent complex precursors, the smooth/étale/unramified trichotomy, the deformation theory.

Grothendieck's key reformulation was relativity: differentials over a base, $Ω_{X / Y}^{1}$ — not absolute differentials of $X$ . The relative perspective makes the construction functorial in both $X$ and $Y$ , base-change-friendly, and applicable to arithmetic schemes over $Spec (Z)$ . This was decisive for the development of Grothendieck topology, étale cohomology, and the Weil conjectures.

The cotangent complex (Quillen 1967, André 1974) extended the notion further: $L_{X / Y}$ is a complex whose $H^{0}$ is $Ω_{X / Y}^{1}$ and whose higher cohomologies measure non-smoothness obstructions. Lurie's $\infty$ -categorical reformulation (2009 Higher Topos Theory) reads $L$ as a cotangent fibration in derived algebraic geometry, with deformation theory as its principal application.

Conceptually, the sheaf of differentials embodies the infinitesimal viewpoint in algebraic geometry: $Ω^{1}$ records the linear approximation to a scheme around each point, dually to the tangent sheaf $T_{X} = Hom (Ω_{X}^{1}, O_{X})$ . This linearisation framework is the algebraic analogue of differential calculus and powers nearly all of modern algebraic geometry — Riemann-Roch, intersection theory, Hodge theory, deformation theory, mirror symmetry.

Bibliography [Master]

Kähler, E., Algebra und Differentialrechnung, Vandenhoeck & Ruprecht 1953 (originally lectured 1939).
Grothendieck, A. & Dieudonné, J., Éléments de Géométrie Algébrique IV, Publ. Math. IHES 20, 24, 28, 32 (1964–67) — the canonical reference for the relative theory.
Hartshorne, R., Algebraic Geometry, Springer 1977 — §II.8.
Vakil, R., The Rising Sea: Foundations of Algebraic Geometry — §21.
Liu, Q., Algebraic Geometry and Arithmetic Curves, Oxford 2002 — Chapter 6.
Quillen, D., Homology of Commutative Rings, MIT mimeographed notes 1967.
André, M., Homologie des Algèbres Commutatives, Springer GMW 206, 1974.
Illusie, L., Complexe Cotangent et Déformations I–II, Springer LNM 239, 283, 1971–72.
Deligne-Illusie, Relèvements modulo $p^{2}$ et décomposition du complexe de de Rham, Invent. Math. 89 (1987), 247–270.
Berthelot, P. & Ogus, A., Notes on Crystalline Cohomology, Princeton 1978.