.

HIGHER DIRECT IMAGES OF KÄHLER p-DIFFERENTIALS FOR k-DU BOIS

SINGULARITIES

ROBERT FRIEDMAN AND RADU LAZA

Abstract. We prove that the higher direct images Rq f∗ ΩpY/S of the sheaves of relative Kähler
arXiv:2205.04729v1 [math.AG] 10 May 2022

p-differentials are locally free and compatible with arbitrary base change for flat proper families
whose fibers have k-Du Bois local complete intersection singularities, for p ≤ k and all q ≥ 0.
This is a direct generalization, for the special case of local complete intersections, of a result of Du
Bois (case k = 0). As applications, we discuss the behavior of Hodge numbers in families and the
unobstructedness of deformations of singular Calabi-Yau varieties.

1. Introduction
The Hodge numbers are constant in a smooth family of complex projective varieties over a
connected base. A powerful way of encoding this fundamental fact is Deligne’s theorem [Del68]:
If f : Y → S is a smooth morphism of complex projective varieties, then the higher direct image
sheaves Rq f∗ ΩpY/S of the relative Kähler differentials are locally free and compatible with base
change. This theorem fails for families of varieties which have singular fibers (and in positive
characteristic). For Y a singular compact complex algebraic variety, Du Bois [DB81] showed that
the Kähler-de Rham complex Ω•Y should be replaced by the filtered de Rham or Deligne-Du Bois
complex Ω•Y whose graded pieces ΩpY ∈ Dbcoh (Y ) play the role of ΩpY (see [PS08, §7.3]). Namely, the
associated spectral sequence with E1p,q = Hq (Y ; ΩpY ) degenerates at E1 and computes H p+q (Y, C)
together with the Hodge filtration associated to the mixed Hodge structure on H ∗ (Y ). However,
the associated Hodge numbers do not behave well in families in general.
The filtered de Rham complex is related to the complex of Kähler differentials via the canonical
comparison Kähler-to-Du Bois map φp : ΩpY → ΩpY . The maps φp are isomorphisms for all p ⇐⇒
Y is smooth. It is thus natural to consider the case when φp is a quasi-isomorphism in a certain
range. Steenbrink [Ste83, §3] introduced the notion of Du Bois singularities, which play a role in
the study of compactifications of moduli. By definition, Y is Du Bois if φ0 is a quasi-isomorphism.
Following [MOPW21] and [JKSY21], we say that Y is k-Du Bois if φp is a quasi-isomorphism for
0 ≤ p ≤ k. Thus 0-Du Bois singularities are exactly the Du Bois singularities in Steenbrink’s
terminology. A key property satisfied by Du Bois singularities is the following:
Theorem 1.1 (Du Bois [DB81, Thm. 4.6]). Let f : Y → S be a flat proper family of complex
algebraic varieties. Assume that some fiber Ys has Du Bois singularities. Then, possibly after
replacing S by a neighborhood of s, for all q ≥ 0, the sheaves Rq f∗ OY are locally free of finite type
and compatible with base change.
The theorem can be interpreted (in particular) as giving a relation between the mixed Hodge
structure H ∗ (Y0 ) of a singular fiber Y0 with the limit mixed Hodge Hlim ∗ associated to a one-
parameter smoothing Y/∆. Roughly speaking, it identifies the “frontier” of the Hodge diamond
of H i (Y0 ) (the p · q = 0 case) with that of Hlim
i . In this version, the theorem (for dimension 2

slc hypersurface singularities) was independently established by Shah [Sha79], and plays a key role
in the study of degenerations of K3 surfaces (e.g. [Sha80]) and related objects (e.g. [KLSV18]).

Research of the second author is supported in part by NSF grant DMS-2101640.

1
Theorem 1.1 continues to have important consequences for the study of compact moduli of varieties
of general type (see e.g. [Kol21], esp. §2.5 of loc. cit.).
The goal of this note is to prove the following generalization of Theorem 1.1 for the case of a
local complete intersection (lci) morphism:
Theorem 1.2. Let f : Y → S be a flat proper family of complex algebraic varieties and let s ∈ S.
Suppose that the fiber Ys has k-Du Bois lci singularities. Then, possibly after replacing S by a
neighborhood of s, the higher direct image sheaves Rq f∗ ΩpY/S of the relative Kähler differentials are
locally free and compatible with base change for 0 ≤ p ≤ k and all q ≥ 0.
As a corollary, we obtain:
Corollary 1.3. Let f : Y → S be a flat proper family of complex algebraic varieties over an
irreducible base. For s ∈ S, suppose that the fiber Ys has k-Du Bois lci singularities. Then,
for every fiber t such that Yt is smooth, dim GrpF H n (Yt ) = dim GrpF H n (Ys ) for every n and for
0 ≤ p ≤ k. 
In the case of hypersurface singularities, results similar to Corollary 1.3 were obtained by Kerr-
Laza [KL20, Cor. 4.2] and [KL20, Thm. 7.6]. These type of results are for example relevant to the
study of the moduli of cubic fourfolds [Laz10]. Specifically, most of the relevant degenerations of
cubic fourfolds are 1-Du Bois. It follows that one can read off the “Type” of such degeneration
directly in terms of the central fiber. In particular, cubic fourfolds with ADE singularities are
Hodge theoretically stable (i.e. any smoothing of such a cubic will have finite monodromy).
Another application of Theorem 1.2 is the following generalization of the results of Kawamata
[Kaw92], Ran [Ran92], and Tian [Tia92] on the unobstructedness of deformations for nodal Calabi-
Yau varieties in any dimension, where the special case of isolated hypersurface singularities was
established in [FL22, §6]:
Corollary 1.4. Let Y be a canonical Calabi-Yau variety (Definition 4.4) which is additionally a
scheme with 1-Du Bois lci (not necessarily isolated) singularities. Then the functor Def (Y ) is
unobstructed.
A key technical step in establishing Theorem 1.2, where the lci assumption seems crucial, is the
following flatness result for relative Kähler differentials. In contrast, in the case p = 0 corresponding
to Theorem 1.1, the flatness is automatic (by assumption) and does not require lci.
Theorem 1.5. Suppose that f : X → S is a flat lci morphism, where S is an arbitrary base space.
Let s ∈ S and suppose that codimXs (Xs )sing ≥ 2k + 1, where Xs = f −1 (s) is the fiber of f over s.
Then, possibly after replacing S by a neighborhood of s, the sheaf of relative differentials ΩpX /S is
flat over S for all p ≤ k. For p = 0, 1, ΩpX /S is flat over S with no assumption on the dimension
of the singular locus.
A brief description of the contents of this paper is as follows. After a quick review of the
definition and the basic facts about higher Du Bois singularities (following [MOPW21], [JKSY21],
and [FL22, §3]) and examples illustrating some of the issues occurring in the singular case, we prove
Theorem 1.5 regarding the flatness of the relative Kähler differentials. After these preliminaries, we
establish Theorem 1.2 and Corollary 1.4. Our argument is close to the original argument ([DBJ74,
Lemma 1]) used to establish Theorem 1.1, following a suggestion of J. Kollár.
Finally, we conclude this introduction with two open questions:
(1) Can the lci assumption be dropped in Theorem 1.2?
(2) There is a companion definition to that of k-Du Bois singularities, namely the k-rational
singularities defined in [KL20, §4], [FL22, §3]. Is there a result analogous to Theorem 1.2
for k-rational singularities?
2
Notations and Conventions. We work in the complex algebraic category (see Remark 4.3 for
some possible generalizations). Following our conventions from previous work [FL22], we use X
(and X ) in the local context (e.g. statements such as Theorem 1.5 are purely local, and hold in
the analytic case as well), whereas Y and π : Y → S are meant to be proper over Spec C or S
respectively (e.g. in Theorem 1.2 where the properness is essential).

Acknowledgement. We have benefited from discussions and correspondence with J. Kollár, M.

Mustaţă, and M. Popa on higher du Bois singularities while preparing [FL22]. The second author
had several related discussions with M. Kerr and M. Saito while preparing previous joint work.

2. Preliminaries
2.1. The filtered de Rham complex. On a smooth scheme Y proper over Spec C or a compact
Kähler manifold, the Hodge-de Rham spectral sequence with E1 page E1p,q = H q (Y ; ΩpY ) degen-
erates at E1 and computes the Hodge structure on H p+q (Y, C). For X not necessarily smooth or
proper over Spec C, Deligne showed that H ∗ (X, C) carries a canonical mixed Hodge structure. Sub-
sequently, Du Bois [DB81] introduced an object Ω•X = (Ω•X , F • Ω•X ) in the filtered derived category
whose graded pieces ΩpX = GrpF Ω•X are analogous to ΩpX in the smooth case (see [PS08, §7.3]). Since
ΩpX is defined locally in the étale topology, it agrees with ΩpX at smooth points. Near the singular
b → X is a log resolution with E ⊆ X
locus, if π : X b the reduced exceptional divisor, Ω• is closely
X
related to Rπ∗ Ω b (log E)(−E). More precisely, a relative version Ω•(X,Σ) of Ω•X is quasi-isomorphic
•
X
to Rπ∗ Ω•b (log E)(−E), where Σ is the singular locus of X; see [PS08, Ex. 7.25].
X
In the proper case, there is the following fundamental result of Du Bois, based on Deligne’s
construction of the mixed Hodge structure on Y :
Theorem 2.1 ([DB81]). If Y is proper over Spec C, the spectral sequence with E1 page Hq (Y ; ΩpY ) =⇒
Hp+q (Y ; Ω•Y ) = H p+q (Y ; C) degenerates at E1 and the corresponding filtration on H ∗ (Y ; C) is the
Hodge filtration associated to the mixed Hodge structure on H ∗ (Y ; C).
As a consequence, we have:
Corollary 2.2. With notation as above,
(i) The natural map H i (Y ; C) → Hi (Y ; Ω•Y /F k Ω•Y ) is surjective for all i and k.
(ii) The spectral sequence with E1 term
(
p,q Hq (Y ; ΩpY ), for p ≤ q;
E1 =
0, for p > k.

converging to Hp+q (Y ; Ω•Y /F k Ω•Y ) degenerates at E1 .

Proof. This is a consequence of the following general fact: If (C • , d, F • C • ) is a filtered complex,
then the associated spectral sequence degenerates at the E1 page ⇐⇒ the differential d is strict
with respect to the filtration F • C • . Then, assuming E1 -degeneration, an easy argument shows that
H i (C • , d) → H i (C • /F k C • , d) is surjective for every i and k. Since the induced filtration on C • /F k C •
is also strict with respect to d, the corresponding spectral sequence for the complex C • /F k C • also
degenerates at E1 . 

2.2. Du Bois and higher Du Bois singularities. There is natural comparison map φ• :
(Ω•X , σ • ) → (Ω•X , F • Ω•X ) [PS08, p.175], where σ • is the trivial or naive filtration on Ω•X . Following
[MOPW21] and [JKSY21], one defines:
3
Definition 2.3. Let X be a complex algebraic variety. Then X is k-Du Bois if the natural maps
φp : ΩpX → ΩpX
are quasi-isomorphisms for 0 ≤ p ≤ k. Note that the case k = 0 coincide with the usual definition
of Du Bois singularities [PS08, Def. 7.34].
d
Example 2.4. Let f (z1 , . . . , zn+1 ) = z1d1 + · · · + zn+1
n+1
define the weighted homogeneous singularity
n+1
X 1
X = V (f ) ⊆ A n+1 . Then X is k-Du Bois at 0 ⇐⇒ ≥ k + 1. In particular, an ordinary
di
i=1 
n−1
double point of dimension n is k-Du Bois for all k ≤ . Thus an ordinary double point of
2
dimension 3 is 1-Du Bois, and in fact is the unique 1-Du Bois hypersurface singularity in dimension
3 (e.g. [NS95, Thm. 2.2]).
A key fact that we shall need is the following:
Theorem 2.5 ([MP21], [MOPW21], [JKSY21]). Let X be a complex algebraic variety with lci
singularities. Assume X is k-Du Bois with k ≥ 1. Then X is normal and regular in codimension
2k, i.e. codim Xsing ≥ 2k + 1.
Proof. The normality claim is [MP21, Cor. 13.5]. For hypersurface singularities, various dimension
bounds (covering the claim of the theorem) were obtained in both [MOPW21] and [JKSY21]. The
lci case follows from [MP21, Thm. F] (numerical characterization of k-Du Bois) and [MP21, Cor.
9.26] (bounds on the dim Xsing in terms of the relevant numerical invariant). 
2.3. Kähler differentials on lci k-Du Bois spaces. It is well known that Kähler p-differentials
on singular spaces can have torsion. For instance this is already case for Ω1C where C is a nodal
curve. However, we have the following:
Theorem 2.6 (Greuel [Gre75, Lemma 1.8]). Suppose that X is an lci singularity of dimension n
and that dim Xsing ≤ d. Then, for all x ∈ X, depthx ΩpX ≥ n − p for p ≤ n − d. More generally, let
f : X → S be a flat lci morphism of relative dimension n over a smooth base S and let Xcrit denote
the critical locus of f , i.e. the points of X where f is not a smooth morphism. If dim S = m and
the relative dimension of Xcrit is at most d then, for every x ∈ X , then depthx ΩpX /S ≥ dim X − p =
n + m − p for p ≤ n − d.
Remark 2.7. (i) Using Theorem 2.5, we immediately recover the following previously known
results:
(1) If X has hypersurface k-Du Bois singularities, then ΩpX is torsion free for 0 ≤ p ≤ k
([JKSY21, Prop. 2.2]).
(2) If X has lci k-Du Bois singularities, then ΩpX is reflexive for 0 ≤ p ≤ k ([MP21, Cor. 13.5]).
(ii) For p = 1 and lci singularities, the situation is well understood by a result of Kunz [Kun86,
Prop. 9.7]: Ω1X satisfies Serre’s condition Sa ⇐⇒ X satisfies Ra (i.e. is regular in codimension a).
Remark 2.8. For k ≥ 2, at least some assumption is needed for the hypotheses in Theorem 1.5.
For example, if X has simple normal crossings singularities, say X is the product of a nodal curve
Spec C[x, y]/(xy) with a smooth manifold of dimension n − 1 and X is the standard smoothing over
A1 given by xy = t, then depthx Ω1X = n at a singular point and Ω2X has torsion in the fiber over
the singular fiber at t = 0, hence is not flat over A1 .
In the same spirit, if π : C → ∆ is a family of smooth projective curves acquiring a single
ordinary double point over 0, with the same local picture as above, the sheaf Ω1C/∆ of relative
Kähler differentials is flat over ∆. However, if C0 is the fiber over 0, the Hodge spectral sequence
4
H q (C0 ; ΩpC0 ) =⇒ Hp+q (C0 ; Ω•C0 ) = H p+q (C0 ; C) does not degenerate at E1 . If C0 is reducible,
the values of h0 (Ω1Ct ) and h1 (Ω1Ct ) jump up at t = 0. Hence R1 π∗ Ω1C/∆ has torsion at 0, and in
particular is not locally free.

3. Flatness of the sheaves of relative Kähler differentials

In this section, we prove the flatness result, Theorem 1.5. Before giving the proof, we note the
following corollary of Greuel’s theorem (Theorem 2.6):
Lemma 3.1. Suppose that f : X → S is a flat lci morphism of relative dimension n over a smooth
base S of dimension m, and that dim(Xt )sing ≤ d, with n − d ≥ 2k + 1 for some integer k ≥ 1 and
every point t ∈ S. Let Xcrit denote the critical locus of f , i.e. the points of X where f is not a
smooth morphism. Then, for every x ∈ X and p ≤ n − d,
depthx ΩpX /S ≥ d + m + 2,

and, for all p ≤ k and every open subset U of X , the restriction map
H 0 (U ; ΩpX /S |U ) → H 0 (U − Xcrit ; ΩpX /S |U − Xcrit )

is an isomorphism.
Proof. By assumption, dim Xcrit ≤ d + m. Note that n − d ≥ 2k + 1 ≥ k, and hence, if p ≤ k, then
p ≤ n − d. Thus Theorem 2.6 implies that, for all x ∈ X ,
depthx ΩpX /S ≥ n + m − p ≥ n − k + m ≥ d + m + k + 1 ≥ d + m + 2 ≥ dim Xcrit + 2.

Then H 0 (U ; ΩpX /S |U ) → H 0 (U − Xcrit ; ΩpX /S |U − Xcrit ) is an isomorphism, by a theorem due in the

analytic case to Scheja [Sch64], [Gre75, Satz 1.2]. 
Proposition 3.2. Suppose that f : X → S is a flat lci morphism of relative dimension n over a
smooth base S of dimension m, and that dim(Xt )sing ≤ d, with n − d ≥ 2k + 1 for some integer
k ≥ 1 and every point t ∈ S. Suppose that X ⊆ A × S, where A is smooth, and let I be the defining
V
ideal of X in A × S. Then there exists a filtration of k π1∗ Ω1A |X = π1∗ ΩkA |X by coherent subsheaves
Ka , 0 ≤ a ≤ k, such that K0 = π1∗ ΩkA |X , depthx Ka ≥ d + m + 2 for all x ∈ X , and, for all a ≥ 0,
a
^
Ka /Ka+1 ∼
= (I/I 2 ) ⊗ ΩX
k−a
/S .

Proof. The proof is by induction on k. The case k = 1 is the conormal sequence

0 → I/I 2 → π1∗ Ω1A |X → Ω1X /S → 0,

where we let K1 be the image of I/I 2 . In general, for all 0 ≤ a ≤ k, let Ka be the image of
Va k−a
(I/I 2 ) ⊗ π1∗ ΩA |X in π1∗ ΩkA |X . Thus {Ka } is the usual Koszul filtration, and the proposition is
V
clear over all points of X − Xcrit . Moreover, Kk is either 0 or k (I/I 2 ). By hypothesis k < n =
V
rank Ω1X /S . Let r = rank I/I 2 . We can clearly assume that r ≥ 1. If r ≥ k, then Kk = k (I/I 2 ). If
V
r < k, then Kk = 0, and the largest p such that Kp 6= 0 is Kr , the image of r (I/I 2 ) ⊗ π1∗ ΩA k−r
|X .
Vr 2 2 ∗ k−r−1 Vr 2 ∗ k−r
In this case, the image of (I/I ) ⊗ (I/I ) ⊗ π1 ΩA |X in (I/I ) ⊗ π1 ΩA |X maps to 0 By
the inductive hypothesis, the sequence
r
^ r
^ r
^
k−r−1 k−r k−r
(I/I 2 ) ⊗ (I/I 2 ) ⊗ π1∗ ΩA |X → (I/I 2 ) ⊗ π1∗ ΩA |X → (I/I 2 ) ⊗ ΩX /S
5
 V
is exact, so there is an induced map ϕr : r (I/I 2 ) ⊗ ΩX
k−r ∗ k
/S → π1 ΩA |X . Then the image of ϕr is
contained in the torsion free sheaf Kr and is equal to Kr over X − Xcrit . Moreover ϕr is injective
over X − Xcrit . By the inductive hypothesis on k,
r
^
k−r
depthx (I/I 2 ) ⊗ ΩX /S ≥ d + m + 2,
V k−r ∼ r
and hence Im ϕr = Kr , and ϕr : r (I/I 2 ) ⊗ ΩX r
/S = K = K /K
r+1 is an isomorphism.

For a general a, there is a commutative diagram

V V
0 −−−−→ J −−−−→ a (I/I 2 ) ⊗ π1∗ ΩA k−a
|X −−−−→ a (I/I 2 ) ⊗ ΩX k−a
/S −
−−−→ 0
 
 ϕa
y y
0 −−−−→ Ka+1 −−−−→ Ka −−−−→ Ka /Ka+1 −−−−→ 0,
V V
where by definition J = Ker{ a (I/I 2 ) ⊗ π1∗ ΩA k−a
|X → a (I/I 2 ) ⊗ ΩX k−a
/S } and ϕa , which exists and
is an isomorphism over X − Xcrit , is yet to be constructed over all of X .
For a fixed k, arguing by descending induction on a, and starting either at k or at r, we can
assume that depthx Ka+1 ≥ d + m + 2. Moreover, Ka+1 |X − Xcrit is a subbundle of Ka |X − Xcrit ,
so the torsion subsheaf of Ka /Ka+1 is supported on Xcrit . If σ is a section of Ka over some open
subset U and σ|V ∈ Ka+1 for an open subset V = U − W of U , where W is an analytic subset of
U , then we may assume that W = Xcrit . Since Ka is torsion free and depthx Ka+1 ≥ d + m + 2,
σ ∈ Ka+1 (U ). Thus Ka /Ka+1 is torsion free.
V V
The kernel I of a (I/I 2 ) ⊗ π1∗ ΩA
k−a
|X → Ka /Ka+1 is equal to J = Ker{ a (I/I 2 ) ⊗ π1∗ ΩA k−a
|X →
Va 2 k−a V a 2 k−a
(I/I ) ⊗ ΩX /S } over X − Xcrit . Moreover, depthx J ≥ depthx (I/I ) ⊗ ΩX /S ≥ d + m + 2.
Thus I ⊆ J , and they are equal over X − Xcrit . Since K /K a a+1 is torsion free, they are equal and
there is an induced injective
V homomorphism ϕ a as in the diagram. Since ϕa is an isomorphism
over X − Xcrit and depthx a (I/I 2 ) ⊗ ΩX k−a
/S ≥ d + m + 2, ϕa is an isomorphism. Finally, since both
depthx K a+1 ≥ d + m + 2 and depthx K /K a a+1 a
≥ d + m + 2, depthx K ≥ d + m + 2 as well. 

Proof of Theorem 1.5. It suffices to consider the case p = k. Since every deformation is (locally)
pulled back from the versal deformation, and we are dealing with lci singularities, by standard
properties of flatness and wedge product under base change we can assume that S is smooth. First,
with no assumption on the singular locus, OX is flat over S by assumption. To see that Ω1X /S is
flat over S, again with no assumption on the singular locus, we have the conormal sequence
u
→ π1∗ Ω1A |X → Ω1X /S → 0.
0 → I/I 2 −

Let I0 be the ideal of X in A. Then the conormal sequence for X /S becomes the corresponding
sequence for Ω1X after tensoring with OS,s /ms . In particular, ū is injective. Hence Ω1X /S = Coker u
is flat over S by the local criterion of flatness [Mat80, (20.E) pp. 150–151].
For k ≥ 2, by induction on k and descending induction on a, Ka+1 and Ka /Ka+1 are flat over
V
S for all a ≥ 1, and hence so is Ka . Let K0a be the image of a (I0 /I02 ) ⊗ ΩA k−a
|X in ΩkA |X.
V
Proposition 3.2 applied to the case S = pt implies that K0a /K0a+1 ∼ =
a k−a
(I0 /I02 ) ⊗ ΩX . We have a
commutative diagram
0 −−−−→ Ka+1 ⊗ (OS,s /ms ) −−−−→ Ka ⊗ (OS,s /ms ) −−−−→ Ka /Ka+1 ⊗ (OS,s /ms ) −−−−→ 0
  
  ∼
y y y=
0 −−−−→ K0a+1 −−−−→ K0a −−−−→ K0a /K0a+1 −−−−→ 0
6
Here, the top row is exact for a ≥ 1 by the flatness of Ka /Ka+1 and induction on k. By descending
induction on a, the above diagram implies that the natural map Ka ⊗ (OS,s /ms ) → K0a is an
isomorphism for all a ≥ 1. In particular, for a = 1, we have the map u : K1 → π1∗ ΩkA |X with
cokernel ΩkX /S , and the reduction ū of u mod ms is the natural inclusion K01 → ΩkA |X. Since ū is
injective, ΩkX /S = Coker u is flat over S by the local criterion of flatness as above. This completes
the proof of Theorem 1.5. 

Corollary 3.3. Suppose that X is an lci k-Du Bois singularity and that f : X → S is a flat
morphism, where S is arbitrary, with X = Xs = f −1 (s) for some s ∈ S. Then possibly after
replacing S by a neighborhood of s, the sheaf of relative differentials ΩpX /S is flat over S for all
p ≤ k.

Proof. This is immediate from Theorem 1.5 and Theorem 2.5. 

4. Proof of the main results

We turn now to the global setting of a deformation of a compact analytic space or proper scheme
Y (Corollary 1.4), and to the question of the local freeness of ΩpY/S (Theorem 1.2).

Theorem 4.1. Let f : Y → Spec A be a proper morphism of complex spaces, where A is an Artin
local C-algebra, with closed fiber Y . Let (F • , d) be a bounded complex of coherent sheaves on Y,
flat over A, where d : F i → F i+1 is A-linear, but not necessarily OY -linear. Finally suppose that
the natural map Hi (Y; F • ) = Ri f∗ F • → Hi (Y ; F • |Y ) is surjective for all i. Then Hi (Y; F • ) is a
finite A-module whose length satisfies:

ℓ(Hi (Y; F • )) = ℓ(A) dim Hi (Y ; F • |Y ).

Proof. This is a minor variation on very standard arguments. Note that Hi (Y; F • ) is a finite A-
module since it is the abutment of a spectral sequence with E1 page E1p,q = H q (Y ; F p ) and such
that all of the differentials in the spectral sequence are A-module homomorphisms. Next we show
that, for every finite A-module M , the natural map

ψM : Hi (Y; F • ) ⊗A M → Hi (Y; F • ⊗A M )

is surjective. The proof is via induction on ℓ(M ), the case ℓ(M ) = 1 being the hypothesis of the
theorem. The inductive step follows from: given an exact sequence 0 → M ′ → M → M ′′ → 0
such that ψM ′ and ψM ′′ are surjective, then ψM is surjective. This follows from the commutative
diagram
Hi (Y; F • ) ⊗A M ′ −−−−→ Hi (Y; F • ) ⊗A M −−−−→ Hi (Y; F • ) ⊗A M ′′ −−−−→ 0
  
ψ ψ ψ
y M ′ y M y M ′′
Hi (Y; F • ⊗A M ′ ) −−−−→ Hi (Y; F • ⊗A M ) −−−−→ Hi (Y; F • ⊗A M ′′ )
where the top row is exact since tensor product is right exact and the second is exact since F • is
A-flat.
To prove the theorem, we argue by induction on ℓ(A). The result is clearly true if ℓ(A) = 1. For
the inductive step, write A as a small extension 0 → I → A → A/I → 0. By flatness, there is an
exact sequence

· · · → Hi (Y; F • ⊗A I) → Hi (Y; F • ) → Hi (Y; F • ⊗A A/I) → · · ·

7
Then the above implies that Hi (Y; F • ) → Hi (Y; F • ⊗A A/I) is surjective for all i. Hence the long
exact sequence breaks up into short exact sequences and thus
ℓ(Hi (Y; F • )) = ℓ(Hi (Y; F • ⊗A I)) + ℓ(Hi (Y; F • ⊗A A/I))
= dim Hi (Y ; F • |Y ) + ℓ(A/I) dim Hi (Y ; F • |Y )
= ℓ(A) dim Hi (Y ; F • |Y ).
This concludes the inductive step. 
Now assume that f : Y → Spec A is a proper morphism, where Y is a scheme of finite type over
Spec C and A is an Artin local C-algebra, with closed fiber Y . Consider the complex Ω•Y/ Spec A and
the quotient complex Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A . For the closed fiber Y , we also have the complex
Ω•Y /σ ≥k Ω•Y .
Lemma 4.2. With notation as above, suppose that Y is k-Du Bois. Then:
(i) For every i, the natural map
Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A ) → Hi (Y ; Ω•Y /σ ≥k Ω•Y )
is surjective.
(ii) The spectral sequence with E1 term
(
p,q H q (Y ; ΩpY ), for p ≤ q;
E1 =
0, for p > k.
converging to Hp+q (Y ; Ω•Y /σ ≥k Ω•Y ) degenerates at E1 . Hence, for every i,
X
dim Hi (Y ; Ω•Y /σ ≥k Ω•Y ) = dim H p (Y ; ΩpY ).
p+q=i
p≤k

Proof. By the k-Du Bois condition, the natural map Ω•Y /σ ≥k Ω•Y → Ω•Y /F k Ω•Y is a quasi-isomorphism
of filtered complexes. Thus there are isomorphisms
Hi (Y ; Ω• /σ ≥k Ω• ) ∼
Y = Hi (Y ; Ω• /F k Ω• )
Y Y Y
q p ∼ p
H (Y ; ΩY ) = Hq (Y ; ΩY ) (p ≤ k).
By Corollary 2.2(ii), the spectral sequence with E1 page Hq (Y ; ΩpY ) for p ≤ k and 0 otherwise
degenerates at E1 . Hence the same is true for the spectral sequence in (ii).
To prove (i), arguing as in [DBJ74], there is a commutative diagram
H i (Y; C) −−−−→ Hi (Y; Ω•Y/ Spec A ) −−−−→ Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A )
 
 
y y
H i (Y ; C) −−−−→ Hi (Y ; Ω•Y ) −−−−→ Hi (Y ; Ω•Y /σ ≥k Ω•Y ).
By Corollary 2.2(i), H i (Y ; C) → Hi (Y ; Ω•Y /σ ≥k Ω•Y ) ∼
= Hi (Y ; Ω•Y /F k Ω•Y ) is surjective. Hence
Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A ) → Hi (Y ; Ω•Y /σ ≥k Ω•Y )
is surjective as well. 
Proof of Theorem 1.2. By Corollary 3.3, possibly after shrinking S, ΩpY/S is flat over S. By a
standard argument, it is enough to show the following: Let A be an Artin local C-algebra and
f : Y → Spec A a proper morphism whose closed fiber Y is isomorphic to Ys . Then Ri f∗ ΩpY/ Spec A =
H i (Y; ΩpY/ Spec A ) is a free A-module, compatible with base change, and in fact for the base change
8
condition it is enough to show that, for every i and quotient A → A/I = A, the natural map
H i (Y; ΩpY/ Spec A ) → H i (Y; ΩpY/ Spec A ) is surjective, where Y = Y ×Spec A Spec A.
By Theorem 4.1 and Lemma 4.2,
ℓ(Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A )) = ℓ(A) dim Hi (Y ; Ω•Y /σ ≥k Ω•Y )
X
= ℓ(A) dim H p (Y ; ΩpY ).
p+q=i
p≤k

On the other hand, by analogy with Lemma 4.2(ii), there is the spectral sequence converging to
Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A ) with E1 page
(
p,q H q (Y; ΩpY ), for p ≤ q;
E1 =
0, for p > k.
X
Thus ℓ(Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A )) ≤ ℓ(H q (Y; ΩpY )), with equality ⇐⇒ the spectral
p+q=i
p≤k
sequence degenerates at E1 . Moreover, ≤ ℓ(A) dim H q (Y ; ΩpY ), with equality holding
ℓ(H q (Y; ΩpY ))
for all q ⇐⇒ for every q and every quotient A → A/I = A, the natural map H q (Y; ΩpY/ Spec A ) →
H q (Y; ΩpY/ Spec A ) is surjective. Combining, we have
X X
ℓ(Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A )) ≤ ℓ(H q (Y; ΩpY )) ≤ ℓ(A) dim H q (Y ; ΩpY )
p+q=i p+q=i
p≤k p≤k

= ℓ(A) dim Hi (Y ; Ω•Y /σ ≥k Ω•Y ) = ℓ(Hi (Y; Ω•Y/ Spec A /σ ≥k Ω•Y/ Spec A ).
Thus all of the inequalities must have been equalities, and in particular H i (Y; ΩpY/ Spec A ) →
H i (Y; ΩpY/ Spec A ) is surjective for every quotient A → A. This completes the proof, by the re-
marks in the first paragraph of the proof. 
Remark 4.3. Instead of assuming that Y is a scheme of finite type proper over Spec C, it is enough
to assume that Y is a compact analytic space with isolated singularities and that there exists a
¯
resolution of singularities of Y which satisfies the ∂ ∂-lemma. It seems likely that there is a more
general result assuming sufficiently strong conditions on every subvariety of Ysing .
We turn now to a proof of Corollary 1.4. First, we define canonical Calabi-Yau varieties following
[FL22, Definition 6.1]:
Definition 4.4. A canonical Calabi-Yau variety Y is a scheme proper over Spec C (resp. a compact
analytic space) which is reduced, Gorenstein, with canonical singularities, and such that ωY ∼
= OY .
Proof of Corollary 1.4. It suffices by [Kaw92] to show that Y has the T 1 -lifting property. In fact, we
show the following somewhat stronger statement: let A be an Artin local C-algebra, π : Y → Spec A
a deformation of Y over A, and I an ideal of A with A = A/I with Y = Y ×Spec A Spec A, the
natural map
Ext1 (Ω1Y/ Spec A , OY ) → Ext1 (Ω1Y/ Spec A , OY )
is surjective. Arguing as in [Kaw92, Lemma 3], we claim that ωY/ Spec A = ∼ OY . To see this, note
that, as Y is 1-Du Bois it is 0-Du Bois, and hence by Theorem 1.1 Ri π∗ OY = H i (Y; OY ) is a
free A-module for all i, compatible with base change. In particular, Rn π∗ OY is a free rank one
A-module. An application of relative duality then shows that (Ri π∗ OY )∨ ∼= Rn−i ωY/ Spec A , and
9
hence Rn−i ωY/ Spec A is a free A-module for all i. Thus R0 π∗ ωY/ Spec A is a free rank one A-module
and the natural map R0 π∗ ωY/ Spec A ⊗A OY → ωY/ Spec A is an isomorphism. Hence ωY/ Spec A ∼ = OY .
By a similar application of relative duality, and since Ri π∗ Ω1Y/ Spec A = H i (Y; Ω1Y/ Spec A ) is a free
A-module for all i, there is an isomorphism of A-modules
Ext1 (Ω1Y/ Spec A , OY ) ∼
= Ext1π (Ω1Y/ Spec A , ωY/ Spec A ) ∼
= HomA (H n−1 (Y; Ω1Y/ Spec A ), A).

Then the T 1 -lifting criterion follows from the fact that H n−1 (Y; Ω1Y/ Spec A ) is a free A-module and
that the natural map H n−1 (Y; Ω1Y/ Spec A ) ⊗A A → H n−1 (Y; Ω1Y/ Spec A ) is an isomorphism. 

Remark 4.5. A similar argument shows that, if Y is a compact analytic space with isolated lci
1-Du Bois singularities such that ωY ∼ = OY and there exists a resolution of singularities of Y
¯
satisfying the ∂ ∂-lemma, then Def (Y ) is unobstructed.

References
[DB81] P. Du Bois, Complexe de de Rham filtré d’une variété singulière, Bull. Soc. Math. France 109 (1981),
no. 1, 41–81.
[DBJ74] P. Du Bois and P. Jarraud, Une propriété de commutation au changement de base des images directes
supérieures du faisceau structural, C. R. Acad. Sci. Paris Sér. A 279 (1974), 745–747. MR 376678
[Del68] P. Deligne, Théorème de Lefschetz et critères de dégénérescence de suites spectrales, Inst. Hautes Études
Sci. Publ. Math. (1968), no. 35, 259–278.
[FL22] R. Friedman and R. Laza, Deformations of singular Fano and Calabi-Yau varieties, arXiv:2203.04823,
2022.
[Gre75] G.-M. Greuel, Der Gauss-Manin-Zusammenhang isolierter Singularitäten von vollständigen Durchschnit-
ten, Math. Ann. 214 (1975), 235–266.
[JKSY21] S. J. Jung, I. K. Kim, M. Saito, and Y. Yoon, Higher Du Bois singularities of hypersurfaces, to appear
in Proc. Lond. Math. Soc. (arXiv:2107.06619), 2021.
[Kaw92] Y. Kawamata, Unobstructed deformations. A remark on a paper of Z. Ran, J. Algebraic Geom. 1 (1992),
no. 2, 183–190.
[KL20] M. Kerr and R. Laza, Hodge theory of degenerations, (II): vanishing cohomology and geometric applica-
tions, arXiv:2006.03953, 2020.
[KLSV18] J. Kollár, R. Laza, G. Saccà, and C. Voisin, Remarks on degenerations of hyper-Kähler manifolds, Ann.
Inst. Fourier (Grenoble) 68 (2018), no. 7, 2837–2882.
[Kol21] J. Kollár, Families of varieties of general type, with the collaboration of K. Altmann and S. Kovács
(available at https://web.math.princeton.edu/∼kollar/FromMyHomePage/modbook.pdf), 2021.
[Kun86] E. Kunz, Kähler differentials, Advanced Lectures in Mathematics, Friedr. Vieweg & Sohn, Braunschweig,
1986.
[Laz10] R. Laza, The moduli space of cubic fourfolds via the period map, Ann. of Math. (2) 172 (2010), no. 1,
673–711.
[Mat80] H. Matsumura, Commutative algebra, second ed., Mathematics Lecture Note Series, vol. 56, Ben-
jamin/Cummings Publishing Co., Inc., Reading, Mass., 1980.
[MOPW21] M. Mustaţă, S. Olano, M. Popa, and J. Witaszek, The Du Bois complex of a hypersurface and the
minimal exponent, arXiv:2105.01245, May 2021.
[MP21] M. Mustaţă and M. Popa, Hodge filtration on local cohomology, Du Bois complex, and local cohomological
dimension, arXiv:2108.05192, August 2021.
[NS95] Y. Namikawa and J. H. M. Steenbrink, Global smoothing of Calabi-Yau threefolds, Invent. Math. 122
(1995), no. 2, 403–419.
[PS08] C. A. M. Peters and J. H. M. Steenbrink, Mixed Hodge structures, Ergebnisse der Mathematik und ihrer
Grenzgebiete. 3. Folge., vol. 52, Springer-Verlag, Berlin, 2008.
[Ran92] Z. Ran, Deformations of Calabi-Yau Kleinfolds, Essays on mirror manifolds, Int. Press, Hong Kong,
1992, pp. 451–457.
[Sch64] G. Scheja, Fortsetzungssätze der komplex-analytischen Cohomologie und ihre algebraische Charakter-
isierung, Math. Ann. 157 (1964), 75–94.
[Sha79] J. Shah, Insignificant limit singularities of surfaces and their mixed Hodge structure, Ann. of Math. (2)
109 (1979), no. 3, 497–536.
10
[Sha80] , A complete moduli space for K3 surfaces of degree 2, Ann. of Math. (2) 112 (1980), no. 3,
485–510.
[Ste83] J. H. M. Steenbrink, Mixed Hodge structures associated with isolated singularities, Singularities, Part
2 (Arcata, Calif., 1981), Proc. Sympos. Pure Math., vol. 40, Amer. Math. Soc., Providence, RI, 1983,
pp. 513–536.
[Tia92] G. Tian, Smoothing 3-folds with trivial canonical bundle and ordinary double points, Essays on mirror
manifolds, Int. Press, Hong Kong, 1992, pp. 458–479.

Columbia University, Department of Mathematics, New York, NY 10027

Email address: rf@math.columbia.edu

Stony Brook University, Department of Mathematics, Stony Brook, NY 11794

Email address: radu.laza@stonybrook.edu

11