# Full text of "Hodge theory meets the minimal model program: a survey of log canonical and Du Bois singularities"

## See other formats

```HODGE THEORY MEETS THE MINIMAL MODEL PROGRAM: A
SURVEY OF LOG CANONICAL AND DU BOIS SINGULARITIES

SANDOR J KOVACS AND KARL SCHWEDE

Abstract. This is a survey of some recent developments in the study of singularities related
to the classification theory of algebraic varieties. In particular, the definition and basic
properties of Du Bois singularities and their connections to the more commonly known
singularities of the minimal model program are reviewed and discussed.

1. Introduction

The primary goal of this note is to survey some recent developments in the study of
singularities related to the minimal model program. In particular, we review the definition
and basic properties of Du Bois singularities and explain how these singularities fit into the
minimal model program and moduli theory.

Since we can resolve singularities [Hir64], one might ask the question why we care about
them at all. It turns out that in various situations we are forced to work with singularities
even if we are only interested in understanding smooth objects.

One reason we are led to study singular varieties is provided by the minimal model program
|KM98j . The main goal is classification of algebraic varieties and the plan is to find reasonably
simple representatives of all birational classes and then classify these representatives. It turns
out that the simplest objects in a birational class tend to be singular. What this really means
is that when choosing a birational representative, we aim to have simple global properties
and this is often achieved by a singular variety. Being singular means that there are points
where the local structure is more complicated than on a smooth variety, but that allows for
the possibility of still having a somewhat simpler global structure and along with it, good
local properties at most points.

Another reason to study singularities is that to understand smooth objects we should
also understand how smooth objects may deform and degenerate. This leads to the need
to construct and understand moduli spaces. And not just moduli for the smooth objects.
Degenerations provide important information as well. In other words, it is always useful to
work with complete moduli problems, i.e., extend our moduli functor so it admits a compact
(and preferably projective) coarse moduli space. This also leads to having to consider singular
varieties.

On the other hand, we have to be careful to limit the kind of singularities that we allow
in order to be able to handle them. One might view this survey as a list of the singularities

Date: September 5, 2009.

2000 Mathematics Subject Classification. 14B05.

The first named author was supported in part by NSF Grant DMS-0554697 and the Craig McKibben and
Sarah Merner Endowed Professorship in Mathematics.

The second named author was partially supported by RTG grant number 0502170 and by a National
Science Foundation Postdoctoral Research Fellowship.

1

that we must deal with to achieve the above stated goals. Fortunately, it is also a class of
singularities with which we have a reasonable chance to be able to work.

In particular, we will review Du Bois singularities and related notions including some
very recent important results. We will also review a family of singularities defined via
characteristic p methods, the Frobenius morphism, and their connections to the other set of
singularities we are discussing.

Definitions and notation 1.1. Let k be an algebraically closed field. Unless otherwise
stated, all objects will be assumed to be defined over k. A scheme will refer to a scheme of
finite type over k and unless stated otherwise, a point refers to a closed point.

For a morphism Y —>■ S and another morphism T ^ S, the symbol will denote Y XgT.
In particular, for t G 5 we write Xf = f^^{t). In addition, if T = SpecF, then Yt will also
be denoted by Yp.

Let X be a scheme and ^ an <^x-iiiodule. The m*'* reflexive power of ^ is the double
dual (or reflexive hull) of the m**^ tensor power of

A line bundle on X is an invertible ffx-^odu\e. A Q-line bundle =Sf on X is a reflexive
^x-module of rank 1 that possesses a reflexive power which is a line bundle, i.e., there exists
an m G N+ such that jg ^ line bundle. The smallest such m is called the index of =Sf .

• For the advanced reader: whenever we mention Weil divisors, assume that X is 5*2 |Har77[
Thm. 8.22A(2)] and think of a Weil divisorial sheaf, that is, a rank 1 reflexive ^x-module
which is locally free in codimension 1. For flatness issues consult |Kol08at Theorem 2].

• For the novice: whenever we mention Weil divisors, assume that X is normal and adopt
the definition |Har77l p. 130].

For a Weil divisor D on X, its associated Weil divisorial sheaf is the ^x-niodule iffxiD)
defined on the open set f/ C X by the formula

and made into a sheaf by the natural restriction maps.

A Weil divisor D on X is a Cartier divisor, if its associated Weil divisorial sheaf, iffx{D)
is a line bundle. If the associated Weil divisorial sheaf, ^x{D) is a Q-line bundle, then D
is a Q- Cartier divisor. The latter is equivalent to the property that there exists an m G N+
such that mD is a Cartier divisor. Weil divisors form an abelian group. Tensoring this group
with Q (over Z) one obtains the group of Q- divisors on X (note that if X is not normal,
some unexpected things can happen in this process, see |Kol92l Chapter 16]).

The symbol ~ stands for linear and = for numerical equivalence of divisors.

Let =Sf be a line bundle on a scheme X. It is said to be generated by global sections if
for every point x G X there exists a global section G iJ°(X, ^) such that the germ
generates the stalk J/f^ as an ^x-module. If =Sf is generated by global sections, then the
global sections define a morphism

a, 6 G r(t/, ^x), b is not a zero divisor anywhere on U, and

0^ : X = P {H^X, ^Y) .

2

S£ is called semi-ample if is generated by global sections for m ^ 0. =^ is called ample
if it is semi- ample and is an embedding for m ^ 0. A line bundle =Sf on X is called hig
if the global sections of ^Sf™" define a rational map 0jfm : X such that X is birational

to 0^m(X) for m 0. Note that in this case need not be generated by global sections,
so is not necessarily defined everywhere. We leave it for the reader the make the obvious
adaptation of these notions for the case of Q-line bundles.

The canonical divisor of a scheme X is denoted by Kx and the canonical sheaf of X is
denoted by ux-

A smooth projective variety X is of general type if ux is big. It is easy to see that this
condition is invariant under birational equivalence between smooth projective varieties. An
arbitrary projective variety is of general type if so is a desingularization of it.

A projective variety is canonically polarized if ujx is ample. Notice that if a smooth
projective variety is canonically polarized, then it is of general type.

Acknowledgements. We would like to thank Kevin Tucker, Zsolt Patakfalvi and the
referee for reading a preliminary draft and making helpful suggestion for improving the
presentation.

2. Pairs and resolutions

For the reader's convenience, we recall a few definitions regarding pairs.
Definition 2.1. A pair (X, A) consists of a normaQ quasi-projective variety or complex
space X and an effective Q-divisor A C X. A morphism of pairs 7 : (X, A) (X, A) is a
morphism 7 : X — > X such that 7(Supp(A)) C Supp(A). A morphism of pairs 7 : (X, A)
(X, A) is called birational if it induces a birational morphism 7 : X ^ X and 7(A) = A. It
is an isomorphism if it is birational and it induces an isomorphism 7 : X X.
Definition 2.2. Let (X, A) be a pair, and x G X a point. We say that (X, A) is snc at x, if
there exists a Zariski-open neighborhood U of x such that U is smooth and A fl f/ is reduced
and has only simple normal crossings (see Section FS.BI for additional discussion). The pair
(X, A) is snc if it is snc at all x G X.

Given a pair (X, A), let (X, A)reg be the maximal open set of X where (X, A) is snc, and
let (X, A)sing be its complement, with the induced reduced subscheme structure.

Remark 2.2.1. If a pair (X, A) is snc at a point x, this implies that all components of A
are smooth at x. If instead of the condition that U is Zariski-open one would only require
this analytically locally, then Definition 12.21 would define normal crossing pairs rather than
pairs with simple normal crossing.

Definition 2.3. A log resolution of (X, A) is a proper birational morphism of pairs it :
(X, A) (X, A) that satisfies the following four conditions:

(2.3.1) X is smooth,

(2.3.2) A = n^^A is the strict transform of A,

(2.3.3) Exc(7r) is of pure codimension 1,

(2.3.4) Supp(A U Exc(7r)) is a simple normal crossings divisor.

Occasionally, we will discuss pairs in the non-normal setting. See Section fS.FI for more details.

3

(2.3.5) the strict transform A of A has smooth support,

then we call vr an embedded resolution of A C X.

In many cases, it is also useful to require that vr is an isomorphism over (X, A)i.eg.

3. Introduction to the singularities of the mmp

Even though we have introduced pairs and most of these singularities make sense for pairs,
to make the introduction easier to digest we will mostly discuss the case when A = 0. As
mentioned in the introduction, one of our goals is to show why we are forced to work with
singular varieties even if our primary interest lies with smooth varieties.

3. A. Canonical singularities

For an excellent introduction to this topic the reader is urged to take a thorough look at
Miles Reid's Young Person's Guide |Rei87] . Here we will only touch on the subject.

Let us suppose that we would like to get a handle on some varieties. Perhaps we want to
classify them or make some computations. In any case, a useful thing to do is to embed the
object in question into a projective space (if we can). Doing so requires a (very) ample line
bundle. It turns out that in practice these can be difficult to find. In fact, it is not easy to
find any non-trivial line bundle on an abstract variety.

One possibility, when X is smooth, is to try a line bundle that is "handed" to us, namely
some (positive or negative) power of the canonical line bundle; ux = detTJ. If X is not
smooth but instead normal, we can construct ujx on the smooth locus and then push it
forward to obtain a rank one refiexive sheaf on all of X (which sometimes is still a line
bundle). Next we will explore how we might "force" this line bundle to be ample in some
(actually many) cases.

Let X be a minimal surface of general type that contains a (— 2)-curve (a smooth rational
curve with self- intersection —2). For an example of such a surface consider the following.

Example 3.1. X = {x^ + + + = 0) C with the Z2-action that interchanges x ^ y
and z ^ w. This action has five fixed points, [1:1: — : — e*] for i = 1, . . . , 5 where e is a

primitive 5**^ root of unity. Consequently the quotient ^ Z2 has five singular points, each a

simple double point of type Ai. Let X ^ X^^^ be the minimal resolution of singularities.
Then X contains five (— 2)-curves, the exceptional divisors over the singularities.

Let us return to the general case, that is, X is a minimal surface of general type that
contains a (— 2)-curve, C C X. As C ~ P^, and X is smooth, the adjunction formula gives
us that Kx -0 = 0. Therefore Kx is not ample.

On the other hand, since X is a minimal surface of general type, it follows that Kx is
semi-ample, that is, some multiple of it is base-point free. In other words, there exists a
morphism,

\mKx\ : X ^ X,,, C F{H\X, ^ximKx))*).

This may be deduced from various results. For example, it follows from Bombieri's classi-
fication of pluri-canonical maps, but perhaps the simplest proof is provided by Miles Reid
|Rei97l E.3].

4

It is then relatively easy to see that this morphism onto its image is independent of m (as
long as mKx is base point free). This constant image is called the canonical model of X, it
will be denoted by Xcan-

The good news is that the canonical line bundle of Xcan is indeed ample, but the trouble
is that Xcan is singular. We might consider this as the first sign of the necessity of working
with singular varieties. Fortunately the singularities are not too bad, so we still have a good
chance to work with this model. In fact, the singularities that can occur on the canonical
model of a surface of general type belong to a much studied class. This class goes by several
names; they are called du Val singularities, or rational double points, or Gorenstein, canonical
singularities. For more on these singularities, refer to |Dur79j . |Rei87j .

3.B, Normal crossings

These singularities already appear in the construction of the moduli space of stable curves
(or if the reader prefers, the construction of a compactificaton of the moduli space of smooth
projective curves). If we want to understand degenerations of smooth families, we have to
allow normal crossings.

A normal crossing singularity is one that is locally analytically (or formally) isomorphic to
the intersection of coordinate hyperplanes in a linear space. In other words, it is a singularity
locally analytically defined as {xiX2 ■ ■ ■ Xr = 0) C A" for some r < n. In particular, as
opposed to the curve case, for surfaces it allows for triple intersections. However, triple
intersections may be "resolved": Let X = {xyz = 0) C A^. Blow up the origin O G A^,
a : BIqA^ — »■ A^ and consider the proper transform of X, a : X — >■ X. Observe that X has
only double normal crossings.

Another important point to remember about normal crossings is that they are not normal.
In particular they do not belong to the previous category. For some interesting and perhaps
surprising examples of surfaces with normal crossings see |Kol07j .

3.C. Pinch points

Another non-normal singularity that can occur as the limit of smooth varieties is the pinch
point. It is locally analytically defined as [xf = X2x'^) C A*^. This singularity is a double
normal crossing away from the pinch point. Its normalization is smooth, but blowing up the
pinch point (i.e., the origin) does not make it any better. (Try it for yourself!)

3.D. Cones

Let C C be a curve of degree d and X C P'^ the projectivized cone over C. As X is a
degree d hypersurface, it admits a smoothing.

Example 3.2. Let S = {x'^ + y'^ + z'^ + tw'^ = 0) C Fl.y..^.^^ x Aj. The special fiber Sq is a
cone over a smooth plane curve of degree d and the general fiber S^, for t 7^ 0, is a smooth
surface of degree c? in P^.

This, again, suggests that we must allow some singularities. The question is, whether we
can limit the type of singularities we must deal with. More particularly to this case, can we
limit the type of cones we need to allow?

First we need an auxiliary computation. By the nature of the computation it is easier to
use divisors instead of line bundles.

5

Commentary 3.3. One of our ultimate goals is to construct a moduli space for canonical
models of varieties. We are already aware that the minimal model program has to deal with
singularities and so we must allow some singularities on canonical models. We would also
like to understand what constraints are imposed if our goal is to construct a moduli space.
The point is that in order to construct our moduli space, the objects must have an ample
canonical class. It is possible that a family of canonical models degenerates to a singular
fiber that has singularities worse than the original canonical models. An important question
then is whether we may resolve the singularities of this special fiber and retain ampleness of
the canonical class. The next example shows that this is not always possible.

Example 3.4. Let W he a smooth variety and X = XiU X2 C W such that Xi and X2
are Cartier divisors in W. Then by the adjunction formula we have

Kx = {Kw + X)\^

Kx, = {Kw + Xi)\^^

Kx2 = + -^2)1^2

Therefore

(3.4.1) Kx\^^ = Kx,+X:,.,\^^
for 2 = 1, 2, so we have that

(3.4.2) Kx is ample -v^ Kx\x. = ^Xi + -^3-j|x is ample for i = 1, 2.

Next, let X be a normal projective surface with Kx ample and an isolated singular point
P G SingX. Assume that X is isomorphic to a cone Sq C as in Example 13.21 locally
analytically near P. Further assume that X is the special fiber of a family S that itself is
smooth. In particular, we may assume that all fibers other than X are smooth. As explained
in fl3.3l) . we would like to see whether we may resolve the singular point P G X and still be
able to construct our desired moduli space, i.e., that K of the resolved fiber would remain
ample. For this purpose we may assume that P is the only singular point of X.

Let T ^ S be the blowing up of P G S and let X denote the proper transform of X. Then
Tq = X U P where P ~ is the exceptional divisor of the blow up. Clearly, a : X ^ X is
the blow up of P on X, so it is a smooth surface and X fl P is isomorphic to the degree d
curve over which X is locally analytically a cone.

We would like to determine the condition on d that ensures that the canonical divisor of
To is still ample. According to (13.4.21) this means that we need that Ke + X\^ and K^ + E\~
be ample.

As P ~ uje ^ ^p2(-3), so ^e^Ke + X|^) ^ fff^^d - 3). This is ample if and only if
d > 3.

As this computation is local near P the only relevant issue about the ampleness of P^ +
P|j^ is whether it is ample in a neighborhood of Ex'- = E\x' next claim this is

equivalent to asking when {K^ + Ex) ■ Ex is positive.

Claim. Let Z he a smooth projective surface with non-negative Kodaira dimension and
r C Z an effective divisor. If [Kz + T) ■ C > for every proper curve C G Z, then Kz + F
is ample.

6

Proof. By the assumption on the Kodaira dimension there exists an m > such that mKz
is effective, hence so is m{Kz + T). Then by the assumption on the intersection number,
{Kz + r)^ > 0, so the statement follows by the Nakai-Moishezon criterium. □

Observe that by the adjunction formula (fT^ + Ex) ■ Ex = degKEx = d{d — 3) as Ex is
isomorphic to a plane curve of degree d. Again, we obtain the same condition as above and
thus conclude that Kr^ may be ample only if > 3.

Now, if we are interested in constructing moduli spaces, then one of the requirements of
being stable is that the canonical bundle be ample. This means that in order to obtain a
compact moduli space we have to allow cone singularities over curves of degree d < 3. The
singularity we obtain for = 2 is a rational double point, but the singularity for d = 3 is
not even rational. This does not fit any of the earlier classes we discussed. It belongs to the
one discussed in the next section.

3.E. Log canonical singularities

Let us investigate the previous situation under more general assumptions.

Computation 3.5. Let D = X]i=o'^«-^*' (-^^ ^ ^ divisor with only normal crossing

singularities in a smooth ambient variety such that Aq = 1. Using a generalized version of
the adjunction formula shows that in this situation (13.4.11) remains true.

r

(3.5.1) Kn\^^ = KD, + Y.X.D,\^^

1=1

Let f : S ^ B a projective family with dimi? = 1, S smooth and K^^ ample for all b E B.
Further let X = for some bo E B a singular fiber and let a : T ^ S be an embedded
resolution of X C S. Finally let Y = a*X = X + Yll=i where X is the proper transform
of X and Fi are exceptional divisors for a. We are interested in finding conditions that are
necessary for Ky to remain ample.

Let Ei'. = Fi\~ be the exceptional divisors for a : X ^ X and for the simplicity of
computation, assume that the Ei are irreducible. For Ky to be ample we need that Ky\~
as well as -f^y|^ for all i are all ample. Clearly, the important one of these for our purposes
is Ky\~ for which by (13.5. ip we have that

r

Ky\^ = K^ + J2XiEi.

i=l

As usual, we may write Kj^ = a*Kx + Yl\=i (^i^i, so we are looking for conditions to
guarantee that (J*Kx + ^(^i + Xi)Ei be ample. In particular, its restriction to any of the Ei
has to be ample. To further simplify our computation let us assume that dimX = 2. Then
the condition that we want satisfied is that for all j,

(3.5.2) ^^(a,, + A,)E, j ■ E, > 0.

7

Let

E+ = ^ \ai + Xi\Ei, and

ai+Xi>0

E- = ^ \ai + Xi\Ei, so

ai+Xi<0

r

Y,ia^ + K)Ei = E+-E_.
1=1

Choose a j such that Ej C Supp-E+. Then E_ ■ Ej > since Ej ^ E_ and fl3.5.2p imphes
that {E+ - E_) ■ Ej > 0. These together imply that E+ ■ Ej > and then that El > 0.
However, the Ei are exceptional divisors of a birational morphism, so their intersection
matrix, {Ei ■ Ej) is negative definite.

The only way this can happen is if E^ = 0. In other words, + Aj < for all i. However,
the Aj are positive integers, so this implies that Ky may remain ample only if < —1 for
all i = 1, . . . , r.

The definition of a log canonical singularity is the exact opposite of this condition. It
requires that X be normal and admit a resolution of singularities, say Y ^ X, such that
all the > —1. This means that the above argument shows that we may stand a fighting
chance if we resolve singularities that are worse than log canonical, but have no hope to do
so with log canonical singularities. In other words, this is another class of singularities that
we have to allow. As we remarked above, the class of singularities we obtained for the cones
in the previous subsection belong to this class. In fact, all the normal singularities that we
have considered so far belong to this class.

The good news is that by now we have covered pretty much all the ways that something
can go wrong and found the class of singularities we must allow. Since we have already found
that we have to deal with some non-normal singularities and in fact in this example we have
not really needed that X be normal, we conclude that we will have to allow the non-normal
cousins of log canonical singularities. These are called semi-log canonical singularities and
the reader can find their definition in the next subsection.

3.F, Semi-log canonical singularities

Semi-log canonical singularities are very important in moduli theory. These are exactly the
singularities that appear on stable varieties, the higher dimensional analogs of stable curves.
However, their definition is rather technical, so the reader might want to skip this section at

As a warm-up, let us first define the normal and more traditional singularities that are
relevant in the minimal model program.

Definition 3.6. A pair {X, A) is called log Q-Gorenstein if Kx + ^ is Q-Cartier, i.e., some
integer multiple of Kx + A is a Cartier divisor. Let (X, A) be a log Q-Gorenstein pair and
/ : X — i> X a log resolution of singularities with exceptional divisor E = UEi. Express the
log canonical divisor of X in terms of Kx + A and the exceptional divisors:

+ A = r{Kx + A) + J2<^^E,

where A = f~^A, the strict transform of A on X and G Q. Then the pair (X, A) has

8

terminal

' ai> 0.
di > 0.
< ttj > — 1.

canonical
pit
kit

log canonical

> singularities, if

for all log resolutions /,
and for all i,

ai > -1 and [AJ < 0.
> -1.

The corresponding definitions for non-normal varieties are somewhat more cumbersome.
We include them here for completeness, but the reader should feel free to skip them and
assume that for instance "semi-log canonical" means something that can be reasonably
considered a non-normal version of log canonical.

Suppose that X is a reduced equidimensional scheme that satisfies the following conditions:

(3.6.1) X satisfies Serre's condition S2 (cf. |Har77[ Thm. 8.22A(2)]).

(3.6.2) X has only simple normal double crossings in codimension 1 (in particular X is
Gorenstein in codimension

Conditions fl3.6|[Tl) and fl3.6ll2|) imply that we may treat the canonical module of X as a
divisorial sheaf even though X is not normal. Further suppose that D is a Q-Weil divisor on
X (again, following |Kol921 Chapter 16], we assume that X is regular at the generic point
of each component in SuppD).

Remark 3.7. Notice that conditions fl3.6|[T|) and fl3.6ll2l) imply that X is seminormal since it
is seminormal in codimension 1 (see [GTSOf Corollary 2.7]).

Set p : X^ ^ X to be the normalization of X and suppose that B is the divisor of the
conductor ideal on X^. We use p~^{D) to denote the puUback of D to X^.

Definition 3.8. We say that (X, D) is semi-log canonical if the following two conditions

(3.8.1) Kx + Dis Q-Cartier, and

(3.8.2) the pair (X^, B + p~^D) is log canonical.

Actually, this is not the original definition of semi-log canonical singularities. The original
definition (which is equivalent to this one) uses the theory of semi- resolutions. See |KSB88j .
[Kol92l Chapter 12], and |Kol08bj for details.

4. Hyperresolutions and Du Bois' original definition

A very important construction is Du Bois's generalized De Rham complex. The original
construction of Du Bois's complex, Vl'^, is based on simplicial resolutions. The reader in-
terested in the details is referred to the original article |DB81j . Note also that a simplified
construction was later obtained in [Car85j and [GNPP88] via the general theory of polyhe-
dral and cubic resolutions. At the end of the paper, we include an appendix in which we
explain how to construct, and give examples of cubical hyperresolutions. An easily accessible
introduction can be found in |Ste85] . Another useful reference is the recent book |PS08j .

Recently the second named author found a simpler alternative construction of (part of)
Du Bois's complex that does not need a simplicial resolution, see |Sch07] and also Section 
below. However we will discuss the original construction because we believe that it is impor-
tant to keep in mind the way these singularities appeared as that explains their usefulness.

hold.

Sometimes a ring that is S2 and Gorenstein in codimension 1 is called quasi-normal.

9

For more on applications of Du Bois's complex and Du Bois singularities see |Ste83] , |Kol95t
Chapter 12], |Kov99l[Kov00b] .

The word "hyperresolution" will refer to either simplicial, polyhedral, or cubic resolution.
Formally, the construction of Q'x is the same regardless the type of resolution used and no
specific aspects of either types will be used.

The following definition is included to make sense of the statements of some of the forth-
coming theorems. It can be safely ignored if the reader is not interested in the detailed
properties of Du Bois's complex and is willing to accept that it is a very close analog of the
De Rham complex of smooth varieties.

Definition 4.1. Let X be a complex scheme (i.e., a scheme of finite type over C) of
dimension n. Let Dfiit{X) denote the derived category of filtered complexes of ^x-inodules

with differentials of order < 1 and D fm^cohiX) the subcategory of D^\t{X) of complexes K' ,
such that for all z, the cohomology sheaves of Gr^^^^^K' are coherent cf. |DB81] . |GNPP8 8].
Let D{X) and Dco\i{X) denote the derived categories with the same definition except that
the complexes are assumed to have the trivial filtration. The superscripts +, — , 6 carry the
usual meaning (bounded below, bounded above, bounded). Isomorphism in these categories
is denoted by . A sheaf ^ is also considered a complex with = ^ and =
for i 7^ 0. If i^' is a complex in any of the above categories, then h^{K' ) denotes the z-th
cohomology sheaf of K' .

The right derived functor of an additive functor F, if it exists, is denoted by RF and
R'F is short for h' o RF. Furthermore, ff , ff^ , and J^^ will denote i?T, R'Tz, and R'J^z
respectively, where F is the functor of global sections, Tz is the functor of global sections
with support in the closed subset Z, and Jifz is the functor of the sheaf of local sections with
support in the closed subset Z. Note that according to this terminology, if 0: F — X is a
morphism and ^ is a coherent sheaf on Y, then R(j)^^ is the complex whose cohomology
sheaves give rise to the usual higher direct images of

Theorem 4.2 [DBSlj 6.3, 6.5]. Let X be a proper complex scheme of finite type and D
a closed suhscheme whose complement is dense in X. Then there exists a unique object
^'x G Oh Dfiit{X) such that using the notation

it satisfies the following properties

(4.2.1) Q'x —qis'Cx, i-^-, Ql'x ^ resolution of the constant sheaf C on X.

(4.2.2) £2(* •) is functorial, i.e., if (p: Y ^ X is a morphism of proper complex schemes
of finite type, then there exists a natural map 0* of filtered complexes

(j)* : ^ R(f)*QiY ■

Furthermore, ^I'x E Ob {F>^fut cohi^)) ^^'^ ^Z*/* proper, then (p* is a morphism
^nD%,^^^,{X).

(4.2.3) Let U C X be an open subscheme of X . Then

(4.2.4) If X is proper, there exists a spectral sequence degenerating at Ei and abutting
to the singular cohomology of X:

= w (X, fi^) ^ i^p+''(x^^ c).

10

If e. : X. ^ X is a hyperresolution, then

In particular, {^x) ~ ^ f^"^ i < 0.
There exists a natural map, 6x — ilx? compatible with
// X is smooth, then

—qis ^X-

In particular,

0.x —qis ^X'

If (p: Y ^ X is a resolution of singularities, then

Suppose that n : Y ^ Y is a projective morphism and X G Y a reduced closed
subscheme such that vr is an isomorphism outside of X. Let E denote the reduced
subscheme ofY with support equal to tt^^{X) and tt' : E ^ X the induced map.
Then for each p one has an exact triangle of objects in the derived category,

QP. . QP, © Rn^n^ R-n'^VLl ^ .

It turns out that Du Bois's complex behaves very much hke the de Rham complex for
smooth varieties. Observe that f l4.2ll4l) says that the Hodge-to-de Rham spectral sequence
works for singular varieties if one uses the Du Bois complex in place of the de Rham complex.
This has far reaching consequences and if the associated graded pieces, turn out to be
computable, then this single property leads to many applications.

Notice that fl4.2|[6!) gives a natural map ffx ^Ix^ '^ill be interested in situations

when this map is a quasi-isomorphism. When X is proper over C, such a quasi- isomorphism
will imply that the natural map

H%X^'', C) ^ H\X, Gx) = M\X,Q^x)

is surjective because of the degeneration at Ei of the spectral sequence in fl4.2lH|) .

Following Du Bois, Steenbrink was the first to study this condition and he christened this
property after Du Bois.

Definition 4.3. A scheme X is said to have Du Bois singularities (or DB singularities for
short) if the natural map ^x Ox from ( ]4.2ll6l) is a quasi-isomorphism.

Remark 4.4. U e : X. X is a hyperresolution of X (see the Appendix for a how to
construct cubical hyperresolutions) then X has Du Bois singularities if and only if the natural
map ffx —>■ R^' *^x. is a quasi-isomorphism.

Example 4.5. It is easy to see that smooth points are Du Bois and Deligne proved that
normal crossing singularities are Du Bois as well cf. |DJ74t Lemme 2(b)].

(4.2.5)

(4.2.6)
(4.2.7)

(4.2.8)
(4.2.9)

We will see more examples of Du Bois singularities in later sections.

11

5. An injectivity theorem and splitting the Du Bois complex

In this section, we state an injectivity theorem involving the duahzing sheaf that plays a
role for Du Bois singularities similar to the role that Grauert-Riemenschneider players for
rational singularities. As an application, we state a criterion for Du Bois singularities related
to a "splitting" of the Du Bois complex.

Theorem 5.1. |Kov99l Lemma 2.2], |Sch09t Proposition 5.11] Let X be a reduced scheme
of finite type over C, x ^ X a (possibly non-closed) point, and Z = {x} its closure. Assume
that X\Z has Du Bois singularities in a neighborhood of x (for example, x may correspond
to an irreducible component of the non-Du Bois locus of X). Then the natural map

is injective for every i.

The proof uses the fact that for a projective X, H'^{X^^, C) —>■ M^{X,Qx) is surjective for
every i > 0, which follows from Theorem 14.21

It would also be interesting and useful if the following generalization of this injectivity
were true.

Question 5.2. Suppose that X is a reduced scheme essentially of finite type over C. Is it
true that the natural map of sheaves

h' [R9{om'x{VPx^^x)) {uox)

is injective for every il

Even though Theorem 15.11 does not answer Question 15.21 it has the following extremely
useful corollary.

Theorem 5.3. [Kov99l Theorem 2.3], |Kol95l Theorem 12.8] Suppose that the natural map
Gx ^Ix ^^'5 ^ ^^ft inverse in the derived category (that is, a map p : £2^ Gx such

that the composition Gx ^^^x — ~^^x is an isomorphism). Then X has Du Bois

singularities.

Proof. Apply the functor Rif{omx{ , ujx) to the maps ^x ^ ilx — ■ Then by the

assumption, the composition

^'x R9{omx{Q3(, ^x) ^

is an isomorphism. Let x G X be a possibly non-closed point corresponding to an irreducible
component of the non-Du Bois locus of X and consider the stalks at x of the cohomology
sheaves of the complexes above. We obtain that the natural map

{R^omx{n'x,^x)),^h^ K).

is surjective for every i. But it is also injective by Theorem 15. 1[ This proves that S : (ujx)x

R!}{om X ^^x) X is a quasi-isomorphism. Finally, applying the functor R9{omff^^{ , {ujx)x)

one more time proves that X is Du Bois at x, contradicting our choice of x G X □

This also gives the following Boutot-like theorem for Du Bois singularities (cf. |Bou87] ).

12

Corollary 5.4. |Kov99l Theorem 2.3], |Kol95l Theorem 12.8] Suppose that f : Y ^ X is a
morphism, Y has Du Bois singularities and the natural map Gx Rf^^Y has a left inverse
in the derived category. Then X also has Du Bois singularities.

Proof. Observe that the composition is an isomorphism

Then apply Theorem 15.31 □

As an easy corollary, we see that rational singularities are Du Bois (which was first observed
in the isolated case by Steenbrink in |Ste83l Proposition 3.7]).

Corollary 5.5. |Kov99j . |SaiOOj If X has rational singularities, then X has Du Bois singu-
larities.

Proof. Let vr : X — > X be a log resolution. One has the following composition i^x ^x
Rtt^GjI. Since X has rational singularities, this composition is a quasi-isomorphism. Apply
Corollary [Ml □

6. Hyperresolution-free characterizations of Du Bois singularities

The definition of Du Bois singularities given via hyperresolution is relatively complicated

(hyperresolutions themselves can be rather complicated to compute, see Appendix B ). In
this section we state several hyperresolution free characterizations of Du Bois singularities.
The first such characterization was given by Steenbrink in the isolated case. Another, more
analytic characterization was given by Ishii and improved by Watanabe in the isolated quasi-
GorensteiiJl case. Finally the second named author gave a characterization that works for
any reduced scheme.

Du Bois gave a relatively simple characterization of an affine cone over a projective variety
being Du Bois in [DBSlj . Steenbrink generalized this criterion to all normal isolated singu-
larities. It is this criterion that Steenbrink, Ishii, Watanabe, and others used extensively to
study isolated Du Bois singularities.

Theorem 6.1. |DB81l Proposition 4.16] |Ste83l 3.6] Let (X, x) he a normal isolated Du Bois
singularity, and it : X ^ X a log resolution of {X, x) such that ir is an isomorphism outside
of X \ {x}. Let E denote the reduced pre-image of x. Then {X,x) is a Du Bois singularity
if and only if the natural map

R'n,G^ R'tt.Ge

is an isomorphism for all i > 0.

Proof. Using Theorem 14. 2[ we have an exact triangle

Since {x}, X and E are all Du Bois (the first two are smooth, and E is snc), we have the
following exact triangle

® R^*^x — ^ R^*^E — •

'A variety X is quasi-Gorenstein if Kx is a Cartier divisor. It is not required that X is Cohen-Macaulay.

13

Suppose first that X has Du Bois singularities (that is, Vfx —qis ^x)- By taking cohomology
and examining the long exact sequence, we see that R'Tt^ff^ K^tt^^^e is an isomorphism
for all i > 0.

So now suppose that R'Tt^.i^^ R^iXt^i^E is an isomorphism for all i > 0. By considering
the long exact sequence of cohomology, we see that K'i^x) — alH > 0. On the other
hand h^{^x) is naturally identified with the seminormalization of iffx-, see Proposition 17.81
below. Thus if X is normal, then iffx — > ^''(ilx) is an isomorphism. □

We now state a more analytic characterization, due to Ishii and slightly improved by
Watanabe. First we recall the definition of the plurigenera of a singularity.

Definition 6.2. For a singularity (X, x), we define the plurigenera {5m}m&i]

Theorem 6.3. |Ish85l Theorem 2.3] [WatSTl Theorem 4.2] Let f : X ^ X be a log resolution
of a normal isolated Gorenstein singularity {X, x) of dimension n > 2. Set E to he the reduced
exceptional divisor (the pre-image of x). Then (X, x) is a Du Bois singularity if and only if
5m{X, x) <\ for any m G N.

In |Sch07] , the second named author gave a characterization of arbitrary Du Bois singulari-
ties that did not rely on hyperresolutions, but instead used a single resolution of singularities.
An improvement of this was also obtained in |ST08l Proposition 2.20]. We provide a proof
for the convenience of the reader.

Theorem 6.4. |Sch07] . [STOSt Proposition 2.20] Let X he a reduced separated scheme of
finite type over a field of characteristic zero. Suppose that X <^Y where Y is smooth and
suppose that tt : Y ^ Y is a proper hirational map with Y smooth and where X = 7r~-'^(X)i.ed;
the reduced pre-image of X , is a simple normal crossings divisor (or in fact any scheme with
Du Bois singularities) . Then X has Du Bois singularities if and only if the natural map
Gx -Rvr^^Y quasi-isomorphism.

In fact, we can say more. There is an isomorphism Rtc^G-y > i]^ such that the natural
map Gx — > 0.x can he identified with the natural map Gx — ^ Rn^G^.

Proof. We first assume that tt is an isomorphism outside of X. Then using Theorem 14. 2[ we
have an exact triangle

Using the octahedral axiom, we obtain the following diagram

+1

+1

a

RiT^nl

RttA

+1

where C is simply the object in the derived category that completes the triangles. But
notice that the vertical arrow a is an isomorphism since Y has rational singularities (in

14

which case each term in the middle column is isomorphic to Gy)- Thus the vertical arrow /3
is also an isomorphism.

One always has a commutative diagram (where the arrows are the natural ones)

Gx -^fx

s ^

Observe that X has Du Bois singularities since it has normal crossings, thus 5 is a quasi-
isomorphism. But then the theorem is proven at least in the case that vr is an isomorphism
outside of X.

For the general case, it is sufficient to show that Rtt^^G-y is independent of the choice of
resolution. Since any two log resolutions can be dominated by a third, it is sufficient to
consider two log resolutions tti : Yi ^ Y and 7r2 : Y2 ^ Y and a map between them p :
Y2 Yi over Y. Let Fi = {7i^\X)),,d and F2 = {7i^\X)),,d = {p-\Fi)\,d. Dualizing the
map and applying Grothendieck duality implies that it is sufficient to prove that UY^iFi) ^
Rp^,{(jjY2{,F2)) is a quasi-isomorphism.

We now apply the projection formula while twisting by uJyI{—Fi). Thus it is sufficient to
prove that

Rp,{uy,/yAF2- P*Fr)) ^ Gy,
is a quasi-isomorphism. But note that F2 — p*Fi = — [p*(l — e)Fi\ for sufficiently small
e > 0. Thus it is sufficient to prove that the pair (Yi, (1 — e)Fi) has kit singularities by
Kawamata-Viehweg vanishing in the form of local vanishing for multiplier ideals; see |Laz04t
9.4]. But this is true since Yi is smooth and Fi is a reduced integral divisor with simple
normal crossings. □

It seems that in this characterization the condition that the ambient variety Y is smooth is
asking for too much. We propose that the following may be a more natural characterization.
For some motivation and for a statement that may be viewed as a sort of converse, see
Conjecture 112.51 and the discussion preceding it.

Conjecture 6.5. Theorem 6.4 should remain true if the hypothesis that Y is smooth is

replaced by the condition that Y has rational singularities.

Having Du Bois singularities is a local condition, so even if X is not embeddable in a
smooth scheme, one can still use Theorem 16.41 by passing to an affine open covering.

To illustrate the utility and meaning of Theorem 16.41 we will explore the situation when
X is a hypersurface inside a smooth scheme Y. Using the notation of Theorem 16. 4[ we note
that we have the following diagram of exact triangles.

Rn^Gy{-X) ^ Rn^Gy ^ R7r^(

X

^y(-X) Gy Gx

Since Y is smooth, /5 is a quasi-isomorphism (as then Y has at worst rational singularities).
Therefore, X has Du Bois singularities if and only if the map a is a quasi-isomorphism.

15

However, a is a quasi-isomorphism if and only if the dual map

(6.5.1) Rtt,uj~(X) ^ UyiX)

is a quasi-isomorphism. The projection formula tells us that Equation 16.5.11 is a quasi-
isomorphism if and only

(6.5.2) Rn,ffy{Kyfy - Ti*X + X)-^ &x

is a quasi-isomorphism. Note however that — 7r*X + X = [—(1 —e)'n*X'\ for £ > and suffi-
ciently close to zero. Thus the left side of Equation 16.5.21 can be viewed as Rn^Gyi^Ky —
(1 — £)7r*X]) for £ > sufficiently small. Note that Kawamata-Viehweg vanishing in the
form of local vanishing for multiplier ideals implies that J{Y, {\ — e)X^ c::^gis RTT^^ffydKy^y —
(1 — e)7!-*X]). Therefore X has Du Bois singularities if and only if J(Y, (1 — e)X) ~ ffx-

Corollary 6.6. If X is a hypersurface in a smooth Y, then X has Du Bois singularities if
and only if the pair (Y, X) is log canonical.

Note that Du Bois hypersurfaces have also been characterized via the Bernstein-Sato
polynomial, see |Sai09l Theorem 0.5].

7. Seminormality of Du Bois singularities

In this section we show that Du Bois singularities are partially characterized by seminor-
mality. First we remind the reader what it means for a scheme to be seminormal.

Definition 7.1. |Swa80j [GT80 ] Suppose that i? is a reduced excellent ring and that S ^ R
is a reduced i?- algebra which is finite as an i?-module. We say that the extension i : i? > S*
is subintegral if the following two conditions are satisfied.

(7.1.1) i induces a bijection on spectra. Spec 5 Speci?.

(7.1.2) i induces an isomorphism of residue fields over every (not necessarily closed)
point of Spec R.

Remark 7.2. In |GT80j . subintegral extensions are called quasi-isomorphisms.

Definition 7.3. |Swa80j |GT80j Suppose that _R is a reduced excellent ring. We say that
R is seminormal if every subintegral extension i? 5* is an isomorphism. We say that a
scheme X is seminormal if all of its local rings are seminormal.

Remark 7.4. In |GT80j . the authors call R seminormal if there is no proper subintegral
extension R S such that S is contained in the integral closure of R (in its total field
of fractions). However, it follows from [Swa80l Corollary 3.4] that the above definition is
equivalent.

Remark 7.5. Seminormality is a local property. In particular, a ring is seminormal if and
only if it is seminormal after localization at each of its prime (equivalently, maximal) ideals.

Remark 7.6. The easiest example of seminormal schemes are schemes with snc singularities.
In fact, a one dimensional variety over an algebraically closed field is seminormal if and only
if its singularities are locally analytically isomorphic to a union of coordinate axes in affine
space.

We will use the following well known fact about seminormality.

16

Lemma 7.7. If X is a seminormal scheme and U ^ X is any open set, then T{U, ^x) is a
seminormal ring.

Proof. We leave it as an exercise to the reader. □

It is relatively easy to see, using the original definition via hyperresolutions, that if X has
Du Bois singularities, then it is seminormal. Du Bois certainly knew this fact, see |DB81l
Proposition 4.9] although he didn't use the word seminormal. Later Saito |SaiOOj proved
that seminormality in fact partially characterizes Du Bois singularities. We give a different
proof of this fact, due to the second named author.

Proposition 7.8. |SaiOOt Proposition 5.2] |Sch09l Lemma 5.6] Suppose that X is a reduced
separated scheme of finite type over C. Then h^{^x) — ^x^^ where &x<^^ is the structure
sheaf of the seminormalization of X .

Proof. Without loss of generality we may assume that X is affine. We need only consider
T^^&E by Theorem 16.41 By Lemma 17. 7[ t^^^e is a sheaf of seminormal rings. Now let
X' = Spec(7r*<^£;) and consider the factorization

X' ^ X.

Note E ^ X' must be surjective since it is dominant by construction and is proper by |Har77t
11.4.8(e)]. Since the composition has connected fibers, so must have p : X' ^ X. On the
other hand, p is a finite map since vr is proper. Therefore p is a bijection on points. Because
these maps and schemes are of finite type over an algebraically closed field of characteristic
zero, we see that r(X, <ffx) ~^ r(-^'; '^x') is a subintegral extension of rings. Since X' is
seminormal, so is r(X', i^x')i which completes the proof. □

8. A MULTIPLIER-IDEAL-LIKE CHARACTERIZATION OF COHEN-MACAULAY DU BOIS

SINGULARITIES

In this section we state a characterization of Cohen-Macaulay Du Bois singularities that
explains why Du Bois singularities are so closely linked to rational and log canonical singu-
larities.

We first do a suggestive computation. Suppose that X embeds into a smooth scheme Y
and that vr : y ^ F is an embedded resolution of X in F that is an isomorphism outside of
X. Set X to be the strict transform of X and set X to be the reduced pre-image of X. We
further assume that X = X U E where E is a. reduced simple normal crossings divisor that
intersects X transversally in another reduced simple normal crossing divisor. Note that E
is the exceptional divisor of vr (with reduced scheme structure). Set E C X be the image of
E. We have the following short exact sequence.

We apply R9{om.ffy{ ,uj'~) followed by i^vr* and obtain the following exact triangle.

Rtt^uj'^ ^ .Rtt^u;-^ R7r^ujj^{E)[dimX]

Using fl4.2ll9l) . the left-most object can be identified with R9{omff^{V^,uj^) and the middle
object can be identified with R9{oin0'^{VlPx,^x)- Recall that X has Du Bois singularities
if and only if the natural map R^Hom ff^iOP^^ij^'x) ~^ ^'x is an isomorphism. Therefore,

17

the object n^uj^[E) is closely related to whether or not X has Du Bois singularities. This
inspired the following result which we state (but do not prove).

Theorem 8.1. |KSS08t Theorem 3.1] Suppose that X is normal and Cohen-Macaulay. Let
TT : X' X be a log resolution, and denote the reduced exceptional divisor of ir by G. Then
X has Du Bois singularities if and only if tx^oJx' (G) — uJx ■

Proof. We will not prove this. The main idea is to show that

Related results can also be obtained in the non-normal Cohen-Macaulay case, see [KSSOSj
for details.

Remark 8.2. The submodule '7:^:Ux'{G) C uJx is independent of the choice of log resolution.
Thus this submodule may be viewed as an invariant which partially measures how far a
scheme is from being Du Bois (compare with |Fuj08| ).

As an easy corollary, we obtain another proof that rational singularities are Du Bois (this
time via the Kempf-criterion for rational singularities).

Corollary 8.3. If X has rational singularities, then X has Du Bois singularities.

Proof. Since X has rational singularities, it is Cohen-Macaulay and normal. Then vr^c^x' =
ux but we also have n^ux' ^ T^^^^x'iG) C ux, and thus n^uJx'{G) = ux as well. Then use
Theorem 18.11 □

We also see immediately that log canonical singularities coincide with Du Bois singularities
in the Gorenstein case.

Corollary 8.4. Suppose that X is Gorenstein and normal. Then X is Du Bois if and only
if X is log canonical.

Proof. X is easily seen to be log canonical if and only if vr^tux'/x (C) — ffx- The projection
formula then completes the proof. □

In fact, a slightly jazzed up version of this argument can be used to show that every
Cohen-Macaulay log canonical pair is Du Bois, see |KSS08l Theorem 3.16].

9. The Kollar-Kovacs splitting criterion

Recently Kollar and the first named author found a rather flexible criterion for Du Bois
singularities.

Theorem 9.1. |KK09j Let f : Y ^ X be a proper morphism between reduced schemes of
finite type over C, W <^ X an arbitrary subscheme, and F: = f~^{W), equipped with the
induced reduced subscheme structure. Let J^wcx denote the ideal sheaf of W in X and J^fcy
the ideal sheaf of F in Y. Assume that the natural map g

^WQX — J-^ Rf*'^F(ZY

admits a left inverse q' , that is, p' o p = id.y^^^^. Then ifY,F, and W all have DB singu-
larities, then so does X .

18

For the proof, please see the original paper.

Remark 9.1.1. Notice that it is not required that / be birational. On the other hand
the assumptions of the theorem and |KovOOal Thm 1] imply that if Y \ F has rational
singularities, e.g., if Y is smooth, then X \ W has rational singularities as well.

This theorem is used to derive various consequences in |KK09j , some of which are formally
unrelated to Du Bois singularities. We will mention some of these in the sequel, but the
interested reader should look at the original article to obtain the full picture.

10. Log canonical singularities are Du Bois

Log canonical and Du Bois singularities are very closely related as we have seen in the
previous sections. This was first observed in |Ish85] . see also |Wat87j and |Ish87] .

Recently, KoUar and the first named author gave a proof that log canonical singularities
are Du Bois using Theorem 19. 1[ We will sketch some ideas of the proof here. There are two
main steps. First, one shows that the non-kit locus of a log canonical singularity is Du Bois
(this generalizes |Amb98j and |Sch08t Corollary 7.3]). Then one uses Theorem 19.11 to show
that this property is enough to conclude that X itself is Du Bois. For the first part we
refer the reader to the original paper. The key point of the second part is contained in the
following Lemma. Here we give a different proof than in |KK09j .

Lemma 10.1. Suppose (X, A) is a log canonical pair and that the reduced non-kit locus of
{X, A) has Du Bois singularities. Then X has Du Bois singularities.

Proof. First recall that the multiplier ideal J^{X, A) is precisely the defining ideal of the
non-kit locus of {X, A) and since {X, A) is log canonical, it is a radical ideal. We set E C X
to be the reduced subscheme of X defined by this ideal. Since the statement is local, we
may assume that X is affine and thus that X is embedded in a smooth scheme Y. We
let 77 : y ^ y be an embedded resolution of {X, A) in Y and we assume that vr is an
isomorphism outside the singular locus of X. Set S to be the reduced-preimage of S (which
we may assume is a divisor in Y) and let X denote the strict transform of X. We consider
the following diagram of exact triangles.

Here the first row is made up of objects in D^^^ii^) needed to make the columns into exact
triangles. Since S has Du Bois singularities, the map e is an isomorphism and so C ~ 0. On
the other hand, there is a natural map i?7r*^j^(-S) R'k^,0'^{K^-ti*{Kx+^)) — J{X,A)
since {X, A) is log canonical. This implies that the map a is the zero map in the derived
category. However, we then see that f3 is also zero in the derived category which implies that
^'^*^T.vjx ^ ^^^^ inverse. Therefore, X has Du Bois singularities (since S UX has
simple normal crossing singularities) by Theorems 15.31 and 16.41 □

19

11. Applications to moduli spaces and vanishing theorems

The connection between log canonical and Du Bois singularities have many useful appli-
cations in moduli theory. We will list a few without proof.

Setup 11.1. Let (p : X ^ B he a flat projective morphism of complex varieties with B
connected. Assume that for all b E B there exists a Q-divisor Dh on Xt, such that {Xb, Dj,)
is log canonical.

Remark 11.2. Notice that it is not required that the divisors Db form a family.

Theorem 11.3. |KK09j Under the assumptions in (111.11) . h^{Xb, Gx^ is independent of
b E B for all i.

Theorem 11.4. |KK09] Under the assumptions in fill.li) if one fiber of cj) is Cohen- Macaulay
(resp. Sk for some k), then all fibers are Cohen- Macaulay (resp. Sk)-

Theorem 11.5. |KK09j Under the assumptions in (111.11) the cohomology sheaves h^{uj'^)
are flat over B, where denotes the relative dualizing complex of (p.

Du Bois singularities also appear naturally in vanishing theorems. As a culmination of the
work of Tankeev, Ramanujam, Miyaoka, Kawamata, Viehweg, Kollar, and Esnault-Viehweg,
Kollar proved a rather general form of a Kodaira-type vanishing theorem in |Kol95l 9.12].
Using the same ideas this was slightly generalized to the following theorem in |KSS08j .

Theorem 11.6 ( |Kol95t 9.12], [KSSOSt 6.2]). Let X be a proper variety and ^ a line bundle
on X. Let =2"^ ~ (ffx{B)), where D = Y^diDi is an effective divisor, and let s be a global
section whose zero divisor is D. Assume that < di < m for every i. Let Z be the scheme
obtained by taking the m*^ root of s (that is, Z = X[^/s] using the notation from |Kol95[
9A]). Assume further that

is surjective.

This, combined with the fact that log canonical singularities are Du Bois yields that
Kodaira vanishing holds for log canonical pairs:

Theorem 11.7. |KSS08l 6.6] Kodaira vanishing holds for Cohen- Macaulay semi-log canon-
ical varieties: Let (X, A) be a projective Cohen- Macaulay semi-log canonical pair and ^ an
ample line bundle on X. Then ^^^) = fori < dimX.

It turns out that Du Bois singularities appear naturally in other kinds of vanishing theo-
rems. Let us cite one of those here.

Theorem 11.8. |GKKP09| 9.3] Let {X,D) be a log canonical reduced pair of dimension
n > 2, TV : X X a log resolution with tt -exceptional set E, and D = Supp(i? + 7r~^D).

H^{Z,Cz)^H^{Z,^z)

is surjective. Then for any collection ofbi>0 the natural map

Then

20

12. Deformations of Du Bois singularities

Given the importance of Du Bois singularities in moduli theory it is an important obvious
question whether they are invariant under small deformation.

It is relatively easy to see from the construction of the Du Bois complex that a general
hyperplane section (or more generally, the general member of a base point free linear system)
on a variety with Du Bois singularities again has Du Bois singularities. Therefore the question
of deformation follows from the following.

Conjecture 12.1. (cf. |Ste83] ) Let D (Z X be a reduced Cartier divisor and assume that
D has only Du Bois singularities in a neighborhood of a point x E D. Then X has only
Du Bois singularities in a neighborhood of the point x.

This conjecture was proved for isolated Gorenstein singularities by Ishii |Ish86] . Also note
that rational singularities satisfy this property, see |Elk78] .

We also have the following easy corollary of the results presented earlier:

Theorem 12.2. Assume that X is Gorenstein and D is normal^ Then the statement of
Conjecture \12. 1\ is true.

Proof. The question is local so we may restrict to a neighborhood of x. If X is Gorenstein,
then so is D as it is a Cartier divisor. Then D is log canonical by (18.41) . and then the pair
(X, D) is also log canonical by inversion of adjunction |Kaw06j . (Recall that if D is normal,
then so is X along D). This implies that X is also log canonical and thus Du Bois. □

It is also stated in |KovOOb[ 3.2] that the conjecture holds in full generality. Unfortunately,
the proof is not complete. The proof published there works if one assumes that the non-
Du Bois locus of X is contained in D. For instance, one may assume that this is the case if
the non-Du Bois locus is isolated.

The problem with the proof is the following: it is stated that by taking hyperplane sections
one may assume that the non-Du Bois locus is isolated. However, this is incorrect. One may
only assume that the intersection of the non-Du Bois locus of X with D is isolated. If one
takes a further general section then it will miss the intersection point and then it is not
possible to make any conclusions about that case.

Therefore currently the best known result with regard to this conjecture is the following:

Theorem 12.3. |KovOOb[ 3.2] Let D G X be a reduced Cartier divisor and assume that D
has only Du Bois singularities in a neighborhood of a point x E D and that X \ D has only
Du Bois singularities. Then X has only Du Bois singularities in a neighborhood of x.

Experience shows that divisors not in general position tend to have worse singularities
then the ambient space in which they reside. Therefore one would in fact expect that if
X \ D is reasonably nice, and D has Du Bois singularities, then perhaps X has even better
ones.

We have also seen that rational singularities are Du Bois and at least Cohen- Macaulay
Du Bois singularities are not so far from being rational cf. 18. 1[ The following result of the
second named author supports this philosophical point.

Theorem 12.4. |Sch07t Thm. 5.1] Let X be a reduced scheme of finite type over a field of
characteristic zero, D a Cartier divisor that has Du Bois singularities and assume that X\D
is smooth. Then X has rational singularities (in particular, it is Cohen-Macaulay) .

Hhis condition is actually not necessary, but the proof becomes rather involved without it.

21

Let us conclude with a conjectural generalization of this statement:

Conjecture 12.5. Let X be a reduced scheme of finite type over a field of characteristic
zero, D a Cartier divisor that has Du Bois singularities and assume that X\D has rational
singularities. Then X has rational singularities (in particular, it is Cohen-Macaulay) .

Essentially the same proof as in (112.21) shows that this is also true under the same additional
hypotheses.

Theorem 12.6. Assume that X is Gorenstein and D is normal^ Then the statement of
Conjecture \12.5\ is true.

Proof. If X is Gorenstein, then so is D as it is a Cartier divisor. Then by (18. 4p D is log
canonical. Then by inversion of adjunction [KawOGj the pair {X, D) is also log canonical
near D. (Recall that if D is normal, then so is X along D).

As X is Gorenstein and X\D has rational singularities, it follows that X\D has canonical
singularities. Then X has only canonical singularities everywhere. This can be seen by
observing that D is a Cartier divisor and examining the discrepancies that lie over D for
(X, D) as well as for X. Therefore, by |Elk81j . X has only rational singularities along D. □

13. Characteristic p > analogs of Du Bois singularities

Starting in the early 1980s, the connections between singularities defined by the action of
the Frobenius morphism in characteristic p > and singularities defined by resolutions of
singularities started to be investigated, cf. |Fed83j . After the introduction of tight closure in
|HH90j . a precise correspondence between several classes of singularities was established. See,
for example, [FW89J, [MS91J . [HW02] . |Smi97j . |Har98] . [MggTj . [SmiOOj . ^Har05] . [HY03] .
|Tak04] . [TW04j . |Tak08j . The second name author partially extended this correspondence
in his doctoral dissertation by linking Du Bois singularities with F-injective singularities, a
class of singularities defined in |Fed83j . The currently known implications are summarized
below.

Log Terminal Rational F- Regular )• F-Rational

Log Canonical Du Bois

+ Gor. & normal

We will give a short proof that normal Cohen-Macaulay singularities of dense F-injective
type are Du Bois, based on the characterization of Du Bois singularities given in Section [HI

Note that Du Bois and F-injective singularities also share many common properties. For
example F-injective singularities are also seminormal |Sch09| Theorem 4.7].

First however, we will define F-injective singularities (as well as some necessary prerequi-
sites).

F-Pure/F-Split

V

F-Injective

'again, this condition is not necessary, but the proof becomes rather involved without it

22

Definition 13.1. Suppose that X is a scheme of characteristic j9 > with absolute Frobenius
map F : X ^ X. We say that X is F -finite if F^i^x is a coherent i^x-niodule. A ring R is
called F -finite if the associated scheme Spec R is F-finite.

Remark 13.2. Any scheme of finite type over a perfect field is F-finite, see for example
|Fed83j .

Definition 13.3. Suppose that (i?, m) is an F-finite local ring. We say that R is F-injective

if the induced Frobenius map H^{R) ^ H^{R) is injective for every i > 0. We say that

an F-finite scheme is F-injective if all of its stalks are F-injective local rings.

Remark 13.4. If [R, m) is F-finite, F-injective and has a dualizing complex, then Rq is also
F-injective for any Q e Spec R. This follows from local duality, see [Sch09, Proposition 4.3]
for details.

Lemma 13.5. Suppose X is a Cohen- Macaulay scheme of finite type over a perfect field k.
Then X is F-injective if and only if the natural map F^ujx — ^ oJx is surjective.

Proof. Without loss of generality (since X is Cohen-Macaulay) we can assume that X is
equidimensional. Set / : X — Spec k to be the structural morphism. Since X is fi-
nite type over a perfect field, it has a dualizing complex u'x = f'k and we set ux =
h~'^^^^{uj'x). Since X is Cohen-Macaulay, X is F-injective if and only if the Frobenius

map Hf'^^{ffx,x) ^ Hf'^^{F^&x,x) is injective for every closed point x G X. By local

duality, see [Har66t Theorem 6.2] or |BH93l Section 3.5], such a map is injective if and only
if the dual map F^ux,x — ^^x,x is surjective. But that map is surjective, if and only if the
map of sheaves F^ux ojx is surjective. □

We now briefly describe reduction to characteristic p > 0. Excellent and far more complete
references include |HH09t Section 2.1] and |Kol96t II. 5. 10]. Also see |Smi01] for a more
elementary introduction.

Let i? be a finitely generated algebra over a field k of characteristic zero. Write R =
k[xi, . . . ,Xn]/I for some ideal I and let S denote k[xi, . . . ,x„]. Let X = Speci? and vr :
X — ^ X a log resolution of X corresponding to the blow-up of an ideal J. Let E denote
the reduced exceptional divisor of vr. Then E is the subscheme defined by the radical of the
ideal J ■ G^.

There exists a finitely generated Z-algebra A G k that includes all the coefficients of the
generators of / and J, a finitely generated A algebra Ra C R, an ideal J a C Ra, and schemes
Xa and Ea of finite type over A such that Ra ® a k = R, JaR = J, Xa x spec a Spec k = X
and Ea XgpecA Spec A; = E with Ea an effective divisor with support defined by the ideal
Ja - ^Xa niay localize A at a single element so that Ya is smooth over A and Ea is a
reduced simple normal crossings divisor over A. By further localizing A (at a single element),
we may assume any finite set of finitely generated Ra modules is A-free (see |Hun96[ 3.4] or
|HR76l 2.3]) and we may assume that A itself is regular. We may also assume that a fixed
affine cover of Ea and a fixed affine cover of Xa are also A-free.

We will now form a family of positive characteristic models of X by looking at all the rings
Rt = Ra ®a k{t) where k{t) is the residue field of a maximal ideal t & T = Spec A. Note that
k{t) is a finite, and thus perfect, field of characteristic p. We may also tensor the various
schemes Xa, Ea, etc. with k{t) to produce a characteristic p model of an entire situation.

23

By making various cokernels of maps free A-modules, we may also assume that maps
between modules that are surjective (respectively injective) over k correspond to surjec-
tive (respectively injective) maps over A, and thus surjective (respectively injective) in our
characteristic p model as well; see |HH09j for details.

Definition 13.6. A ring R of characteristic zero is said to have dense F-injective type if for
every family of characteristic p ^ models with A chosen sufficiently large, we have that a
Zariski dense set of those models (over Spec A) have F-injective singularities.

Theorem 13.7. |Sch09] Let X be a reduced scheme of finite type over C and assume that
it has dense F -injective type. Then X has Du Bois singularities.

Proof. We only provide a proof in the case that X is normal and Cohen-Macaulay. For a
complete proof, see |Sch09] . Let vr : X — ^ X be a log resolution of X with exceptional divisor
E. We reduce this entire setup to characteristic p ^ such that the corresponding X is
F-injective. Let F*^ : X — > X be the e-iterated Frobenius map.
We have the following commutative diagram.

F^uJx ^

where the horizontal arrows are induced by the dual of the Frobenius map, F^ffx and

the vertical arrows are the natural maps induced by vr. By hypothesis, is surjective. On the
other hand, for e > sufficiently large, the map labeled p is an isomorphism. Therefore the
map o p is surjective which implies that the map (3 is also surjective. But as this holds for
a dense set of primes, it must be surjective in characteristic zero as well, and in particular,
as a consequence X has Du Bois singularities. □

It is not known whether the converse of this statement is true:

Open Problem 13.8. If X has Du Bois singularities, does it have dense F-injective type?

Since F-injective singularities are known to be closely related to Du Bois singularities, it
is also natural to ask how F-injective singularities deform cf. Conjecture 112. 1[ In general,
this problem is also open.

Open Problem 13.9. If a Cartier divisor D m X has F-injective singularities, does X have
F-injective singularities near Dl

In the case that X (equivalently D) is Cohen-Macaulay, the answer is affirmative, see
|Fed83j . In fact, Fedder defined F-injective singularities partly because they seemed to de-
form better than F-pure singularities (the conjectured analog of log canonical singularities).

Appendix A. Connections with Buchsbaum rings

In this section we discuss the links between Du Bois singularities and Buchsbaum rings.
Du Bois singularities are not necessarily Cohen-Macaulay, but in many cases, they are Buchs-
baum (a weakening of Cohen-Macaulay).

Recall that a local ring (i?, m, k) has quasi- Buchsbaum singularities if mif^(i?) = for all
i < dim R. Further recall that a ring is called Buchsbaum if r'^™^i?r„(i?) is quasi- isomorphic
to a complex of A;- vector spaces. Here r'^™^ is the brutal truncation of the complex at the

24

dim R location. Note that this is not the usual definition of Buchsbaum singularities, rather
it is the so-called Schenzel's criterion, see |Sch82j . Notice that Cohen-Macaulay singularities
are Buchsbaum (after truncation, one obtains the zero-object in the derived category).

It was proved by Tomari that isolated Du Bois singularities are quasi-Buchsbaum (a proof
can be found in |Ish85l Proposition 1.9]), and then by Ishida that isolated Du Bois singulari-
ties were in fact Buchsbaum. Here we will briefiy review the argument to show that isolated
Du Bois singularities are quasi-Buchsbaum since this statement is substantially easier.

Proposition A.l. Suppose that {X,x) is an isolated Du Bois singularity with R = Gx,x-
Then R is quasi-Buchsbaum.

Proof. Note that we may assume that X is affine. Since Spec R is regular outside its the
maximal ideal m, it is clear that some power of m annihilates H^{R) for all i < dimi?. We
need to show that the smallest power for which this happens is 1. We let vr : X ^ X be a
log resolution with exceptional divisor E as in Theorem 16. 1[ Since X is affine, we see that
H'^{R) ~ H'~\X \ {m}, ^x) ^ H'~\X \ E, G^) for all i > 0. Therefore, it is enough to
show that mH'^~^{X \ E, Gj^) =0 for all i < dimX. In other words, we need to show that
mH'{X \ E, 0^) = for all z < dimX - 1.
We examine the long exact sequence

■ • • ^ H-^-\X \ E, ff^) H},iX, G^) H\X, G^) ^ll\X\E,G~)^---

Now, H'^{X, G^) = R\Tra o rr^) {ffx) which vanishes for i < dimX by the Maths dual
of Grauert-Riemenschneider vanishing. Therefore H^{X\E,G'j^) ~ W{X,G'^) for i <
dimX - 1. Finally, since X is Du Bois, H\X, G^) = H\E, Ge) by Theorem EH But it
is obvious that mH^{E, Ge) = since E is a. reduced divisor whose image in X is the point
corresponding to m. The result then follows. □

It is also easy to see that isolated F-injective singularities are also quasi-Buchsbaum.

Proposition A. 2. Suppose that {R,xn) is a local ring that is F-injective. Further suppose
that Spec R \ {m} is Cohen-Macaulay. Then [R, m) is quasi-Buchsbaum.

Proof. Since the punctured spectrum of R is Cohen-Macaulay, H^{R) is annihilated by some
power of m for i < dimi?. We will show that the smallest such power is 1. Choose c G m.
Since R is F-injective, F^ : if^(_R) — H^{R) is injective for all e > 0. Choose e large
enough so that c^^ HU^R) is zero for all i < e. However, for any element z G H^{R),
F^{cz) = c^^F^{z) G c^^H^{R) = for i < dimi?. This implies that cz = and so
mi/^(i?) = for i < dimi?. □

Perhaps the most interesting open question in this area is the following:

Open Problem A. 3 (Takagi). Are F-injective singularities with isolated non-CM locus
Buchsbaum?

Given the close connection between F-injective and Du Bois singularities, this question
naturally leads to the next one:

Open Problem A. 4. Are Du Bois singularities with isolated non-CM locus Buchsbaum?

25

Appendix B. Cubical hyperresolutions

For the convenience of the reader we include a short appendix explaining the construction
of cubical hyperresolutions, as well as several examples. We follow |GNPP88] and mostly
use their notation.

First let us fix a small universe to work in. Let Schred denote the category of reduced
schemes. One should note that the usual fibred product of schemes X XsY need not be
reduced, even when X and Y are reduced. We wish to construct the fibred product in the
category of reduced schemes. Given any scheme W (reduced or not) with maps to X and Y
over 5", there is always a unique morphism W ^ X x g Y , which induces a natural unique
morphism Wj-ed {X Xs ^)red- It is easy to see that (X X5 F)red is the fibred product in
the category of reduced schemes.

Let us denote by 1 the category {0} and by 2 the category {0^1}. Let n be an integer
> —1. We denote by the product of n + 1 copies of the category 2 = {0 ^ 1} [GNPP88]
I, 1.15]. The objects of are identified with the sequences a = (ao, ai, . . . , such
that ai G {0, 1} for < i < n. For n = —1, we set = {0} and for n = we have
^0 = {0 ~* We denote by □„ the full subcategory consisting of all objects of except
the initial object (0, . . . , 0). Clearly, the category □+ can be identified with the category of
□„ with an augmentation map to {0}.

Definition B.l. A diagram of schemes is a functor \$ from a category to the category of
schemes. A finite diagram of schemes is a diagram of schemes such that the aforementioned
category C has finitely many objects and morphisms; in this case such a functor will be called
a C-scheme. A morphism of diagrams of schemes \$ : C°p —>■ Schred to : D°p —>■ Schred is
the combined data of a functor P : C°p — > D°p together with a natural transformation of
functors 77 : \$ — >^ o P.

Remark B.2. With the above definitions, the class of (finite) diagrams of schemes can be
made into a category. Likewise the set of C-schemes can also be made into a category (where
the functor P : C°p —* C°p is always chosen to be the identity functor).

Remark B.3. Let / be a category. If instead of a functor to the category of reduced schemes,
one considers a functor to the category of topological spaces, or the category of categories,
one can define /-topological spaces, and /-categories in the obvious way.

If X. : /°P —>■ Schred is an /-scheme, and i G Ob/, then Xi will denote the scheme
corresponding to i. Likewise if G Mor/ is a morphism (p : j i, then will denote the
corresponding morphism X^p : Xi Xj. If / : Y. —>■ X. is a morphism of /-schemes, we
denote by fi the induced morphism Yi ^ Xi. If X. is an /-scheme, a closed sub- /-scheme
is a morphism of /-schemes g : Z. ^ X. such that for each i & I, the map gi : Zi Xi is a
closed immersion. We will often suppress the g of the notation if no confusion is likely to arise.
More generally, any property of a morphism of schemes (projective, proper, separated, closed
immersion, etc..) can be generalized to the notion of a morphism of /-schemes by requiring
that for each object i of /, gi has the desired property (projective, proper, separated, closed
immersion, etc..)

Definition B.4. |GNPP88| I, 2.2] Given a morphism of /-schemes f ■ Y. ^ X. , we
define the discriminant of f to be the smallest closed sub- /-scheme Z. of X. such that
fi : (Yi — (f~^{Zi))) — i> {Xi — Zi) is an isomorphism for all %.

26

Definition B.5. |GNPP88| I, 2.5] Let 5*. be an /-scheme, f : X. S. a proper morphism
of /-schemes, and D. the discriminant of /. We say that / is a resolutioii^ of S. if X.
is a smooth /-scheme (meaning that each Xi is smooth) and dim f~^(Di) < dimS'j, for all
2 e Ob/.

Remark B.6. This is the definition found in |GNPP88] . Note that the maps are not required
to be surjective (of course, the ones one constructs in practice are usually surjective).

Consider the following example: the map k[x,y]/{xy) —>■ k[x] which sends y to 0. I claim
that the associated map of schemes is a "resolution" of the *-scheme. Spec k[x, y]/{xy). The
discriminant is Spec k[x,y]/{x). The pre-image however is simply the origin on k[x], which
has lower dimension than "1". Resolutions like this one are sometimes convenient to consider.

On the other hand, this definition seems to allow something it perhaps shouldn't. Choose
any variety X of dimension greater than zero and a closed point z & X. Consider the map
z ^ X and consider the *-scheme X. The discriminant is all of X. However, the pre-image
of X is still just a point, which has lower dimension than X itself, by hypothesis.

In view of these remarks, sometimes it is convenient to assume also that dim Di < dim Si
for each i G Ob /. In the resolutions of /-schemes that we construct (in particular, in the
ones that are used to that prove cubic hyperresolutions exist), this always happens.

Let / be a category. The set of objects of / can be given the following pre-order relation,
i < j ii and only if Hom/(z, j) is nonempty. We will say that a category / is ordered if this
pre-order is a partial order and, for each i G Ob /, the only endomorphism of i is the identity
[GNPP88| I, C, 1.9]. Note that a category / is ordered if and only if all isomorphisms and
endomorphisms of / are the identity.

It turns out of that resolutions of /-schemes always exist under reasonable hypotheses.

Theorem B.7. |GNPP88| I, Theorem 2.6] Let S be an I-scheme of finite type over a field k.
Suppose that k is a field of characteristic zero and that I is a finite ordered category. Then
there exists a resolution of S .

In order to construct a resolution Y. of an /-scheme X. , it might be tempting to simply
resolve each Xi, set Yi equal to that resolution, and somehow combine this data together.
Unfortunately this cannot work, as shown by the example below.

Example B.8. Consider the pinch point singularity,

X = Spec k[x, y, z]/ {x^y — z^) = Spec k[s, t^, st]

and let Z be the closed subscheme defined by the ideal (s, st) (this is the singular set). Let
/ be the category {0 1}. Consider the /-scheme defined by Xq = X and Xi = Z (with
the closed immersion as the map). Xi is already smooth, and if one resolves Xq, (that is,
normalizes it) there is no compatible way to map Xi (or even another birational model of
Xi) to it, since its pre-image by normalization will be two-to-one onto Z G X\ The way
this problem is resolved is by creating additional components. So to construct a resolution
Y. we set Yi = Z = Xi (since it was already smooth) and set Yq = Xq ]J Z where Xq is the
normalization of Xq. The map Yi — > Yq just sends Yi (isomorphically) to the new component
and the map Yq Xq is the disjoint union of the normalization and inclusion maps.

'A resolution is a distinct notion from a cubic hyperresolution.

27

One should note that although the theorem proving the existence of resolutions of I-
schemes is constructive, |GNPP88] . it is often easier in practice to construct an ad-hoc
resolution.

Now that we have resolutions of /-schemes, we can discuss cubic hyperresolutions of
schemes, in fact, even diagrams of schemes have cubic hyperresolutions! First we will discuss
a single iterative step in the process of constructing cubic hyperresolutions. This step is
called a 2-resolution.

Definition B.9. |GNPP88l I, 2.7] Let S be an /-scheme and Z. aDf x /-scheme. We say
that Z. is a 2-resolution of 5* if Z. is defined by the Cartesian square (pullback, or fibred
product in the category of (reduced) /-schemes) of morphisms of /-schemes below

Zii^ *- Zqi

f

Zio^ ^ Zqo

where

i) Zoo = S.

ii) Zoi is a smooth /-scheme.

iii) The horizontal arrows are closed immersions of /-schemes.

iv) / is a proper /-morphism

v) Zio contains the discriminant of /; in other words, / induces an isomorphism of
(Zoi)i - (Zii), over {Zoo)i - (Zio)i, for all z G Ob/.

Clearly 2-resolutions always exist under the same hypotheses that resolutions of /-schemes
exist: set Zoi to be a resolution, Zio to be discriminant (or any appropriate proper closed
sub- /-scheme that contains it), and Zu its (reduced) pre-image in Zoi-

Consider the following example.

Example B.IO. Let / = {0} and let S be the /-scheme Speck[t'^,t^]. Let Zoi = =
SpecA;[t] and Zoi — > S' = Zoo be the map defined by k[t'^,t^] k[t]. The discriminant of
that map is the closed subscheme of S = Zoo defined by the map (p : k[t'^,t^] —>■ k that
sends t"^ and to zero. Finally we need to define Zu. The usual fibered product in the
category of schemes is k[t]/{t'^), but we work in the category of reduced schemes, so instead
the fibered product is simply the associated reduced scheme (in this case Spec k[t]/(t)). Thus
our 2-resolution is defined by the diagram of rings pictured below.

k[t]/{t)

k[t]/{t) k[t]

k[t^t^]

We need one more definition before defining a cubic hyperresolution,

28

Definition B.ll. |GNPP88l I, 2.11] Let r be an integer greater than or equal to 1, and let
be a X /-scheme, for 1 < n < r. Suppose that for all n, 1 < n < r, the D^.i x I-
schemes Xq^,^ and X". are equal. Then we define, by induction on r, a x /-scheme

Z. =Ted{X],Xl...,X:)

that we call the reduction of [X] , . . . , XI), in the following way: If r = 1, one defines
Z. =X],iir = 2 one defines Z. .' = red(X} , ) by

J X^^ , ifa = (0,0),

for all (3 G □d', with the obvious morphisms. If r > 2, one defines Z. recursively as
red(red(Xi,...,X:-i),Xr).

Finally we are ready to define cubic hyperresolutions.

Definition B.12. jGNPP88l I, 2.12] Let S be an /-scheme. A cubic hyperresolution aug-
mented over S* is a x /-scheme Z. such that

Z. =red(X.\...,X:),

where

(B.12.1) X] is a 2-resolution of S,

(B.12.2) for 1 < n < r, XI'+i is a 2-resolution of Xf , and
(B.12.3) Za is smooth for all a G Dr.

Now that we have defined cubic hyperresolutions, we should note that they exist under
reasonable hypotheses

Theorem B.13. |GNPP88l I, 2.15] Let S be an I-scheme. Suppose that k is a field of
characteristic zero and that I is a finite (bounded) ordered category. Then there exists Z. ,
a cubic hyperresolution augmented over S such that

dim Za < dim S* — |a| -|- 1, Va G

Below are some examples of cubic hyperresolutions.

Example B.14. Let us begin by computing cubic hyperresolutions of curves so let C be a
curve. We begin by taking a resolution vr : C ^ C (where C is just the normalization). Let
P be the set of singular points of C; thus P is the discriminant of tt. Finally we let E be
the reduced exceptional set of tt, therefore we have the following Cartesian square

E

C

It is clearly already a 2-resolution of C and thus a cubic-hyperresolution of C.

Example B.15. Let us now compute a cubic hyperresolution of a scheme X whose singular
locus is itself a smooth scheme, and whose reduced exceptional set of a strong resolution

TT : X ^ X is smooth (for example, any cone over a smooth variety). As in the previous

29

example, let E be the singular locus of X and E the reduced exceptional set of tt, Then the
Cartesian square of reduced schemes

E

TT

E

^X

is in fact a 2-resolution of X, just as in the case of curves above.

The obvious algorithm used to construct cubic hyperresolutions does not construct hy-
perresolutions in the most efficient or convenient way possible. For example, applying the
obvious algorithm to the intersection of three coordinate planes gives us the following.

Example B.16. Let X UY U Z he the three coordinate planes in A^. In this example we
construct a cubic hyperresolution using the obvious algorithm. What makes this construction
different, is that the dimension is forced to drop when forming the discriminant of a resolution

of a diagram of schemes.

Yet again we begin the algorithm by taking a resolution and the obvious one is tt :

{XUYUZ) {XUYUZ). The discriminant is S = {X D Y) U {X D Z) Li {¥ D Z),

the three coordinate axes. The fiber product making the square below Cartesian is simply

the exceptional set E^{{XnY)U{Xn Z)) 1[{{Y n X) U (F n Z)) U((^ n X) U (Z n Y)).

We now need to take a 2-resolution of the 2-scheme (f) : E ^ B. We take the obvious
resolution that simply separates irreducible components. This gives us E ^ B mapping to
(J) : E ^ B. The discriminant oi E ^ E is a. set of three points Xq, Yq and Zq corresponding
to the origins in X, Y and Z respectively. The discriminant of the map S — > S is simply
identified as the origin Aq of our initial scheme X U Y U Z (recall B is the union of the
three axes). The union of that with the images of Xo, Yq and Zq is again just Aq. The fiber
product of the diagram

B=((xny)u(xnz)) U ((ynx)u(ynz)) u ((znx)u(zny)) ^ (^U'^U^)

B={xnY)u{xnz)u{Ynz)

^ XUYUZ

{E^B)

({Xo, Fo, ^0} ^ {A^}) ^{(j>:E^B)

can be viewed as {Qi, . . . , Qq} — > {Pi, P2, P3} where Qi and Q2 are mapped to Pi and so on
(remember E was the disjoint union of the coordinate axes of X, of Y, and of respectively

30

Z, so E has six components and thus six origins). Thus we have the following diagram
{Qi, . . . , Qe} ^ E

which we can combine with previous diagrams to construct a cubic hyperresolution.

Remark B.17. It is possible to find a cubic hyperresolution for the three coordinate planes
in in a different way. Suppose that S is the union of the three coordinate planes {X,
Y, and Z) of A^. Consider the ^2 or scheme defined by the diagram below (where the
dotted arrows are those in but not in 02).

xnYnz

vnz

xnY

Y

xnz

z

X

■■■■^ ¥

■■^XUYUZ

One can verify that this is also a cubic hyperresolution of X UY U Z.

Now let us discuss sheaves on diagrams of schemes, as well as the related notions of push
forward and its right derived functors.

Definition B.18. |GNPP88| I, 5.3-5.4] Let X. be an /-scheme (or even an /-topological
space). We define a sheaf (or pre-sheaf) of abelian groups F' on X. to be the following
data:

(B.18.1) A sheaf (pre-sheaf) F^ of abelian groups over Xj, for all i G Ob/, and
(B.18.2) An X(^-morphism of sheaves F"^ : F^ — > {X(f,)^F^ for all morphisms (p : i ^ j of
I, required to be compatible in the obvious way.

Given a morphism of diagrams of schemes /. : X.
functor for sheaves on X. .

Y. one can construct a push-forward

Definition B.19. |GNPP88l I, 5.5] Let X. be an /-scheme, Y. a J-scheme, F" a sheaf
on X. , and /. : X. — > F. a morphism of diagrams of schemes. We define (/. )*/^' in the
following way. For each j G Ob J we define

((/.),F-)^ = lim(F^),(^FO

31

where the inverse hmit traverses all pairs (i, 0) where : f{i) — j is a morphism in J°p.

Remark B.20. In many applications, J will simply be the category {0} with one object and
one morphism (for example, cubic hyperresolutions of schemes). In that case one can merely
think of the limit as traversing /.

Remark B.21. One can also define a functor /*, show that it has a right adjoint and that
that adjoint is /* as defined above [GNPP88| I, 5.5].

Definition B.22. |GNPP88| I, Section 5] Let X. and Y. be diagrams of topological spaces
over / and J respectively, \$ : / — >^ J a functor, /. : X. Y. a \$-morphism of topological
spaces. If G' is a sheaf over Y. with values in a complete category C, one denotes by f*G'
the sheaf over X. defined by

for all i G Ob/. One obtains in this way a functor

/* : Sheaves(y. , C) —> Sheaves(X. , C)

Given an /-scheme X. , one can define the category of sheaves of abelian groups Ab(X. )
on X. and show that it has enough injectives. Next, one can even define the derived category
D'^{X. , Ab(X. )) by localizing bounded below complexes of sheaves of abelian groups on X.
by the quasi-isomorphisms (those that are quasi- isomorphisms on each i & I). One can also
show that (/. )* as defined above is left exact so that it has a right derived functor /?(/. )*
|GNPP88| I, 5.8-5.9]. In the case of a cubic hyperresolution of a scheme f : X. — ^ X,

R{{f.),F-) = Rhm{Rfi,F')

where the limit traverses the category / of X. .

Final Remark. We end our excursion into the world of hyperresolutions here. There are
many other things to work out, but we will leave them for the interested reader. Many
"obvious" statements need to be proved, but most are relatively straightforward once one
gets comfortable using the appropriate language. For those and many more statements,
including the full details of the construction of the Du Bois complex and many applications,

References

[Ainb98] F. Ambro: The locus of log canonical singularities, preprint, 1998. larXiv:math.AG7 98 6 67

[Bou87] J.-F. BOUTOT: Singularites rationnelles et quotients par les groupes reductifs, Invent. Math. 88
(1987), no. 1, 65-68. MR877006 (88a:14005)

[BH93] W. Bruns and J. HerzoG: Cohen- Macaulay rings, Cambridge Studies in Advanced Mathe-
matics, vol. 39, Cambridge University Press, Cambridge, 1993. MR1251956 (95h:13020)

[Car85] J. A. Carlson: Polyhedral resolutions of algebraic varieties, Trans. Amer. Math. Soc. 292
(1985), no. 2, 595-612. MR808740 (87i:14008)

[DB81] P. Du BoiS: Complexe de de Rham filtre d'une variete singuliere. Bull. Soc. Math. France 109
(1981), no. 1, 41-81. MR61 3848 (82j:1 4006)

[DJ74] P. Dubois and P. Jarraud: Une propriete de commutation au changement de base des images
directes superieures du faisceau structural, C. R. Acad. Sci. Paris Ser. A 279 (1974), 745-747.
MR0376678 (51 #12853)

[Dur79] A. H. Durfee: Fifteen characterizations of rational double points and simple critical points,
Enseign. Math. (2) 25 (1979), no. 1-2, 131-163. MR543555 (80nn:14003)

32

[Elk78] R. Elkik: Singularites rationnelles et deformations, Invent. Math. 47 (1978), no. 2, 139-147.
MR501926 (80c:14004)

[ElkSl] R. Elkik: Rationalite des singularites canoniques, Invent. Math. 64 (1981), no. 1, 1-6. I\/IR621766
(83a:14003)

[Fed83] R. Fedder: F -purity and rational singularity, Trans. Amer. Math. Soc. 278 (1983), no. 2,
461-480. MR701505 (84h:13031)

[FW89] R. Fedder and K. Watanabe: A characterization of F -regularity in terms of F -purity, Com-
mutative algebra (Berkeley, CA, 1987), Math. Sci. Res. Inst. Publ., vol. 15, Springer, New York,
1989, pp. 227-245. MR1015520 (91k:13009)

[Fuj08] O. FuJiNO: Theory of non-lc ideal sheaves-basic properties-, preprint, 2008. arXiv:0801 .2198

[GKKP09] D. Greb, S. Kebekus, S. J. KovAcs, and T. Peternell: Extension theorems for differential
forms and Bogomolov-Sommese vanishing on log canonical varieties II, in preparation, 2009.

[GT80] S. Greco and C. TraversO: On seminormal schemes, Compositio Math. 40 (1980), no. 3,
325-365. MR571055 (81j:14030)

[GNPP88] F. Guillen, V. Navarro Aznar, P. Pascual Gainza, and F. Puerta: Hyperresolutions
cubiques et descente cohomologique. Lecture Notes in Mathematics, vol. 1335, Springer- Verlag,
Berhn, 1988, Papers from the Seminar on Hodge-Dehgne Theory held in Barcelona, 1982.
MR972983 (90a:14024)

[Har98] N. HarA: A characterization of rational singularities in terms of injectivity of Frobenius maps,

Amer. J. Math. 120 (1998), no. 5, 981-996. MR1 646049 (99h:1 3005)
[Har05] N. Hara: A characteristic p analog of multiplier ideals and applications, Comm. Algebra 33

(2005), no. 10, 3375-3388. MR21 75438 (2006f:1 3006)
[IIW02] N. Hara and K.-I. Watanabe: F-regular and F-pure rings vs. log terminal and log canonical

singularities, J. Algebraic Geom. 11 (2002), no. 2, 363-392. MR1 87411 8 (2002k:1 3009)
[HY03] N. Hara and K.-I. Yoshida: A generalization of tight closure and multiplier ideals. Trans.

Amer. Math. Soc. 355 (2003), no. 8, 3143-3174 (electronic). MR1 974679 (2004i;1 3003)
[Har66] R. Hartshorne: Residues and duality. Lecture notes of a seminar on the work of A.

Grothendieck, given at Harvard 1963/64. With an appendix by P. Deligne. Lecture Notes in

Mathematics, No. 20, Springer- Verlag, Berlin, 1966. MR0222093 (36 #5145)
[Har77] R. Hartshorne: Algebraic geometry, Springer- Verlag, New York, 1977, Graduate Texts in

Mathematics, No. 52. MR0463157 (57 #31 16)
[Hir64] H. HironakA: Resolution of singularities of an algebraic variety over a field of characteristic

zero. I, II, Ann. of Math. (2) 79 (1964), 109-203; ibid. (2) 79 (1964), 205-326. MR0199184(33

#7333)

[HH90] M. HOCHSTER and C. Huneke: Tight closure, invariant theory, and the Briangon- Skoda the-
orem, J. Amer. Math. Soc. 3 (1990), no. 1, 31-116. MR1017784 (91g:13010)

[HH09] M. HoCHSTER and C. Huneke: Tight closure in equal characteristic zero, A preprint of a
manuscript, http://www.math.lsa.umich.edu/hochster/dk.ps, 2009.

[HR76] M. HoCHSTER AND J. L. ROBERTS: The purity of the Frobenius and local cohomology. Advances
in Math. 21 (1976), no. 2, 117-172. MR041 71 72 (54 #5230)

[Hun96] C. Huneke: Tight closure and its applications, GBMS Regional Conference Series in Mathemat-
ics, vol. 88, Published for the Conference Board of the Mathematical Sciences, Washington, DC,
1996, With an appendix by Melvin Hochster. MR1 377268 (96m:13001)

[Ish85] S. ISHii: On isolated Gorenstein singularities, Math. Ann. 270 (1985), no. 4, 541-554. MR776171
(86j:32024)

[Ish86] S. ISHii: Small deformations of normal singularities. Math. Ann. 275 (1986), no. 1, 139-148.
MR849059 (87i:14003)

[Ish87] S. ISHii: Isolated Q-Gorenstein singularities of dimension three. Complex analytic singulari-
ties. Adv. Stud. Pure Math., vol. 8, North- Holland, Amsterdam, 1987, pp. 165-198. MR894292

(89d:32016)

[Kaw06] M. Kawakita: Inversion of adjunction on log canonicity. larXiv:math. AG/051 1 2541
[KSB88] J. KOLLAR AND N. I. Shepherd-Barron: Threefolds and deformations of surface singularities.
Invent. Math. 91 (1988), no. 2, 299-338. MR922803 (88m:14022)

33

[Kol95] J. KoLLAR: Shafarevich maps and automorphic forms, M. B. Porter Lectures, Princeton Uni-
versity Press, Princeton, NJ, 1995. MR1341589 (96i:14016)

[Kol96] J. KOLLAR: Rational curves on algebraic varieties, Ergebnisse der Mathematik und ihrer Gren-
zgebiete. 3. Folge. A Series of Modern Surveys in Mathematics, vol. 32, Springer- Verlag, Berlin,
1996. MR1440180 (98c:14001)

[Kol07] J. KOLLAR: Two examples of surfaces with normal crossing singularities, preprint, 2007.
arXiv:0705.0926v2

[KolOSa] J. Kollar: Hulls and husks, preprint, 2008. ■arXiv:0805.0576v2 [math. AG]
[KolOSb] J. Kollar: Semi log resolution, preprint, 2008. arXiv:081 2.3592 [math.AG]

[KK09] J. Kollar and S. J. KovacS: Log canonical singularities are Du Bois, preprint, 2009.
^rXiv:0902.064S/2 [math.AG]

[KM98] J. Kollar and S. Morl Birational geometry of algebraic varieties, Cambridge Tracts in Math-
ematics, vol. 134, Cambridge University Press, Cambridge, 1998, With the collaboration of C.
H. Clemens and A. Corti, Translated from the 1998 Japanese original. MR1 658959 (2000b:14018)

[Kol92] J. Kollar et. AL: Flips and abundance for algebraic threefolds, Societe Mathematique de
France, Paris, 1992, Papers from the Second Summer Seminar on Algebraic Geometry held at
the University of Utah, Salt Lake City, Utah, August 1991, Asterisque No. 211 (1992). MR1225842
(94f:14013)

[Kov99] S. J. KovACS: Rational, log canonical, Du Bois singularities: on the conjectures of Kollar and
Steenbrink, Compositio Math. 118 (1999), no. 2, 123-133. MR1 71 3307 (2001 g:1 4022)

[KovOOa] S. J. KoVACS: A characterization of rational singularities, Duke Math. J. 102 (2000), no. 2,
187-191. MR1749436 (2002b:14005)

[KovOOb] S. J. KovACS: Rational, log canonical, Du Bois singularities. II. Kodaira vanishing and small
deformations, Compositio Math. 121 (2000), no. 3, 297-304. MR1 761 628 (2001 m:1 4028)

[KSS08] S. J. KovACS, K. E. Schwede, and K. E. Smith: The canonical sheaf of Du Bois singularities,
preprint, 2008. arXiv:0801 .1 541v2 [math.AG]

[Laz04] R. Lazarsfeld: Positivity in algebraic geometry. II, Ergebnisse der Mathematik und ihrer
Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and
Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 49, Springer- Verlag,
Berlin, 2004, Positivity for vector bundles, and multiplier ideals. MR2095472 (2005k:1 4001b)

[MS91] V. B. Mehta and V. Srinivas: Normal F-pure surface singularities, J. Algebra 143 (1991),
no. 1, 130-143. MR1 128650 (92j:1 4044)

[MS97] V. B. Mehta and V. Srinivas: A characterization of rational singularities, Asian J. Math. 1
(1997), no. 2, 249-271. MR1491985 (99e:13009)

[PS08] C. A. M. Peters and J. H. M. Steenbrink: Mixed Hodge structures, Ergebnisse der Math-
ematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results
in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics],
vol. 52, Springer- Verlag, Berhn, 2008. MR2393625

[Rei87] M. Reid: Young person's guide to canonical singularities. Algebraic geometry, Bowdoin, 1985
(Brunswick, Maine, 1985), Proc. Sympos. Pure Math., vol. 46, Amer. Math. Soc, Providence,
RI, 1987, pp. 345-414. MR927963 (89b:14016)

[Rei97] M. Reid: Chapters on algebraic surfaces. Complex algebraic geometry (Park City, UT, 1993),
IAS/Park City Math. Ser., vol. 3, Amer. Math. Soc, Providence, RI, 1997, pp. 3-159. MR1442522
(98d:14049)

[SaiOO] M. Saito: Mixed Hodge complexes on algebraic varieties. Math. Ann. 316 (2000), no. 2, 283-331.
MR1741272 (2002h:14012)

[Sai09] M. Saito: On the Hodge filtration of Hodge modules, Mosc. Math. J. 9 (2009), no. 1, 161-191.
[Sch82] P. SCHENZEL: Applications of dualizing complexes to Buchsbaum rings, Adv. in Math. 44 (1982),

no. 1, 61-77. MR654548(83j:13011)
[Sch07] K. Schwede: A simple characterization of Du Bois singularities. Compos. Math. 143 (2007),

no. 4, 813-828. MR2339829
[Sch08] K. Schwede: Centers of F-purity, to appear in Mathematische Zeitschrift, 2008. larXiv:0807. 1 6541

34

[Sch09] K. SciiWEDE: F-injective singularities are Du Bois, Amer. J. Math. 131 (2009), no. 2, 445-473.
MR2503989

[ST08] K. SCHWEDE AND S. Takagi: Rational singularities associated to pairs, Michigan Math. J. 57

(2008), 625-658.

[Smi97] K. E. Smith: F-rational rings have rational singularities, Amer. J. Math. 119 (1997), no. 1,

159-180. MR1428062 (97k:13004)
[SmiOO] K. E. Smith: The multiplier ideal is a universal test ideal, Comm. Algebra 28 (2000), no. 12,

5915-5929, Special issue in honor of Robin Hartshorne. MR1 80861 1 (2002d: 13008)
[SmiOl] K. E. Smith: Tight closure and vanishing theorems. School on Vanishing Theorems and Effective

Results in Algebraic Geometry (Trieste, 2000), ICTP Lect. Notes, vol. 6, Abdus Salam Int. Cent.

Theoret. Phys., Trieste, 2001, pp. 149-213. MR1919458 (2003f:13005)
[Ste83] J. H. M. Steenbrink: Mixed Hodge structures associated with isolated singularities, Singu-
larities, Part 2 (Areata, Calif., 1981), Proc. Sympos. Pure Math., vol. 40, Amer. Math. Soc,

Providence, RI, 1983, pp. 513-536. MR713277 (85d:32044)
[Ste85] J. H. M. Steenbrink: Vanishing theorems on singular spaces, Asterisque (1985), no. 130,

330 341, Differential systems and singularities (Luminy, 1983). MR804061 (87j:14026)
[Swa80] R. G. Swan: On seminormality, J. Algebra 67 (1980), no. 1, 210-229. MR595029 (82d:13006)
[Tak04] S. Takagi: An interpretation of multiplier ideals via tight closure, J. Algebraic Geom. 13 (2004),

no. 2, 393-415. MR2047704 (2005c:13002)
[Tak08] S. Takagi: A characteristic p analogue of pit singularities and adjoint ideals, Math. Z. 259

(2008), no. 2, 321-341. MR2390084 (2009b:13004)
[TW04] S. Takagi and K.-i. Watanabe: On F-pure thresholds, J. Algebra 282 (2004), no. 1, 278-297.

MR2097584 (2006a:13010)

[Wat87] Watanabe, Kimio: On plurigenera of normal isolated singularities. II, Complex analytic singu-
larities. Adv. Stud. Pure Math., vol. 8, North-Holland, Amsterdam, 1987, pp. 671-685. MR894312
(89h:32029)

Sandor J KovAcs: University of Washington, Department of Mathematics, Seattle, WA 98195, USA 