You are on page 1of 145

GRASSMANNIANS AND SINGULARITIES

YI HU
arXiv:2109.02968v1 [math.AG] 7 Sep 2021

Abstract. Let X be an integral scheme of finite presentation over a field. Let


q be a singular closed point of X. We prove that there exists an open subset V
of X containing q such that V admits a resolution, that is, there exists a smooth
scheme Ve and a proper birational morphism from Ve onto V .

Regularities are all alike; every

singularity is singular in its own way.

Contents

1. Introduction 2
2. A Quick Tour: the main idea and approach 12
3. Primary Plücker Relations and In-homogenized Plücker-Ideal 22
3.1. Plücker relations 22
3.2. Primary Plücker equations with respect to a fixed affine chart 25
3.3. In-homogenized Plücker ideal with respect to a fixed affine chart 27
3.4. Ordering the set of all primary Plücker equations 31
4. A Singular Local Birational Model V for Grd,E 33
4.1. The construction of V ⊂ RF 33
4.2. ̟-divisors, ̺-divisors, and the standard charts of RF 42
5. ϑ-Blowups 46
5.1. Some conventions on blowups 46
5.2. Main binomial equations: revisited 50
5.3. ϑ-centers and ϑ-blowups 51
5.4. Properties of ϑ-blowups 53
1
2 YI HU

6. ℘-Blowups 61
6.1. The initial setup: (℘(11) r0 ) 61
6.2. ℘-centers and ℘-blowups in (℘(kτ ) rµ ) 63
6.3. Properties of ℘-blowups 70
6.4. Proper transforms of defining equations in (℘(kτ ) rµ sh ) 73
7. ð-Blowups 79
7.1. The initial setup: (ð(11) r0 ) 79
7.2. ð-centers and ð-blowups in (ð(kτ ) rµ ) 80
7.3. Properties of ð-blowups 86
7.4. Combining all ϑ-blowups, ℘-blowups, and ð-blowups 90
7.5. Proper transforms of defining equations in (ð(kτ ) rµ sh ) 92
8. Γ-schemes and Their Transforms 97
8.1. Γ-schemes 97
8.2. F -transforms of Γ-schemes in VF[k] 98
8.3. ϑ-transforms of Γ-schemes in Veϑ [k]
104
8.4. ℘-transforms of Γ-schemes in Ve(℘(kτ ) rµ sh ) 111
8.5. ð-transforms of Γ-schemes in Ve(ð r s )
(kτ ) µ h
117
9. Main Statements on the Final Scheme Veð 124
10. Local Resolution of Singularity 136
11. Geometric Resolution 143
References 144

1. Introduction

Let V be an integral scheme of finite presentation over a field k. We say V admits a


resolution if there exists a smooth scheme Ve over k and a proper birational morphism
from Ve onto V .

Theorem 1.1. (Local Resolution, Theorem 10.5) Let X be a scheme of finite pre-
sentation over a field k. Assume further that X is integral and singular. Take any
GRASSMANNIANS AND SINGULARITIES 3

singular closed point q ∈ X. Then, there exists an open subset V of X containing


the point q such that V admits a resolution.

Mnëv showed ([16]) that every integral singularity type of finite type defined
over Z appears in some configuration space of points on the projective plane. This
result is called Mnëv’s universality theorem in literature. Lafforgue ([11] and [12])
strengthened and proved the same statement scheme-theoretically. Also, Lee and
Vakil ([13]) proved the similar scheme-theoretic statement on incidence schemes of
points and lines on the projective plane. Using Gelfand-MacPerson correspondence,
Lafforgue’s version of Mnëv’s universality theorem is equivalent to the statement
that every integral singularity type of finite type defined over Z appears in some
thin Schubert cell on the Grassmannian Gr3,E of three-dimensional linear subspaces
of a fixed vector space E of dimension greater than 3. Every thin Schubert cell is an
open subset of a unique closed subscheme of an affine chart of the Grassmannian.
This unique closed subscheme of that affine chart of Gr3,E is called a Γ-scheme in
this article.
We approach Theorem 1.1 via a detour through Mnëv’s universality theorem by
first resolving all the aforementioned Γ-schemes that are integral, hence also, all the
thin Schubert cells of Gr3,E that are integral.
Following Lafforgue’s presentation of [12], suppose we have a set of vector spaces,
E1 , · · · , En such that every Eα , 1 ≤ α ≤ n, is of dimension 1 over a field k (or, a
free module of rank 1 over Z), for some positive integer n > 1. We let

E = E1 ⊕ . . . ⊕ En .

Then, the Grassmannian Grd,E , defined by

Grd,E = {linear subspaces F ֒→ E | dim F = d},

is a projective variety defined over Z, for any fixed integer 1 ≤ d < n. (For local
resolution of singularities, it suffices to focus on Gr3,E ; in this article, we still consider
the general Grassmannian Grd,E : see the third paragraph of §3.)
4 YI HU

We have a canonical decomposition


M
∧d E = Ei1 ⊗ · · · ⊗ Eid ,
i=(i1 <···<id )∈Id,n
where Id,n is the set of all sequences of d distinct integers between 1 and n.
This gives rise to the Plücker embedding of the Grassmannian by

Grd,E ֒→ P(∧d E) = {(pi )i∈Id,n ∈ Gm \(∧d E\{0})},

F −→ [∧d F ],
where Gm is the multiplicative group.
As a closed subscheme of P(∧d E), the Grassmanian Grd,E is defined by the Plücker
ideal I℘ , generated by all Plücker relations, whose typical member is expressed
succinctly, in this article, as
X
(1.1) F : sgn(s)pus pvs
s∈SF

where SF is an index set, us , vs ∈ Id,n for any s ∈ SF , and sgn(s) is the ± sign
associated with the term pus pvs (see (3.3) and (3.4) for details).
Given the above Plücker equation, we introduce the projective space PF which
comes equipped with the homogeneous coordinates [x(us ,vs ) ]s∈SF . Then, correspond-
ing to each Plücker relation (1.1), there is a linear homogeneous equation in PF ,
called the induced linearized Plücker relation,
X
(1.2) LF : sgn(s)x(us ,vs )
s∈SF

(see Definition 4.13). We set ΛF := {(us , vs ) | s ∈ SF }.


As the question about resolution of singularity type is local, we can focus on an
affine chart Um = (pm 6= 0) of the Plücker projective space P(∧d E) for some fixed
m ∈ Id,n . We can identify the coordinate ring of Um with the polynomial ring
k[xu ]u∈Id,n \{m} . For any Plücker relation F , we let F̄ be the in-homogenization of
F on the chart Um . Given this chart, we then explicitly describe a set of Plücker
relations, called m-primary Plücker relations, listed under a carefully chosen total
order “ <℘ ”,
Fm = {F̄1 <℘ · · · <℘ F̄Υ },
GRASSMANNIANS AND SINGULARITIES 5

n

with Υ = d
− 1 − d(n − d), such that they together generate the in-homogenized
ideal I℘,m of the Plücker ideal I℘ on the chart. Further, on the chart Um , if we
set pm = 1 and set xu = pu for any u ∈ Id,n \{m}, then any m-primary relation
F̄ ∈ Fm admits the following in-homogenized expression
X
F̄ : sgn(sF )xuF + sgn(s)xus xvs ,
s∈SF \{sF }

where xuF is called the leading variable of F̄ whose term is called the leading term
of F̄ and sF ∈ SF is the index for the leading term. (See (3.8) and (3.10) for details.)
Next, we can introduce the natural rational map
Q
Θ̄[Υ],Gr : Um ∩ Grd,E 

(1.3) / Um ❴ ❴ ❴/ F̄ ∈Fm PF

Y
[xu ]u∈Id,n −→ [xu xv ](u,v)∈ΛF
F̄ ∈Fm

where [xu ]u∈Id,n is the in-homogenized Plücker coordinates of a point of Um ∩ Grd,E .


We let Vm be the closure of the graph of the rational map Θ̄[Υ],Gr . Then, we
obtain the following diagram
 Q
Vm  / RF := Um × F̄ ∈Fm PF

 
d,E 
Um ∩ Gr / Um .

The scheme Vm is singular, in general, and is birational to Um ∩ Grd,E . (The


reader is recommended to read §2 to see the purpose of introducing the model Vm .)
As the necessary and crucial steps to achieve our ultimate goal, we are to perform
some specific sequential embedded blowups for (Vm ⊂ RF ).
For the purpose of applying induction, employed mainly for the purpose of proofs,
we also introduce the following rational map.
For any positive integer m, we set [m] := {1, · · · , m}.
Then, for any k ∈ [Υ], we have the rational map
Q
Θ̄[k],Gr : Um ∩ Grd,E 

(1.4) / Um ❴ ❴ ❴/ i∈[k] PF i
6 YI HU
Y
[xu ]u∈Id,n −→ [xu xv ](u,v)∈ΛFi
i∈[k]

We let VF[k] be the closure of the graph of the rational map Θ̄[k],Gr. Then, we obtain
the following diagram
 Q
VF[k]  / RF[k] := Um × i∈[k] PF i

 
d,E 
Um ∩ Gr / Um .

The scheme VF[k] is birational to Um ∩ Grd,E .


Set RF[0] := Um . There exists a forgetful map

RF[j] −→ RF[j−1] , for any j ∈ [Υ].

In the above notations, we have

Vm = VF[Υ] , RF = RF[Υ] .

Proposition 1.2. (Proposition 4.11) The scheme Vm , as a closed subscheme of


Q
RF = Um × F̄ ∈Fm PF , is defined by the following relations, for all F̄ ∈ Fm ,

(1.5) BF,(s,t) : x(us ,vs ) xut xvt − x(ut ,vt ) xus xvs , ∀ s, t ∈ SF \{sF },

(1.6) BF,(sF ,s) : x(us ,vs ) xuF − x(m,uF ) xus xvs , ∀ s ∈ SF \{sF },

(1.7) Bpre-q ,
X X
(1.8) LF : sgn(s)x(us ,vs ) , F̄ : sgn(s)xus xvs
s∈SF s∈SF
P
with F̄ expressed as sgn(sF )xuF + sgn(s)xus xvs , where Bpre-q is the set
s∈SF \{sF }

of binomial equations of pre-quotient type (see Definition 4.9).

Our construction of the desired embedded blowups on Vm ⊂ RF is based upon


the set of all binomial relations of (1.6):

Bmn = {BF,(sF ,s) | F̄ ∈ Fm , s ∈ SF \{sF }}.

An element BF,(sF ,s) of Bmn is called a main binomial relation. We also let

Bres = {BF,(s,t) | F̄ ∈ Fm , s, t ∈ SF \{sF }}.


GRASSMANNIANS AND SINGULARITIES 7

An element BF,(s,t) of Bres is called a residual binomial relation. The residual bino-
mial relations or binomial relations of pre-quotient type in Bpre-q play no roles in
the construction of the aforesaid embedded blowups.
To apply induction, we provide a total order on the set SF \{sF } and list it as

SF \{sF } = {s1 < · · · < stF }

where (tF + 1) is the number of terms in the relation F . This renders us to write
BF,(sF ,s) as B(kτ ) where F = Fk for some k ∈ [Υ] and s = sτ for some τ ∈ [tFk ].
We can now synopsize the process of the embedded blowups for (Vm ⊂ RF ).
It is divided into three sequential blowups, named as, ϑ-, ℘-, ð-blowups, listed in
the order of occurrence. (For the specific purpose of each of the three sequential
blowups, the reader is referred to §2.)
• On ϑ-sets, ϑ-centers, and ϑ-blowups.
For any primary Plücker relation F̄k ∈ Fm , we introduce the corresponding ϑ-
set ϑ[k] = {xuF , x(m,uF ) } and the corresponding ϑ-center Zϑ[k] = XuF ∩ X(m,uF )
k k k k

where Xu = (xu = 0) and X(m,u ) = (x(m,u ) = 0). We set Reϑ := RF .


Fk Fk Fk Fk [0]

Then, inductively, we let Reϑ[k] → Reϑ[k−1] be the blowup of Reϑ[k−1] along (the proper
transform of) the ϑ-center Zϑ[k] for all k ∈ [Υ]. This gives rise to the sequential
ϑ-blowups

(1.9) Reϑ := Reϑ[Υ] −→ · · · −→ Reϑ[k] −→ Reϑ[k−1] −→ · · · −→ Reϑ[0] .

Each morphism Reϑ[k] → Reϑ[k−1] is a smooth blowup, meaning, the blowup of a


smooth scheme along a smooth closed center. For any k, we let Veϑ ⊂ Reϑ be the
[k] [k]

proper transform of V in Reϑ[k] . We set Veϑ := Veϑ[Υ] .


• On ℘-sets, ℘-centers, and ℘-blowups.
For any main binomial B(kτ ) ∈ Bmn , there correspond to a finite integer ρ(kτ )
depending on (kτ ) and a finite integer σ(kτ )µ depending on (kτ )µ for any µ ∈ [ρ(kτ ) ].
We set Re(℘ r ) = Reϑ . For each (k, τ, µ, h) as above, there corresponds a ℘-set
(11) 0

φ(kτ )µh consisting of two special divisors on an inductively defined scheme Re(℘(kτ ) rµ−1 ) ;
its corresponding ℘-center Zφ(kτ )µh is the scheme-theoretic intersection of the two
divisors. We let Z℘ = {Zφ(kτ )µh | k ∈ [Υ], τ ∈ [tFk ], µ ∈ [ρ(kτ ) ], h ∈ [σ(kτ )µ ]}, totally
8 YI HU

ordered lexicographically on the indexes (k, τ, µ, h). We set Re℘(11) r1 s0 := Rϑ . Then,


inductively, we let we Re(℘ r s ) → Re(℘ r s ) be the blowup of Re(℘ r s ) along
(kτ ) µ h (kτ ) µ h−1 (kτ ) µ h−1

(the proper transform of) the ℘-center Zφ(kτ )µh . This gives rise to the sequential ℘-
blowups

(1.10) Re℘ → · · · → Re(℘(kτ ) rµ sh ) → Re(℘(kτ ) rµ sh−1 ) → · · · → Reϑ ,

where Re℘ := Re(℘(ΥtF ) rρ(Υt sσ(Υt ρ ) . Each morphism Re(℘(kτ ) rµ sh ) → Re(℘(kτ ) rµ sh−1 )
Υ FΥ ) FΥ ) (ΥtFΥ )

is a smooth blowup. For any (kτ )µh, we let Ve(℘(kτ ) rµ sh ) ⊂ Re(℘(kτ ) rµ sh ) be the proper
transform of V in Re(℘(kτ ) rµ sh ) . We set Ve℘ := Ve(℘(ΥtF ) rρ(Υt ) sσ(Υt ) ρ(Υt ) ) .
Υ FΥ FΥ FΥ

• On ð-sets, ð-centers, and ð-blowups.


For any main binomial B(kτ ) ∈ Bmn , there correspond to a finite integer κ(kτ )
depending on (kτ ) and a finite integer ς(kτ )µ depending on (kτ )µ for any µ ∈ [κ(kτ ) ].
We set Re(ð r ) = Re℘ . For each (k, τ, µ, h) as above, there corresponds an ð-set
(11) 0

ψ(kτ )µh consisting of two special divisors on an inductively defined scheme Re(ð(kτ ) rµ−1 ) ;
its corresponding ð-center Zψ(kτ )µh is the scheme-theoretic intersection of the two
divisors. We let Zð = {Zψ(kτ )µh | k ∈ [Υ], τ ∈ [tFk ], µ ∈ [κ(kτ ) ], h ∈ [ς(kτ )µ ]}, totally
ordered lexicographically on the indexes (k, τ, µ, h). We set Reð r s := R℘ . Then, (11) 1 0

inductively, we let we Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 ) be the blowup of Re(℘(kτ ) rµ sh−1 ) along
(the proper transform of) the ð-center Zψ(kτ )µh . This gives rise to the sequential ð-
blowups

(1.11) Reð → · · · → Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 ) → · · · → Reð ,

where Reð := Re(ð(ΥtF ) rκ(Υt sς(Υt ρ ) . Each morphism Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 )
Υ FΥ ) FΥ ) (ΥtFΥ )

is a smooth blowup. For any (kτ )µh, we let Ve(ð(kτ ) rµ sh ) ⊂ Re(ð(kτ ) rµ sh ) be the proper
transform of V in Re(ð(kτ ) rµ sh ) . We set Veð := Ve(ð(ΥtF ) rκ(Υt ) sς(Υt ) ρ(Υt ) ) .
Υ FΥ FΥ FΥ

The schemes Veð ⊂ Reð are our final destination.


To study the local structure of Veð ⊂ Reð , we approach it by induction via the
sequential blowups (1.9), (1.10), and (1.11).
Definition 5.1 introduces the covering standard affine charts V for any of the
smooth schemes Reϑ[k] , Re(℘(kτ ) rµ sh ) , Re(ð(kτ ) rµ sh ) , as above.
GRASSMANNIANS AND SINGULARITIES 9

(⋆) Proposition 5.9 introduces coordinate variables for any standard affine chart
V of Reϑ and provides explicit geometric meaning for every coordinate variable.
[k]

Proposition 5.12 provides explicit description and properties of the local defining
equations of the scheme Veϑ[k] ∩ V on any standard affine chart V of Reϑ[k] .
(⋆) Proposition 6.7 introduces coordinate variables for any standard affine chart V
of Re(℘ r s ) and provides explicit geometric meaning for every coordinate variable.
(kτ ) µ h

Proposition 6.9 provides explicit description and properties of the local defining
equations of the scheme Ve℘(kτ ) rµ sh ∩ V on any standard affine chart V of Re(℘(kτ ) rµ sh ) .
(⋆) Proposition 7.5 introduces coordinate variables for any standard affine chart V
of Re(ð r s ) and provides explicit geometric meaning for every coordinate variable.
(kτ ) µ h

Proposition 7.10 provides explicit description and properties of the local defining
equations of the scheme Veð(kτ ) rµ sh ∩ V on any standard affine chart V of Re(ð(kτ ) rµ sh ) .
To summarize the progress, we depict the diagram (1.12) below.
Thus far, we have obtained the first two rows of the diagram:
(⋆) In the first row: each morphism Reℏ → Reℏ′ is Reϑ[k] → Reϑ[k−1] , or Re(ð(kτ ) rµ sh ) →
Re(ð(kτ ) rµ sh−1 ) , or Re(℘(kτ ) rµ sh ) → Re(℘(kτ ) rµ sh−1 ) , and is a smooth blowup; each RF[j] →
RF[j−1] is a projection, a forgetful map.
(⋆) In the second row: each morphism Veℏ → Veℏ′ is Veϑ[k] → Veϑ[k−1] , or Ve(ð(kτ ) rµ sh ) →
Ve(ð(kτ ) rµ sh−1 ) , or Ve(℘(kτ ) rµ sh ) → Ve(℘(kτ ) rµ sh−1 ) ; this morphism as well as each VF[j] →
VF[j−1] is surjective, projective, and birational.
To explain the third and fourth rows of the diagram, we go back to the fixed chart
Um . This is the affine space which comes equipped with the coordinate variables
VarUm := {xu | u ∈ Id,n \{m}}. Let Γ be any subset of VarUm and let ZΓ be the
subscheme of Um defined by the ideal generated by all the elements of Γ together
with all the in-homogenized m-primary Plücker relations F̄ with F̄ ∈ Fm . This is
a Γ-scheme mentioned in the beginning of this introduction. The precise relation
between a given thin Schubert cell and its corresponding Γ-scheme is given in (10.2).
Our gaol is to resolve the Γ-scheme ZΓ when it is integral and singular.
10 YI HU

(1.12)
Reð / ··· / Reℏ / Reℏ′ / ··· / RF[j] / RF[j−1] · · · / Um
O O O O O O

? ? ? ? ? ?
Veð / ··· / Veℏ / Veℏ′ / ··· / VF[j] / VF[j−1] · · · / Um ∩ Grd,E
O O O O O O

? ? ? ? ? ?
Zeð,Γ / ··· / Zeℏ,Γ / Zeℏ′ ,Γ / ··· / ZF[j] ,Γ / ZF[j−1]) ,Γ · · · / ZΓ
O O O O O O
=
? ? ? ? ?
Zeð,Γ
† / ··· / Zeℏ,Γ
† / Zeℏ†′ ,Γ / ··· / †
ZF [j] ,Γ
/ †
ZF [j−1]) ,Γ
··· / ZΓ ,

where all vertical uparrows are closed embeddings.


Let Γ be a subset VarUm . Assume that ZΓ is integral. Then, starting from ZΓ ,
step by step, via induction within every of the sequential ϑ-, ℘-, and ð-blowups, we
are able to construct the third and fourth rows in the diagram (1.12) such that
(⋆) every closed subscheme in the third row, ZF[j] ,Γ , respectively, Zeℏ , admits
explicit local defining equations in any standard chart of its corresponding smooth
scheme in the first row;
(⋆) every closed subscheme in the fourth row ZF†
[j] ,Γ
, respectively, Zeℏ† , is an irre-
ducible component of its corresponding subscheme ZF ,Γ , respectively, Zeℏ , such that [j]

the induced morphism †


ZF [j] ,Γ
−→ ZΓ , respectively, Zeℏ† −→ ZΓ is surjective, projec-
tive, and birational.
(⋆) the left-most Zeð,Γ is smooth; so is Zeð,Γ

, now a connected component of Zeð,Γ.
In this article, a scheme X is smooth if it is a disjoint union of connected smooth
schemes of possibly various dimensions.
The closed subscheme ZF[j] ,Γ , called an F -transform of ZΓ , is constructed in
Lemma 8.3; the closed subscheme Zϑ[j] ,Γ , called a ϑ-transform of ZΓ , is constructed
in Lemma 8.5; the closed subscheme Ze(℘ rµ s ),Γ , called a ℘-transform of ZΓ , is (kτ ) h

constructed in Lemma 8.7; the closed subscheme Ze(ð(kτ ) rµ sh ),Γ , called an ð-transform
of ZΓ , is constructed in Lemma 8.9.
GRASSMANNIANS AND SINGULARITIES 11

Our main theorem on the Grassmannian is

Theorem 1.3. (Theorems 9.6 and 9.7) Let F be either Q or a finite field with p
elements where p is a prime number. Let Γ be any subset of VarUm . Assume that
ZΓ is integral. Let Zeð,Γ be the ð-transform of ZΓ in Veð . Then, Zeð,Γ is smooth over
F. In particular, the induced morphism Zeð,Γ

→ ZΓ is a resolution over F, if ZΓ is
singular.

The proof of Theorem 1.3 (Theorems 9.6 and 9.7) is based upon the explicit
description of the main binomials and linearized Plücker defining equations of Zeð,Γ
(Corollary 8.11) and detailed calculation and careful analysis on the Jacobian of
these equations (§9).
Theorem 1.1 is a direct consequence of Theorem 1.3, combining with Lafforgue’s
version of Mnëv’s unversality theorem (Theorems 10.2 and 10.4).
In general, consider any fixed singular integral scheme X. By Theorem 1.1, X
can be covered by finitely many affine open subsets such that every of these affine
open subsets of X admits a resolution. It remains to glue finitely many such local
resolutions to obtain a global one. This is being pursued.
We learned that Hironaka posted a preprint on resolution of singularities in posi-
tive characteristics [5].
In spite of the current article, the author is not in a position to survey the topics
of resolution of singularities, not even very briefly. We refer to Kollár’s book [10] for
an extensive list of references on resolution of singularities. Over characteristic zero,
there have been some recent progresses since the book [10]: risking inadvertently
omitting some other’s works, let us just mention two very recent ones: [2] and [15].

The approach presented in this paper was inspired by [7]. The two articles,
however, are mathematically independent.
The author is grateful to the anonymous reviewers for their very helpful questions
and constructive suggestions, especially for pointing out the insufficiency of the
previous version. He thanks János Kollár and Chenyang Xu for the suggestion to
write a summary section, §2, to lead the reader a quick tour through the paper.
12 YI HU

A List of Fixed Notations Used Throughout

[n]: the set of all integers from 1 to n, {1, 2 · · · , n}.


Id,n : the set of all sequences of integers {(1 ≤ u1 < · · · < ud ≤ n)}.
P(∧d E): the projective space with Plücker coordinates pi , i ∈ Id,n .
I℘ : the ideal of k[pi ]i∈Id,n generated by all Plücker relations.
Um : the affine chart of P(∧d E) defined by pm 6= 0 for some fixed m ∈ Id,n .
Fm : the set of m-primary Plücker equations.
LFm : the set of all linearized m-primary Plücker equations.

Υ := nd − 1 − d(n − d): the cardinality of Fm ;
V: a standard affine chart of an ambient smooth scheme;
Bmn : the set of all main binomial relations;
Bres : the set of all residual binomial relations;
Bpre-q : the set of all binomial relations of pre-quotient type;
Bq : the set of all binomial relations of quotient type;
B: Bmn ⊔ Bres ⊔ Bq ;
Γ: a subset of Um .
A\a: A\{a} where A is a finite set and a ∈ A.
|A|: the cardinality of a finite set A.
k: a fixed based field.

2. A Quick Tour: the main idea and approach

This section may be skipped entirely if the reader prefers to dive into the main
text immediately. However, reading this section first is recommended.

• A detour to Gr3,E via Mnëv’s universality.


By Mnëv’s universality, any singularity over Z appears in a thin Schubert cell of
the Grassmannian Gr3,E of three dimensional linear subspaces in a vector space E.
Consider the Plücker embedding Gr3,E ⊂ P(∧3 E) with Plücker coordinates pijk .
A thin Schubert cell of Gr3,E is a nonempty intersection of codimension one Schubert
cells of Gr3,E ; it corresponds to a matroid d of rank 3 on the set [n]. Any Schubert
GRASSMANNIANS AND SINGULARITIES 13

divisor is defined by pijk = 0 for some (ijk). Thus, a thin Schubert cell Gr3,E of
d
the matroid d is an open subset of the closed subscheme Z Γ of Gr3,E defined by
{pijk = 0 | pijk ∈ Γ} for some subset Γ of all Plücker variables. The thin Schubert
cell must lie in an affine chart (pm 6= 0) for some m ∈ I3,n . Thus, Gr3,E is an open
d
subset of a closed subscheme of Gr3,E ∩ (pm 6= 0) of the following form

ZΓ = {pijk = 0 | pijk ∈ Γ} ∩ Gr3,E ∩ (pm 6= 0).

This is a closed affine subscheme of the affine chart (pm 6= 0). We aim to resolve
ZΓ , hence also the thin Schubert cell Gr3,E , when both are integral and singular.
d
• Minimal set of Plücker relations for the chart (pm 6= 0).
Up to permutation, we may assume that m = (123) and the chart is

Um := (p123 6= 0).

We write the in-homogenized coordinates of Um as

{xabc | (abc) ∈ I3,n \{(123)}.

As a closed subscheme of the affine space Um , ZΓ is defined by

{xijk = 0, F̄ = 0 | xijk ∈ Γ},

where F̄ rans over all in-homogenized Plücker relations. We need to pin down some
explicit Plücker relations to form a minimal set of generators of Gr3,E ∩ Um .
They are of the following forms:

(2.1) F̄(123),1uv = x1uv − x12u x13v + x13u x12v ,

(2.2) F̄(123),2uv = x2uv − x12u x23v + x23u x12v ,

(2.3) F̄(123),3uv = x3uv − x13u x23v + x23u x13v ,

(2.4) F̄(123),abc = xabc − x12a x3bc + x13a x2bc − x23a x1bc ,

where u < v ∈ [n]\{1, 2, 3} and a < b < c ∈ [n]\{1, 2, 3}. Here, [n] = {1, · · · , n}.
In a nutshell, we have the set

(2.5) Fm = {F̄(123),iuv , 1 ≤ i ≤ 3; F̄(123),abc }


14 YI HU

Every relation of Fm is called m-primary. Here, m = (123).

• Equations of Γ-schemes and likely singularities. Hence, as a closed subscheme


of the affine space Um , ZΓ is defined by

ZΓ = {xu = 0, F̄(123),iuv , 1 ≤ i ≤ 3, F̄(123),abc | xu ∈ Γ},

for all u < v ∈ [n]\{1, 2, 3} and a < b < c ∈ [n]\{1, 2, 3}.


Upon setting xu = 0 with u ∈ Γ, we obtain the affine coordinate subspace

Um,Γ ⊂ Um

such that ZΓ , as a closed subscheme of the affine subspace Um,Γ , is defined by

(2.6) {F̄(123),iuv |Γ, 1 ≤ i ≤ 3; F̄(123),abc |Γ },

where F̄ |Γ denotes the restriction of F̄ to the affine subspace Um,Γ . These are in
general truncated Plücker equations, some of which may be identically zero.
We do not analyze singularities of ZΓ .
But, we make a quick observation: when all the terms of the truncated Plücker
equations, F̄(123),iuv |Γ , 1 ≤ i ≤ 3 and F̄(123),abc |Γ , vanish at a point, then a singularity
is likely to occur.
Thus, first, we would like to “remove” all the zero factors of all the terms of

F̄(123),iuv , 1 ≤ i ≤ 3, F̄(123),abc .

To achieve this, it is workable if we can separate all the terms of the above Plücker
relations. (Years had been passed before we came back to the right approach.)

• Separating the terms of Plücker relations.


Motivated by [7], we establish a local model Vm , birational to the chart Um ∩Gr3,E ,
such that in a specific set of defining binomial equations of Vm , all the terms of the
above Plücker relations are separated.
To explain, we introduce the projective space PF for each and every Plücker
P
relation F = s∈SF sgn(s)pus pvs with [x(us ,vs ) ]s∈SF as its homogeneous coordinates.
GRASSMANNIANS AND SINGULARITIES 15

We then let Vm be the closure of the graph of the rational map Θ̄[Υ],Gr of (1.3)
in the case of Gr3,E . (This is motivated by an analogous construction in [7].) By
calculating the multi-homogeneous kernel of the homomorphism

(2.7) ϕ̄ : k[(xw ); (x(u,v) )] −→ k[xw ]

x(u,v) → xu xv ,
we determine a set of defining relations of Vm as a closed subscheme of the smooth
ambient scheme
Y
RF := Um × PF .
F̄ ∈Fm

These defining relations, among many others, include the following binomials

(2.8) x1uv x(12u,13v) − x12u x13v x(123,1uv) , x1uv x(13u,12v) − x13u x12v x(123,1uv) ,

x2uv x(12u,23v) − x12u x23v x(123,2uv) , x2uv x(23u,12v) − x23u x12v x(123,2uv) ,

x3uv x(13u,23v) − x13u x23v x(123,3uv) , x3uv x(23u,13v) − x23u x12v x(123,3uv) ,

xabc x(12a,3bc) − x12a x3bc x(123,abc) , xabc x(13a,2bc) − x13a x2bc x(123,abc) ,

xabc x(23a,1bc) − x23a x1bc x(123,abc) .

We see that the terms of all the m-primary Plücker relations of (2.5) are separated
into the two terms of the above binomials.
To distinguish, we call xu (e.g., x12u ) a ̟-variable and Xu = (xu = 0) a ̟-divisor;
we call x(u,v) (e.g., x(12u,13v) ) a ̺-variable and X(u,v) = (x(u,v) = 0) a ̺-divisor.
The defining relations also include the linearized Plücker relations as in (1.2):
X
LF = sgn(s)x(us ,vs ) , ∀ F̄ ∈ Fm .
s∈SF

The set of all linearized Plücker relations is denoted by LFm .


There are many other extra defining relations.
As it turns out, through years of “trial and error”, “removing” all the zero factors
of all the binomial relations of (2.8) suffices for our ultimate purpose.
The geometric intuition behind the above sufficiency is as follows. The equations
of (2.8) alone together with LFm only define a reducible closed scheme, in general.
The roles of other extra relations (to be discussed soon) are to pin down its main
16 YI HU

component Vm . As the process of “removing” zero factors goes, a process of some


specific blowups, all the boundary components are eventually blown out of existence,
making the proper transforms of (2.8) together with the linearized Plücker relations
generate the ideal of the final blowup scheme Veð of Vm , on all charts.
We thus call the binomial equations of (2.8) the main binomials. The set of main
binomials is denoted Bmn . The set Bmn is equipped with a somewhat carefully chosen
total ordering (see (5.11)).
The defining relations of Vm in RF also include many other binomials: we classify
them as residual binomials (see Definition 4.12) and binomials of pre-quotient type
(see Definition 4.9). The set of residual binomials is denoted Bres ; the set of binomials
of pre-quotient type is denoted Bpre-q .
Together, the equations in the following sets

Bmn , Bres , Bpre-q , LFm , Fm

define the scheme Vm in the smooth ambient scheme RF . See Corollary 4.11.
When we focus on an arbitrarily fixed chart V of RF , binomials of pre-quotient
type of Bpre-q can be further reduced to binomials of quotient type whose set is
q
denoted by BV . See Definition 4.19 and Proposition 4.20. (For the reason to use the
term “of quotient type”, see [7].)
As mentioned in the introduction, for the purpose of inductive proofs, we also
need the rational map Θ̄[k],Gr of (1.4), and we let VF[k] be the closure of the rational
map of Θ̄[k],Gr, for all k ∈ [Υ]. In this notation, Vm = VF[Υ] . We let
Y
RF[k] := U × PF i .
i∈[k]

This is a smooth scheme and contains VF[k] as a closed subscheme. Further, we have
the natural forgetful map

(2.9) RF[k] −→ RF[k−1] .

• The process of “removing” zero factors of main binomials.


To remove zero factors of main binomials, we either work on each primary Plücker
relation individually (in the case of a ϑ-blowup), or work on each main binomial
GRASSMANNIANS AND SINGULARITIES 17

individually (in the case of a ℘- or an ð-blowup). To this end, we need to provide a


total order on the set Fm .
We let {F̄(123),iuv , 1 ≤ i ≤ 3} go first, then followed by {F̄(123),abc }. Within
{F̄(123),iuv , 1 ≤ i ≤ 3}, we say F̄(123),iuv < F̄(123),ju′ v′ if (uv) < (u′v ′ ) lexicographically,
or when (uv) = (u′ v ′ ) , i < j. Within {F̄(123),abc }, we say F̄(123),abc < F̄(123),a′ b′ c′ if
(abc) < (a′ b′ c′ ) lexicographically. This ordering is compatible with that of Bmn .
The purpose of “removing” zero factors is achieved through sequential blowups
based upon factors of main binomials and their proper transforms. We break
the sequential blowups into three types, named as ϑ-, ℘-, and ð-blowups; besides
“removing” zero factors, each serves its own corresponding purpose.

⋆ On ϑ-blowups.
From the main binomial equations of (2.8), we select the following closed centers

Zϑ : (xiuv = 0)∩(x(123,iuv) = 0), i ∈ [3]; (xabc = 0)∩(x(123,abc) = 0), a 6= b 6= c ∈ [n]\[3].

We order the sets {(uv)} and {(abc)} lexicographically; we order {(iuv)}, written as
{(i, (uv)) | i ∈ [3]}, reverse-lexicographically. We then let {(iuv)} go before {(abc)}.
This way, the set Zϑ is equipped with a total order induced from the above orders
on the indexes.
We then blow up RF along (the proper transforms of) the centers in Zϑ , in the
above order. This gives rise to the sequence (1.9) in the introduction

Reϑ := Reϑ[Υ] −→ · · · −→ Reϑ[k] −→ Reϑ[k−1] −→ · · · −→ Reϑ[0] .

For any k ∈ [Υ], we let Veϑ[k] ⊂ Reϑ[k] be the proper transform of Vm in Reϑ[k] . We
then set Veϑ = Veϑ[Υ] and Reϑ = Reϑ[Υ] .
Besides removing the zero factors as displayed in the centers of Zϑ , ϑ-blowups also
make the proper transforms of the residual binomial equations become dependent
on the proper transforms of the main binomial equations on any standard chart. In
particular, it also leads to the conclusion Veϑ ∩ Xϑ,(m,uk ) = ∅ for all k ∈ [Υ] where
Xϑ,(m,uk ) is the proper transform of the ̺-divisor X(m,uk ) = (x(m,uk ) = 0)
Thus, upon completing ϑ-blowups, we can discard all the residual binomials Bres
from consideration.
18 YI HU

⋆ On ℘-blowups.
Here, we continue the process of “removing” zero factors of the proper transforms
of the main binomials. From now on, we focus on each main binomial individually,
starting from the first one.
The first main binomial equation of (2.8) is

B145 : x145 x(124,135) − x124 x135 x(123,145) .

The proper transforms of all the variables of B145 may assume zero value on Veϑ
except x(123,145) since Veϑ ∩ Xϑ,(123,145) = ∅, where Xϑ,(123,145) is the proper transform
of X(123,145) = (x(123,145) = 0) in Reϑ . For each and every term of B145 , we pick a
“zero” factor to form a pair, but, we do not pick any ̺-variable. We do not pick
x(123,145) because Veϑ ∩ Xϑ,(123,145) = ∅; we do not pick x(124,135) for a good reason.
Such a pair is called a ℘-set with respect to B145 . Then, there are two such pairs.

(2.10) φ1 = (x145 , x124 ), φ2 = (x145 , x135 ).

(The fact that there are only two ℘-sets for the first equation is an accident; for
general main binomial, the number of its corresponding ℘-sets may be huge.) The
common vanishing loci of the variables in ℘-sets give rise to the ℘-centers

(2.11) Zφ1 = Xϑ,145 ∩ Xϑ,124 , Zφ2 = Xϑ,145 ∩ Xϑ,135 ,

where Xϑ,u is the proper transform of Xu in Reϑ . We can then blow up Reϑ along
(the proper transforms) of Zφ1 and Zφ2 .
We then move on to the next main binomial equation

B245 : x245 x(124,235) − x124 x235 x(123,245) .

Note that x124 appears in φ1 , hence the minus term of the proper transform of B245
acquires the exceptional parameter ζ created by the blowup along Zφ1 through the
variable x124 (x124 either turns into the exceptional parameter ζ or acquires it).
This is an additional “zero” factor in the minus term of the proper transform of
B245 . Then, for each and every term of the proper transform of B245 , we pick a
factor, including exceptional parameters, to form a pair, again, we do not pick any
̺-variables. Such a pair is called a ℘-set with respect to B245 . The common vanishing
GRASSMANNIANS AND SINGULARITIES 19

loci of the variables in ℘-sets give rise to ℘-centers in the previously obtained blowup
scheme. We can blow up that scheme along (the proper transforms) of these centers.
We then move on to B345 , repeat the above, and so on.
This gives rise to the sequential blowups (1.10) in the introduction

Re℘ → · · · → Re(℘(kτ ) rµ sh ) → Re(℘(kτ ) rµ sh−1 ) → · · · → Reϑ .

An intermediate blowup scheme in the above is denoted by Re(℘(kτ ) rµ sh ) . Here (kτ )


is the index of a main binomial. As the process of ℘-blowups goes on, more and
more exceptional parameters may be acquired and appear in the proper transform
of the later main binomial B(kτ ) , resulting more pairs of zero factors, hence more
corresponding ℘-sets and ℘-centers. The existence of the index rµ , called round
µ, is due to the need to deal with the situation when an exceptional parameter
with exponent greater than one is accumulated in the proper transform of the main
binomial B(kτ ) (such a situation does not occur for the first few main binomials).
The index sh , called step h, simply indicates the corresponding step of the blowup.
Besides removing the zero factors, the reason that we exclude ̺-variables from
℘-sets is to help to control the binomial equations of quotient type.

⋆ On ð-blowups.
Here, we finalize the process of “removing” zero factors of the proper transforms
of all the main binomials. Like in the ℘-blowups, we focus on each main binomial
relation individually, starting from the first one. The construction is analogous to
that of ℘-blowup.
By induction, suppose we are now considering a main binomial B(kτ ) . For each
and every term of the proper transform of the main binomial B(kτ ) in the previously
obtained blowup scheme, we pick a possible “zero” factor to form a pair. Here, we
do not exclude any variable any more. Such a pair is called ð-sets with respect to
B(kτ ) . They give rise to ð-centers with respect to B(kτ ) . The set of all ð-center comes
equipped with a total order. We then blow up the previously obtained scheme along
(the proper transforms) of these ð-centers.
20 YI HU

This gives rise to the final sequential blowups (1.11) in the introduction

Reð → · · · → Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 ) → · · · → Re℘ .

Here again, the index (kτ ) indicates the main binomial B(kτ ) ; the existence of the
index rµ is due to the need to deal with the excessive accumulation of exceptional
parameters; sh simply indicates the corresponding step of the blowup.
In the above, the constructions of ℘- and ð-blowups are discussed in terms of
coordinate variables of the proper transforms of the main binomials on local charts.
In the main text, the constructions of all these ℘- and ð-blowups, like ϑ-blowups,
are done globally via induction.

(From the previous discussions, one sees that the process of ℘- and ð-blowups
is highly inefficient. To provide a concrete example for the whole process, Gr(2, n)
would miss some main points; Gr(3, 6) would be too long to include, and also,
perhaps not too helpful as far as showing (a resolution of) a singularity is concerned.)

• Γ-schemes and their ϑ-, ℘-, ð-transforms.


Fix any integral Γ-scheme ZΓ , considered as a closed subscheme of Um ∩ Gr3,E .
Our goal is to resolve ZΓ when it is singular.
As in the introduction, we have the following instrumental diagram (2.12).
The first two rows follow from the above discussion; we only need to explain the
third and fourth rows.
Here, when ZF[j−1]) (resp. Zeℏ′ ,Γ ) is not contained in the corresponding blowup
center, ZF (resp. Zeℏ,Γ) is, roughly, obtained from the proper transform of ZF
[j]) [j−1])

(resp. Zeℏ′ ,Γ ). When ZF[j−1]) (resp. Zeℏ′ ,Γ ) is contained in the corresponding blowup
center, then ZF (resp. Zeℏ,Γ) is, roughly, obtained from a canonical rational slice
[j])

of the total transform of ZF[j−1]) (resp. Zeℏ′ ,Γ ) under the morphism VF[j] → VF[j−1]
(resp. Veℏ → Veℏ′ ) in the second row. Moreover, every ZF (resp. Zeℏ,Γ ) admits ex-
[j])

plicit defining equations over any standard affine chart of the corresponding smooth
scheme in the first row. Furthermore, in every case, ZF (resp. Zeℏ,Γ ) contains an
[j])

irreducible component †
ZF [j] ,Γ
(resp. Zeℏ,Γ

) such that it maps onto ZΓ birationally.
GRASSMANNIANS AND SINGULARITIES 21

(2.12)
Reð / ··· / Reℏ / Reℏ′ / ··· / RF[j] / RF[j−1] · · · / Um
O O O O O O

? ? ? ? ? ?
Veð / ··· / Veℏ / Veℏ′ / ··· / VF[j] / VF[j−1] · · · / Um ∩ Gr3,E
O O O O O O

? ? ? ? ? ?
Zeð,Γ / ··· / Zeℏ,Γ / Zeℏ′ ,Γ / ··· / ZF[j] ,Γ / ZF[j−1]) ,Γ · · · / ZΓ
O O O O O O
=
? ? ? ? ?
Zeð,Γ
† / ··· / Zeℏ,Γ
† / Zeℏ†′ ,Γ / ··· / †
ZF [j] ,Γ
/ †
ZF [j−1]) ,Γ
··· / ZΓ .

• Smoothness by Jacobian of main binomials and linearized Plücker relations.


We are now ready to explain the smoothness of Zeð,Γ when ZΓ is integral. We first
investigate the smoothness of Veð which is a special case of Zeð,Γ when Γ = ∅.
The question is local. So we focus on an affine chart of V of Reð . Proposition 7.10
mn
provides key properties for the defining equations BV q
, LV,Fm , BV of Veð ∩ V ⊂ V.
As envisioned, we find that the scheme Veð is smooth on the chart V by some
explicit calculations and careful analysis on the Jacobian of the main binomial re-
mn
lations of BV and linearized Plücker relations of LV,Fm . This implies that on the
mn
chart V, the main binomial relations of BV and the linearized Plücker relations of
LV,Fm together generate the ideal of Veð ∩ V. Thus, as a consequence, the binomials
q
of quotient type BV can be discarded from consideration.
Then, the similar calculations and analysis on the Jacobian of the induced main
mn
binomial relations of BV and the induced linearized Plücker relations of LV,Fm for
Zeð,Γ implies that Zeð,Γ is smooth as well, on all charts V. In particular, Ze† , now a ð,Γ

connected component of Zeð,Γ , is smooth, too.


This implies that Zeð,Γ

−→ ZΓ is a resolution, if ZΓ is singular.
The above are done in §9.

• Local resolution via Mnëv universality.


22 YI HU

Upon reviewing Lafforgue’s version of Mnëv universality, we can use the resolution
Zeð,Γ

−→ ZΓ to obtain a local resolution for any singular integral affine scheme. This
is done in §10. The proof in §10 uses only standard techniques.

3. Primary Plücker Relations and In-homogenized Plücker-Ideal

The purpose of this section is to describe a minimal set of Plücker relations so


that they generate the Plücker ideal for a given chart. The approach of this article
depends on these explicit relations. The entire section is elementary.
Fix a pair of positive integers n > 1 and 1 ≤ d < n. In this section, we focus on
Grassmannians Grd,E where E = E1 ⊕ · · · ⊕ En is as introduced in the introduction.
For application to resolution of singularity, it suffices to consider Gr3,E . However,
we choose to work on the general case of Grd,E for the following two reasons. (1)
Working on Gr3,E instead of Grd,E saves us little space or time: if we focus on (2.5)
P
but not the general form s∈SF xus xvs in the construction of ϑ-, ℘-, and ð-blowups,
then the proofs of some key propositions would have to be somewhat case by case,
less conceptual, and hence may be lengthier. However, it is always good to frequently
use the equations of (2.5) and (2.8) as examples to help to understand the notations
and the process. We caution here that replying only on Plücker equations of the
form F̄(123),iuv , 1 ≤ i ≤ 3 from (2.5) (they correspond to Plücker equations of Gr2,E )
might miss some crucial points. (2) As a convenient benefit, the results obtained
and proofs provided for Grd,E here can be directly cited in [8].
All the results of this section are elementary and some might have already been
known. Nonetheless, the development in the current section is instrumental for our
approach. Hence, some good details are necessary.
We make a convention. Let A be a finite set and a ∈ A. Then, we write

A\a := A\{a}.

Also, we use |A| to denote the cardinality of the set A.

3.1. Plücker relations.


Fix a pair of positive integers (n, d) with n > 1 and 1 ≤ d < n. We denote the
set {1, · · · , n} by [n]. We let Id,n be the set of all sequences of distinct integers
GRASSMANNIANS AND SINGULARITIES 23

{1 ≤ u1 < · · · < ud ≤ n}. An element of Id,n is frequently written as u = (u1 · · · ud ).


We also regard an element of Id,n as a subset of d distinct integers in [n]. For
instance, for any u, m ∈ Id,n , u\m takes its set-theoretic meaning. Also, u ∈ [n]\u if
and only if u 6= ui for all 1 ≤ i ≤ d.
As in the introduction, suppose we have a set of vector spaces, E1 , · · · , En such
that every Eα , 1 ≤ α ≤ n, is of dimension 1 over k (or, a free module of rank 1 over
Z), and, we let

E := E1 ⊕ . . . ⊕ En .

For any fixed integer 1 ≤ d < n, the Grassmannian, defined by

Grd,E = {F ֒→ E | dim F = d},

is a projective variety defined over Z.


We have the canonical decomposition
M
∧d E = Ei1 ⊗ · · · ⊗ Eid .
i=(i1 ,··· ,id )∈Id,n

This gives rise to the Plücker embedding of the Grassmannian:

Grd,E ֒→ P(∧d E) = {(pi )i∈Id,n ∈ Gm \(∧d E\{0})},

F −→ [∧d F ],

where Gm is the multiplicative group.


The group (Gm )n /Gm , where Gm is embedded in (Gm )n as the diagonal, acts on
P(∧d E) by

t · pi = ti1 · · · tid pi

where t = (t1 , · · · , tn ) is (a representative of) an element of (Gm )n /Gm and i =


(i1 , · · · , id ). This action leaves Grd,E invariant. The (Gm )n /Gm -action on Grd,E will
only be used in §10.
The Grassmannian Grd,E as a closed subscheme of P(∧d E) is defined by a set of
specific quadratic relations, called Plücker relations. We describe them below.
24 YI HU

For narrative convenience, we will assume that pu1 ···ud is defined for any sequence
of d distinct integers between 1 and n, not necessarily listed in the sequential order
of natural numbers, subject to the relation

(3.1) pσ(u1 )···σ(ud ) = sgn(σ)pu1 ···ud

for any permutation σ on the set [n], where sgn(σ) denotes the sign of the permu-
tation. Furthermore, also for convenience, we set

(3.2) pu := 0,

for any u = (u1 · · · ud ) of a set of d integers in [n] if ui = uj for some 1 ≤ i 6= j ≤ d.


Now, for any pair (h, k) ∈ Id−1,n × Id+1,n with

h = {h1 , · · · , hd−1 } and k = {k1 , · · · , kd+1 },

we have the Plücker relation:


d+1
X
(3.3) Fh,k = (−1)λ−1 ph1 ···hd−1 kλ pk1 ···kλ ···kd+1 ,
λ=1

where “kλ ” means that kλ is deleted from the list.


To make the presentation concise, we frequently succinctly express a general Plücker
relation as
X
(3.4) F = sgn(s)pus pvs ,
s∈SF

for some index set SF , with us , vs ∈ Id,n , where sgn(s) is the ± sign associated with
the quadratic monomial term pus pvs . We note here that sgn(s) depends on how
every of us and vs is presented, per the convention (3.1).

Definition 3.1. Consider any Plücker relation F = Fh,k for some pair (h, k) ∈
Id−1,n × Id+1,n . We let tF + 1 be the number of terms in F . We then define the rank
of F to be tF − 2. We denote this number by rank (F ).

The integer tF , as defined above, will be frequently used throughout.


GRASSMANNIANS AND SINGULARITIES 25

Example 3.2. Consider the Grassmannian Gr(3, 6). Then, the Plücker relation

F(16),(3456) : p163 p456 − p164 p356 + p165 p346

is of rank zero; the Plücker relation

F(12),(3456) : p123 p456 − p124 p356 + p125 p346 − p126 p345

is of rank one.

Let Z[pi ]i∈Id,n be the homogeneous coordinate ring of the Plücker projective space
P(∧d E) and I℘ ⊂ Z[pi ]i∈Id,n be the homogeneous ideal generated by all the Plücker
relations (3.3) or (3.4). We call I℘ the Plücker ideal for the Grassmannian Grd,E .
Then, the graded quotient ring Z[pu ]u∈Id,n /I℘ is the homogeneous coordinate ring
of Grd,E , called the Grassmannian algebra.

3.2. Primary Plücker equations with respect to a fixed affine chart.


In this subsection, we focus on a fixed affine chart of the Plücker projective space
P(∧d E).
Fix any m ∈ Id,n . In P(∧d E), we let

Um := (pm ≡ 1)

stand for the open chart defined by pm 6= 0. Then, the affine space Um comes
equipped with the coordinate variables xu = pu /pm for all u ∈ Id,n \m. In practical
calculations, we will simply set pm = 1, whence the notation (pm ≡ 1) for the chart.
We let
Um (Gr) = Um ∩ Grd,E

be the corresponding induced open chart of Grd,E .


The chart Um (Gr) is canonically an affine space. Below, we explicitly describe
 
n
Υ := − 1 − d(n − d)
d
many specific Plücker relations with respect to the chart Um , called the m-primary
Plücker relations, such that their restrictions to the chart Um define Um (Gr) as a
closed subscheme of the affine space Um .
26 YI HU

To this end, we write m = (m1 · · · md ). We set


m
Id,n = {u ∈ Id,n | |u\m| ≥ 2} ⊂ Id,n

where u and m are also regarded as subsets of integers, and |u\m| denotes the
m
cardinality of u\m. In words, u ∈ Id,n if and only if u = (u1 , · · · , ud ) contains at
least two elements distinct from elements in m = (m1 , · · · , md ). It is helpful to write
explicitly the set I \Im :
d,n d,n

m = {m} ∪ {{u} ∪ (m \ m ) | for all u ∈ [n]\m and all 1 ≤ i ≤ d},


Id,n \Id,n i

/ m if and only if u 6= mi for any 1 ≤ i ≤ d. Then, one calculates and finds


where u ∈
 
m n
|Id,n | = Υ = − 1 − d(n − d),
d
where |Im | denotes the cardinality of Im .
d,n d,n

Further, let a = (a1 · · · ak ) be a list of some elements of [n], not necessarily


mutually distinct, for some k < n. We will write

va = v(a1 · · · ak ) = (va1 · · · ak ) and av = (a1 · · · ak )v = (a1 · · · ak v),

each is considered as a list of some elements of [n], for any v ∈ [n]\a.


m . We let u denote the smallest
Now, take any element u = (u1 , · · · , ud ) ∈ Id,n 0

integer in u\m. We then set

h = u \ u0 and k = (u0 m1 · · · md ),

where u \ u0 = u \ {u0 } and u is regarded as a set of integers.


This gives rise to the Plücker relation Fh,k , taking of the following form

(3.5) Fh,k = p(u\u0 )u0 pm − p(u\u0 )m1 pu0 (m\m1 ) + · · · + (−1)d p(u\u0 )md pu0 (m\md ) ,

where m \ mi = m \ {mi } and m is regarded as a set of integers, for all i ∈ [d].


We give a new notation for this particular equation: we denote it by
d
X
(3.6) Fm,u = p(u\u0 )u0 pm + (−1)i p(u\u0 )mi pu0 (m\mi ) ,
i=1
m . To simplify the notation, we introduce
because it only depends on m and u ∈ Id,n

u r = u \ u0 , m
ci = m \ mi , for all i ∈ [d].
GRASSMANNIANS AND SINGULARITIES 27

Then, (3.6) becomes


d
X
(3.7) Fm,u = pm pur u0 + (−1)i pur mi pu0 m
di .
i=1
r
We point out here that u and u u0 may differ by a permutation.

Definition 3.3. We call the Plücker equation Fm,u of (3.7) a primary Plücker
equation for the chart Um = (pm ≡ 1). We also say Fm,u is m-primary. The term
pm pu is called the leading term of Fm,u .

(One should not confuse Fm,u with the expression of a general Plücker equation
Fh,k : we have (m, u) ∈ I2d,n for the former and (h, k) ∈ Id−1,n × Id+1,n for the latter.)
One sees that the correspondence between Im and the set of m-primary Plücker
d,n

equations is a bijection.

3.3. In-homogenized Plücker ideal with respect to a fixed affine chart.


Following the previous subsection, we continue to fix an element m ∈ Id,n and
focus on the chart Um of P(∧d E).
We will write Id,n \m for Id,n \{m}.
m , by (3.7), it gives rise to the m-primary equation
Given any u ∈ Id,n
d
X
Fm,u = pm pur u0 + (−1)i pur mi pu0 m
di .
i=1

If we set pm = 1 and let xw = pw , for all w ∈ Id,n \m, then it becomes


d
X
(3.8) F̄m,u = xur u0 + (−1)i xur mi xu0 m
di .
i=1

Definition 3.4. We call the relation (3.8) the in-homogenized (or the localized)
m-primary Plücker relation corresponding to u ∈ Im . We call the unique distin-
d,n

guished variable, xu (which may differ xur u0 by a sign), the leading variable of the
in-homogenized Plücker relation F̄m,u .

Throughout this paper, we often express an m-primary Plücker equation F as


X X
(3.9) F = sgn(s)pus pvs = sgn(sF )pm pus + sgn(s)pus pvs
F
s∈SF s∈SF \sF
28 YI HU

where sF is the index for the leading term of F , and SF \sF := SF \{sF }. Then,
upon setting pm = 1 and letting xw = pw for all w ∈ Id,n \m, we can write the
corresponding in-homogenized m-primary Plücker equation F̄ as
X X
(3.10) F̄ = sgn(s)xus xvs = sgn(sF )xuF + sgn(s)xus xvs
s∈SF s∈SF \sF

where xuF := xus is the leading variable of F̄ .


F

Definition 3.5. Let F be an m-primary Plücker relation and F̄ its in-homogenization


with respect to the chart Um . We set tF̄ = tF and rank (F̄ ) = rank (F ).

For any u ∈ Id,n \m, we let xu = pu /pm for all u ∈ Id,n \m. Then, we can
identify the coordinate ring of Um with k[xu ]u∈Id,n \m . We let I℘,m be the ideal
of k[xu ]u∈Id,n \m obtained from the Plücker ideal I℘ be setting pm = 1 and letting
xu = pu for all u ∈ Id,n \m. We call I℘,m the in-homogenized Plücker ideal on the
chart Um .

Definition 3.6. For any u ∈ Im d,n , we define the m-rank of u (resp. xu ) to be the

rank of its corresponding primary Plücker equation Fm,u .

Proposition 3.7. The in-homogenized Plücker ideal I℘,m is generated by


m }.
Fm := {F̄m,u | u ∈ Id,n

Consequently, the chart Um (Gr) = Um ∩ Grd,E comes equipped with the set of free
variables
m}
VarUm := {xu | u ∈ Id,n \{m}\Id,n
and is canonically isomorphic to the affine space with the above variables as its
coordinate variables.

Proof. (This proposition is elementary; it serves as the initial check of an induction


for some later proposition; we provide sufficient details for completeness.)
It suffices to observe that for any u ∈ Im , its corresponding in-homogenized
d,n

Plücker primary Plücker equation F̄m,u is equivalent to an expression of the leading


variable xu as a polynomial in the free variables of VarUm . For instance, one can
check this by induction on the m-rank, rank (u), of u, as follows.
GRASSMANNIANS AND SINGULARITIES 29

Suppose rank (u) = 0. Then, up to a permutation, we may write

u = (m \ {mi mj })vu

where mi , mj ∈ m for some 1 ≤ i, j, ≤ d, and u < v ∈


/ m. Then, we have

(3.11) F̄u,u : xu + (−1)i xur mi xum j


ci + (−1) xur mj xum
cj ,

where ur = (m \ {mi mj })v. One sees that xur mi , xum


ci , xur mj and xum
cj belong to

VarUm . Hence, the statement holds.


Now suppose that rank (u) > 0. Using (3.8), we have
d
X
F̄m,u : xu ru
0 + (−1)i xur mi xu0 m
ci .
i=1

ci , i ∈ [d], belong to VarUm . Note also that


Note that all variables xu0 m

rank (ur mi ) = rank (u) − 1,

provided that pur mi is not identically zero, that is, it is a well-defined Plücker variable
(see (3.2)). Thus, applying the inductive assumption, any such xur mi is a polynomial
in the variables of VarUm . Therefore, F̄m,u , is equivalent to an expression of xu as
a polynomial in the variables of VarUm .
m } and let V (J)
Let J be the ideal of k[xu ]u∈Id,n \m generated by {F̄m,u | u ∈ Id,n
the subscheme of Um defined by J. By the above discussion, V (J) is canonically
isomorphic to the affine space of dimension d(n − d) with the variables of VarUm as
its coordinate variables. Since Um (Gr) ⊂ V (J), we conclude Um (Gr) = V (J). 

Definition 3.8. We call the variables in


m}
VarUm := {xu | u ∈ Id,n \m\Id,n

the m-basic Plücker variables. When m is fixed and clear from the context, we just
call them basic variables.

Only non-basic Plücker variables correspond to m-primary Plücker equations.


Observe that for all Plücker relations F , we have 0 ≤ rank(F ) ≤ d − 2. Hence,
for any 0 ≤ r ≤ d − 2, we let
r m
Fm = {F̄m,u | rank(Fm,u ) = r, u ∈ Id,n }.
30 YI HU

Then, we have
[
r
Fm = Fm .
0≤r≤d−2

Then, one observes the following easy but useful fact.

m with rank (F
Proposition 3.9. Fix any 0 ≤ r ≤ d−2 any u ∈ Id,n m m,u ) = r. Then,
the leading variable xu of F̄m,u does not appear in any relation in
0 r−1 r
Fm ∪ · · · ∪ Fm ∪ (Fm \ F̄m,u ).

To close this subsection, we raise a concrete question. Fix the chart (pm ≡ 1). In
k[xu ]u∈Id,n \m , according to Proposition 3.7, every in-homogenized Plücker equation
F̄h,k on the chart Um can be expressed as a polynomial in the in-homogenized
m
primary Plücker relations F̄m,u with u ∈ Id,n . It may be useful in practice to find
such an expression explicitly for an arbitrary Fh,k . For example, for the case of
Gr(2, 5), this can be done as follows.

Example 3.10. For Gr(2, 5), we have five Plücker relations:

F1 = p12 p34 − p13 p24 + p14 p23 , F2 = p12 p35 − p13 p25 + p15 p23 ,

F3 = p12 p45 − p14 p25 + p15 p24 , F4 = p13 p45 − p14 p35 + p15 p34 ,

F5 = p23 p45 − p24 p35 + p25 p34 .

On the chart (p45 ≡ 1), F3 , F4 , and F5 are primary. One calculates and finds

p45 F1 = p34 F3 − p24 F4 + p14 F5 ,

p45 F2 = p35 F3 − p25 F4 + p15 F5 .

In addition, the Jacobian of the in-homogenized Plücker equations of F̄3 , F̄4 , F̄5 with
respect to all the variables, x12 , x14 , x15 , x13 , x35 , x34 , x23 , x24 , x25 , is given by
 
1 x25 x24 0 0 0 0 0 0
 
 0 0 0 1 x x 0 0 0 .
 14 15 
0 0 0 0 0 0 1 x35 x34

There, one sees visibly a (3 × 3) identity minor.


GRASSMANNIANS AND SINGULARITIES 31

3.4. Ordering the set of all primary Plücker equations.


Fix m ∈ Id,n . We consider the set
m}
Fm = {F̄m,u | u ∈ Id,n

as in Proposition 3.7. As in the paragraph above Proposition 3.9, we have


[
r
Fm = Fm ,
1≤r≤d−2

r
where Fm m } for all 0 ≤ r ≤ d − 2.
= {F̄m,u | rank(F̄m,u ) = r, u ∈ Id,n
We will provide a total order on the set Fm . This ordering will be fixed and used
throughout.
We first provide a partial order ≺℘ on the set Fm .

i j
Definition 3.11. Let F̄ ∈ Fm and Ḡ ∈ Fm . Then, we say

F̄ ≺℘ Ḡ if i < j.

This gives rise to the partially ordered set (Fm , ≺℘ ).

In what follows, we extend ≺℘ to make Fm a totally ordered set.

Definition 3.12. Let K be any fixed totally ordered finite set, with its order denoted
by <. Consider any two subsets η ⊂ K and ζ ⊂ K with the cardinality n for some
positive integer n. We write η = (η1 , · · · , ηn ) (respectively, of ζ = (ζ1 , · · · , ζn )) as
an array according to the ordering of K. We say η <lex ζ if the left most nonzero
number in the vector η − ζ is negative, or more explicitly, if we can express

η = {t1 < · · · < tr−1 < sr < · · · }

ζ = {t1 < · · · < tr−1 < tr < · · · }

such that sr < tr for some integer r ≥ 1. We call <lex the lexicographic order induced
by (K, <).
Likewise, we say η <invlex ζ if the right most nonzero number in the vector η − ζ
is negative, or more explicitly, if we can express

η = {· · · < sr < tr+1 < · · · < tn }


32 YI HU

ζ = {· · · < tr < tr+1 < · · · < tn }


such that sr < tr for some integer r ≥ 1. We call <invlex the reverselexicographic
order induced by (K, <).

This definition can be applied to the set

Id,n = {(i1 < i2 < · · · < id ) | 1 ≤ iµ ≤ n, ∀ 1 ≤ µ ≤ d}

for all d and n. Thus, we have equipped the set Id,n with both the lexicographic
ordering “ <lex ” and the reverse lexicographic ordering “ <invlex ”.

Definition 3.13. Consider any u, v ∈ Id,n \m.


ci u and v = m
Suppose u = m m , for some m , m ∈
cj v are two elements of Id,n \m\Id,n i j

m and u, v ∈ [n]\m. We say


u <℘ v
ci <lex m
if u < v or when u = v, m cj .
m . We say
Suppose u and v are two elements of Id,n

u <℘ v

if one of the following three holds:


• rankm u < rankm v;
• rankm u = rankm v, u\m <lex v\m;
• rankm u = rankm v, u\m = v\m, and m ∩ u <lex m ∩ v.

Definition 3.14. Consider any two Plücker variables xu and xv . We say

xu <℘ xv if u <℘ v.
i
Consider any two distinct primary equations, F̄m,u , F̄m,v ∈ Fm of the same rank
i for some 0 ≤ i ≤ d − 2, with u 6= v. We say

F̄m,u <℘ F̄m,v if u <℘ v.

The above together with Definition 3.11 provides an induced total order on the
set Fm . We denote the totally ordered set by (Fm , <℘ ). Hence, we can write

Fm = {F̄1 <℘ · · · <℘ F̄Υ }.


GRASSMANNIANS AND SINGULARITIES 33

In what follows, when comparing two Plücker variables xu and xv or two m-


primary Plücker equations, we exclusively use <℘ . Thus, throughout, for simplicity,
we will simply write < for <℘ . A confusion is unlikely.

4. A Singular Local Birational Model V for Grd,E

The purpose of this section is to establish a local model V , birational to Grd,E , such
that all terms of all m-primary Plücker equations can be separated in the defining
main binomial relations of V in a smooth ambient scheme RF .

4.1. The construction of V ⊂ RF .


Consider the fixed affine chart Um of the Plücker projective space P(∧d E). For
P
any F̄ ∈ Fm , written as F = s∈SF sgn(s)pus pvs , we let PF be the projective space
with homogeneous coordinates written as [x(us ,vs ) ]s∈SF . For convenience, we let

(4.1) ΛF = {(us , vs ) | s ∈ SF }.

This is an index set for the homogeneous coordinates of the projective space PF . To
avoid duplication, we make a convention:

(4.2) x(us ,vs ) = x(vs ,us ) , ∀ s ∈ SF , ∀ F̄ ∈ Fm .

If we write (us , vs ) in the lexicographical order, i.e., we insist us <lex vs , then the
ambiguity is automatically avoided. Howerer, the convention is still useful.

Definition 4.1. We call x(us ,vs ) a ̺-variable of PF , or simply a ̺-variable. To


distinguish, we call a Plücker variable, xu with u ∈ Id,n \m, a ̟-variable.

Fix k ∈ [Υ]. We introduce the natural rational map


Q
(4.3) Θ[k] : P(∧d E) ❴ ❴ ❴/ i∈[k] PF i
Y
[pu ]u∈Id,n −→ [pu pv ](u,v)∈ΛFi
i∈[k]

where [pu ]u∈Id,n is the homogeneous Plücker coordinates of a point in P(∧d E). When
restricting Θ[k] to Um , it gives rise to
Q
(4.4) Θ̄[k] : Um ❴ ❴ ❴/ i∈[k] PFi
34 YI HU

We let
 Q
(4.5) PF[k]  / P(∧d E) × i∈[k] PFi

be the closure of the graph of the rational map Θ[k] , and


 Q
(4.6) Um,F[k]  / RF[k] := Um × i∈[k] PF i

be the closure of the graph of the rational map Θ̄[k] .

Definition 4.2. Fix any k ∈ [Υ]. We let

R[k] = k[pu ; x(vs ,us ) ]u∈Id,n ,s∈SFi ,i∈[k].

A polynomial f ∈ R[k] is called multi-homogeneous if it is homogenous in [pu ]u∈Id,n


and is homogenous in [x(vs ,us ) ]s∈SFi , for every i ∈ [k]. A multi-homogeneous polyno-
mial f ∈ R[k] is ̺-linear if it is linear in [x(vs ,us ) ]s∈SFi , whenever it contains some
̺-variables of PFi , for any i ∈ [k].

We set R0 := k[pu ]u∈Id,n . Then, corresponding to the embedding (4.5), there


exists a degree two homomorphism

(4.7) ϕ[k] : R[k] −→ R0 , x(us ,vs ) → pus pvs

for all s ∈ SFi , i ∈ [k].


We then let

(4.8) ϕ̄[k] : R[k] −→ R0 , x(us ,vs ) → xus xvs

for all s ∈ SFi , i ∈ [k], be the in-homogenization of ϕ[k] with respect to the chart
Um = (pm ≡ 1). This corresponds to the embedding (4.6).
We are mainly interested in the case when k = Υ. Hence, we set

R := R[Υ] , ϕ := ϕ[Υ] , ϕ̄ := ϕ̄[Υ] .

We let kerm-h ϕ[k] (resp. kerm-h ϕ̄[k] ) denote the set of all multi-homogeneous
polynomials in ker ϕ[k] (resp. ker ϕ̄[k] ).
GRASSMANNIANS AND SINGULARITIES 35
Q
Lemma 4.3. The scheme PF[k] , as a closed subscheme of P(∧d E) × i∈[k] PFi , is
defined by kerm-h ϕ[k]. In particular, the scheme Um,F[k] , as a closed subscheme of
Q
RF[k] = Um × i∈[k] PFi , is defined by kerm-h ϕ̄[k] .

Proof. This is immediate. 

We need to investigate kerm-h ϕ[k] .


Consider any f ∈ kerm-h ϕ[k] . We express it as the sum of its monic monomials
(monomials with constant coefficients 1)
X
f= mi .
P
We have ϕ[k](f ) = ϕ[k] (mi ) = 0 in R0 . Thus, the set of the monic monomials
{mi } can be grouped into minimal groups to form partial sums of f so that the
images of elements of each group are identical and the image of the partial sum of
each minimal group equals 0 in R0 . When ch.k = 0, this means each minimal group
consists of a pair (mi , mj ) and its partial sum is the difference mi − mj . When
ch.k = p > 0 for some prime number p, this means each minimal group consists
of either (1): a pair (mi , mj ) and mi − mj is a partial sum of f ; or (2): exactly p
elements mi1 , · · · , mip and mi1 + · · · + mip is a partial sum of f . But, the relation
mi1 + · · · + mip is always generated by the relations mia − mib for all 1 ≤ a, b ≤ p.
Thus, regardless of the characteristic of the field k, it suffices to consider binomials
m − m′ ∈ kerm-h ϕ[k] .

Example 4.4. Consider Gr3,E . Then, the following binomials belong to kerm-h ϕ[k]
for any fixed k ∈ [Υ].
Fix a, b, c ∈ [k], all being distinct.

x(12a,13b) x(13a,12c) x(12b,13c)

(4.9) −x(13a,12b) x(12a,13c) x(13b,12c)

Fix a, b, c, ā, b̄, c̄ ∈ [k], all being distinct.

x(12a,3bc) x(13a,2b̄c̄) x(13ā,2bc) x(12ā,3b̄c̄)

(4.10) −x(13a,2bc) x(12a,3b̄c̄) x(12ā,3bc) x(13ā,2b̄c̄) ,


36 YI HU

Fix a, b, c, a′ , ā, b̄, c̄ ∈ [k], all being distinct.

x(12a,13a′ ) x(13a,2bc) x(12a′ ,3b̄c̄) x(12ā,3bc) x(13ā,2b̄c̄)

(4.11) −x(13a,12a′ ) x(12a,3bc) x(13a′ ,2b̄c̄) x(13ā,2bc) x(12ā,3b̄c̄) .

These binomials are arranged so that one sees visibly the matching for multi-homogeneity.

Lemma 4.5. Fix any i ∈ [k]. We have

(4.12) pu′ pv′ x(u,v) − pu pv x(u′ ,v′ ) ∈ kerm-h ϕ[k] .

where x(u,v) , x(u′ ,v′ ) are any two distinct ̺-variables of PFi . Likewise, we have

(4.13) xu′ xv′ x(u,v) − xu xv x(u′ ,v′ ) ∈ kerm-h ϕ̄[k] .

Proof. This is trivial. 

Let Al (resp. Pl ) be the affine (resp. projective) space of dimension l for some
positive integer l with coordinate variables (x1 , · · · , xl ) (resp. with homogeneous
coordinates [x1 , · · · , xl ]). A monomial m is square-free if x2 does not divide m for
every coordinate variable x in the affine space. A polynomial is square-free if all of
its monomials are square-free.

For any m − m′ ∈ kerm-h ϕ[k] , we define deg̺ (m − m′ ) to be the total degree of


m (equivalently, m′ ) in ̺-variables of R[k] .
P
For any F = s∈SF pus pvs with F̄ ∈ Fm and s ∈ SF , we write Xs = x(us ,vs ) .
Recall here that we have set ϕ = ϕ[Υ] : R = R[Υ] → R0 .

Lemma 4.6. Suppose a binomial m − m′ ∈ kerm-h ϕ[k] with deg̺ (m − m′ ) > 0.


Then, we have
Qℓ Qℓ
m=h i=1 mi and m′ = h i=1 m′i
for some monomial h in R[k] and some positive integer ℓ such that for every i ∈ [ℓ]
mi − m′i ∈ kerm-h ϕ[k] and is of the following form:

(4.14) ϕ(X1 )X2 − ϕ(X1′ )X2′ ,

where every of X1 , X2 , X1′ , and X2′ is a monomial of R in ̺-variables only (i.e.,


without ̟-variables; here we allow X1 = X1′ = 1) such that
GRASSMANNIANS AND SINGULARITIES 37

(1) X1 X2 − X1′ X2′ ∈ kerm-h ϕ and is ̺-linear;


(2) ϕ(X1 X2 ) (equivalently, ϕ(X1′ X2′ )) is a square-free monomial;
(3) for any F̄ ∈ Fm and s ∈ SF , suppose xus xvs divides m (resp. m′ ), then
Xs = x(us ,vs ) divides X1 (resp. X1′ ) in one of the relations of (4.14).

Proof. We prove by induction on deg̺ (m − m′ ).


Suppose deg̺ (m − m′ ) = 1.
Then, we can write
m − m′ = f x(u,v) − gx(u′ ,v′ )

for some f, g ∈ R0 , and two ̺-variables of PFi , x(u,v) and x(u′ ,v′ ) for some i ∈ [k]. If
x(u,v) = x(u′ ,v′ ) , then one sees that f = g and m − m′ = 0. Hence, we assume that
x(u,v) 6= x(u′ ,v′ ) . Then, we have

f pu pv = gpu′ pv′ .

Because x(u,v) and x(u′ ,v′ ) are two distinct ̺-variables of PFi , one checks from the
definition that the two sets
{pu , pv }, {pu′ , pv′ }

are disjoint. Consequently,


pu pv | g, pu′ pv′ | f.

Write
g = g1 pu pv , f = f1 pu′ pv′ .

Then we have
pu pv pu′ pv′ (f1 − g1 ) = 0 ∈ R0 .

Hence, f1 = g1 . Then, we have

m − m′ = h(pu′ pv′ x(u,v) − pu pv x(u′ ,v′ ) )

where h := f1 = g1 . This implies the statement, in this case. Observe that in such
a case, we have that m − m′ is generated by the relations of (4.12).
Suppose Lemma 4.6 holds for deg̺ < e for some positive integer e > 1.
Consider deg̺ (m − m′ ) = e.
38 YI HU

By the multi-homogeneity of (m − m′ ), we can write

(4.15) m − m′ = nXs − n′ Xt

such that Xs and Xt are the ̺-variables of PFi corresponding to some s, t ∈ SFi for
some i ∈ [k], and n, n′ ∈ R[k] .
If s = t, then m − m′ = Xs (n − n′ ). Hence, the statement follows from the
inductive assumption applied to (n − n′ ) ∈ kerm-h ϕ[k] since deg̺ (n − n′ ) = e − 1.
We suppose now s 6= t. Let f¯ = nxus xvs − n′ xut xvt . Then, f¯ ∈ kerm-h ϕ[k].
Further, deg f¯ < e. By the inductive assumption, we can write
̺


Y ℓ
Y

nxus xvs = h(xus xvs ns ) ni , n xut xvt = h(xut xvt nt ) n′j
i=1 j=1

for some integer ℓ ≥ 1, with n0 = xus xvs ns and n′0 = (xut xvt nt ) such that for each
0 ≤ i ≤ ℓ, it determines (matches) a unique 0 ≤ i′ ≤ ℓ such that (ni − n′i′ ) is of
the form of (4.14) and verifies all the properties (1) - (3) of Lemma 4.6. Consider
n0 = xus xvs ns . It matches n′0′ . By the multi-homogeneity of n0 − n′0′ , (4.14) and
(1) of Lemma 4.6, we can write n′0′ = xut′ xvt′ nt′ for some t′ ∈ SFi and nt′ ∈ R[k] .
Therefore, by switching t with t′ if t 6= t′ , and re-run the above arguments, without
loss of generality, we can assume t′ = t and n0 = xus xvs ns matches n′0 = (xut xvt nt ).
Further, by re-indexing {n′j | j ∈ [ℓ]} if necessary, we can assume that ni matches
n′i for all 1 ≤ i ≤ l.
Qℓ Qℓ
Now, note that we have n = h(ns ) i=1 ni , n′ = h(n′t ) i=1 ni . Hence

Y ℓ
Y

m = h(ns Xs ) ni , m = h(n′t Xt ) ni .
i=1 i=1

We let m0 = ns Xs and mi = ni for all i ∈ [ℓ]; m′0 = (nt Xt ) and mi = n′i for all
i ∈ [ℓ]. Then, one checks directly that Lemma 4.6 holds for m − m′ .
This proves the lemma. 

Definition 4.7. Let Bb[k] be the set of all binomial relations of (4.14) that verify
Lemma 4.6 (1) - (3); let Be[k] the in-homogenizations with respect to (pm ≡ 1) of all
binomial relations of Bb[k] .
GRASSMANNIANS AND SINGULARITIES 39

Corollary 4.8. The ideal kerm-h ϕ[k] is generated by Bb[k] . Consequently, the ideal
kerm-h ϕ̄[k] is generated by Be[k] .

Proof. Take any binomial (m − m′ ) ∈ kerm-h ϕ[k] with deg̺ (m − m′ ) > 0. We


express, by Lemma 4.6,

Y ℓ
Y

m − m = h( mi − m′i )
i=1 i=1

such that mi − m′i ∈ Bb[k] for all i ∈ [ℓ]. Then, we have



Y ℓ−1
Y ℓ−1
Y ℓ
Y

m − m = h( mi − m′ℓ mi + m′ℓ mi − m′i )
i=1 i=1 i=1 i=1

ℓ−1
Y ℓ−1
Y ℓ−1
Y
= h((mℓ − m′ℓ ) mi + m′ℓ ( mi − m′i )).
i=1 i=1 i=1
Thus, by a simple induction on the integer ℓ, the corollary follows. 

We now let Θ̄[k],Gr be the restriction of Θ̄[k] to Um (Gr) = Um ∩ Grd,E :


Q
(4.16) Θ̄[k],Gr : Um (Gr) ❴ ❴ ❴/ i∈[k] PFi
Y
[xu ]u∈Id,n −→ [xu xv ](u,v)∈ΛFi .
i∈[k]

We let
 Q 
(4.17) Vm,F[k]  / Um (Gr) × i∈[k] PFi
/ RF[k]

be the closure of the graph of the rational map Θ̄[k],Gr.


Then, one sees that Vm,F[k] is the proper transform of Um (Gr) in Um,F[k] under
the birational morphism Um,F[k] −→ Um .
Since we focus on the chart Um , below we write VF[k] = Vm,F[k] .
By construction, there exists the natural forgetful map

(4.18) RF[k] −→ RF[k−1]

and it induces a birational morphsim

(4.19) ρF[k] : VF[k] −→ VF[k−1] .


40 YI HU

Definition 4.9. We let Bi (resp. B[k] ) be the set of all binomial relations in (4.13)
for any fixed i ∈ [k] (resp. for all i ∈ [k]). We set Bpre-q = Be[k] \B[k] . An element of
[k]
pre-q
B[k] is called a binomial of pre-quotient type.
Qk
Lemma 4.10. The scheme VF[k] , as a closed subscheme of RF[k] = Um × i=1 PFi ,

is defined by the following relations

(4.20) BFi ,(s,t) : x(us ,vs ) xut xvt − x(ut ,vt ) xus xvs , ∀ s, t ∈ SFi \sFi , i ∈ [k],

(4.21) BFi ,(sFi ,s) : x(us ,vs ) xuF − x(m,uF ) xus xvs , ∀ s ∈ SFi \sFi , i ∈ [k],
i i

pre-q
(4.22) B[k] ,
X
(4.23) LFi : sgn(s)x(us ,vs ) , i ∈ [k],
s∈SFi
X
(4.24) F̄j : sgn(s)xus xvs , j ∈ [Υ]
s∈SFj
P
where F̄i is expressed as sgn(sFi )xuF + s∈SFi \sFi sgn(s)xus xvs for every i ∈ [k].
i

Proof. By Lemma 4.3 and Corollary 4.8, the scheme Um,F[k] , as a closed subscheme
of RF[k] is defined by the relations in (4.20), (4.21), and (4.22). Thus, VF[k] , being
the proper transform of Um (Gr), as a closed subscheme of RF[k] , is defined by the
relations in (4.20), (4.21), and (4.22) together with the proper transforms of the
P
in-homogenized Plücker equations F̄ = s∈SF sgn(s)xus xvs for all F̄ ∈ Fm .
It suffices to show that under the presence of (4.20) and (4.21), LFi and F̄i imply
each other for all i ∈ [k].
Fix any i ∈ [k]. Take any s ∈ SFi . Consider the binomial relations of Bi

(4.25) x(u,v) xus xvs − xu xv x(us ,vs ) ,

for all (u, v) ∈ ΛFi (cf. (4.1)). By multiplying sgn(s) to (4.25) and adding together
all the resulted binomials, we obtain,

(4.26) xus xvs LFi = x(us ,vs ) F̄i , mod (hBi i),

where hBi i is the ideal generated by the relations in Bi . As neither of xus xvs and
x(us ,vs ) belong to the ideal of VF[k] , and VF[k] is integral, we see that LFi and F̄i
imply each other for all i ∈ [k]. 
GRASSMANNIANS AND SINGULARITIES 41

For conciseness, we set the following

Vm := Vm,F[Υ] , Um,F := Um,F[Υ] , RF := RF[Υ] .

Then, we have the following diagram


  Q
Vm  / Um,F  / RF = Um × F̄ ∈Fm PF

  

Um (Gr)  / Um Um .

In what follows, we will sometimes write V for Vm , as we will exclusively focus


on the chart Um , throughout, unless otherwise stated.
We also set
pre-q
Bpre-q = B[Υ] .

By the case of Lemma 4.10 when k = Υ, we have


Q
Corollary 4.11. The scheme Vm , as a closed subscheme of RF = Um × F̄ ∈Fm PF ,
is defined by the following relations

(4.27) BF,(s,t) : x(us ,vs ) xut xvt − x(ut ,vt ) xus xvs , ∀ s, t ∈ SF \sF

(4.28) BF,(sF ,s) : x(us ,vs ) xuF − x(m,uF ) xus xvs , ∀ s ∈ SF \sF ,

(4.29) Bpre-q ,
X
(4.30) LF : sgn(s)x(us ,vs ) ,
s∈SF
X
(4.31) F̄ : sgn(s)xus xvs ,
s∈SF
P
for all F̄ ∈ Fm with F̄ being expressed as sgn(sF )xuF + s∈SF \sF sgn(s)xus xvs .

Definition 4.12. A binomial equation BF,(sF ,s) of (4.28) with s ∈ SF \sF is called
a main binomial equation. We let

BFmn = {BF,(sF ,s) | s ∈ SF \sF } and Bmn = ⊔F̄ ∈Fm BFmn .

A binomial equation BF,(s,t) of (4.27) with s, t ∈ SF \sF and s 6= t is called a residual


binomial equation. We let

BFres = {BF,(s,t) | s, t ∈ SF \sF } and Bres = ⊔F̄ ∈Fm BFres .


42 YI HU

Recall that an element of Bpre-q is called a binomial relation of pre-quotient type.


P
Definition 4.13. Given any F̄ ∈ Fm , written as F̄ = s∈SF sgn(s)xus xvs , the
linear equation LF as in (4.30)
X
LF : sgn(s)x(us ,vs )
s∈SF

is called the linearized Plücker equation with respect to F̄ (or F ).

We observe here that


Y X X
(4.32) dim( PF ) = |SF \sF | = |BFmn | = |Bmn |,
F̄ ∈Fm F̄ ∈Fm F̄ ∈Fm

where |K| denotes the cardinality of a finite set K.

4.2. ̟-divisors, ̺-divisors, and the standard charts of RF .


From earlier, we have the set ΛF = {(us , vs ) | s ∈ SF }. This is an index set for
the homogeneous coordinates of the projective space PF , and is also an index set for
all the variables that appear in the linearized Plücker equation L̄F of (4.30). To be
used later, we also set
ΛFm = ⊔F̄ ∈Fm ΛF .
Q
Definition 4.14. Consider the scheme RF = Um × F̄ ∈Fm PF . Recall that the
affine chart Um comes equipped with the coordinate variables {xu }u∈Id,n \m . For any
u ∈ Id,n \m, we set
Xu := (xu = 0) ⊂ RF .

We call Xu the Plücker divisor, in short, the ̟-divisor, of RF associated with u.


We let D̟ be the set of all ̟-divisors on the scheme RF .
Q
Definition 4.15. In addition to the ̟-divisors, the scheme RF = Um × F̄ ∈Fm PF
also comes equipped with the divisors

X(u,v) := (x(u,v) = 0)

for all (u, v) ∈ ΛFm . We call X(u,v) the ̺-divisor corresponding to (u, v). We let
D̺ be the set of all ̺-divisors of RF .
GRASSMANNIANS AND SINGULARITIES 43

Definition 4.16. Fix k ∈ [Υ]. For every F̄i ∈ Fm with i ∈ [k], choose and fix an
arbitrary element sFi ,o ∈ SFi . Then, the scheme RF[k] is covered by the affine open
charts of the form
Y Y
Um × (x(us ,vs ) ≡ 1) ⊂ RF[k] = Um × PF i .
Fi ,o Fi ,o
i∈[k] i∈[k]

We call such an affine open subset a standard chart of RF[k] , often denoted by V.

Fix any standard chart V as above. We let


Y Y
V′ = Um × (x(us ,vs ) ≡ 1) ⊂ RF[k−1] = Um × PF i .
F ,o F ,o i i
i∈[k−1] i∈[k−1]

Then, this is a standard chart of RF[k−1] , uniquely determined by V. We say V lies


over V′ . In general, suppose V′′ is a standard chart of RF[j] with j < k − 1. Via
induction, we say V lies over V′′ if V′ lies over V′′ .
Note that the standard chart V of RF[k] in the above definition is uniquely indexed
by the set

(4.33) ΛoF[k] = {(usF ,o , vsF ,o ) ∈ ΛFi | i ∈ [k]}.


i i

Given ΛoF[k] , we let


[
Λ⋆F[k] = ( ΛFi )\ΛoF[k] .
i∈[k]

We set ΛoFm := ΛoF[Υ] and Λ⋆Fm := ⋆


ΛF[Υ] .
To be cited as the initial cases of certain inductions later on, we need the following
two propositions.

Proposition 4.17. Consider any standard chart


Y
V = Um × (x(us ,vs ) ≡ 1)
F ,o F ,o i i
i∈[k]

of RF[k] , indexed by ΛoF[k] as above. It comes equipped with the set of free variables

VarV = {xV,w , xV,(u,v) | w ∈ Id,n \m, (u, v) ∈ Λ⋆F[k] }

and is isomorphic to the affine space with the variables in VarV as its coordinate
variables. Moreover, on the standard chart V, we have
(1) the divisor Xw ∩ V is defined by (xV,w = 0) for every w ∈ Id,n \ m;
44 YI HU

(2) the divisor X(u,v) ∩ V is defined by (xV,(u,v) = 0) for every (u, v) ∈ Λ⋆F[k] .

Proof. Recall that Um = (pm ≡ 1). Then, we let xV,w = xw for all w ∈ Id,n \m. Now
consider every i ∈ [k]. Upon setting x(us , vs ) ≡ 1, we let xV,(us ,vs ) = x(us ,vs ) be
Fi ,0 Fi ,0

the in-homogenization of x(us ,vs ) for all s ∈ SFi \sFi ,o . From here, the statement is
straightforward to check. 

Proposition 4.18. Let the notation be as in Propsotion 4.17. Then, the scheme
VF[k] ∩ V, as a closed subscheme of V is defined by the following relations

(4.34) xV,(us ,vs ) xV,ut xV,vt − xV,(ut ,vt ) xV,us xV,vs , ∀ s, t ∈ SFi \sFi , i ∈ [k],

(4.35) xV,(us ,vs ) xV,uF − xV,(m,uF ) xV,us xV,vs , ∀ s ∈ SFi \sFi , i ∈ [k],
i i

pre-q
(4.36) BV,[k] ,
X
(4.37) LV,Fi : sgn(s)xV,(us ,vs ) , i ∈ [k],
s∈SFi
X
(4.38) F̄V,j : sgn(s)xV,us xV,vs , k < j ≤ Υ.
s∈SFj

pre-q pre-q
where the equations of BV,[k] are the in-homogenizations of the equations of B[k] .

Proof. By applying Lemma 4.10, it suffices to show that F̄V,i is redundant on the
chart V for all i ∈ [k]. Fix any i ∈ [k]. By (4.26) and take s = sFi ,o on the chart V
(cf. Definition 4.16), we obtain,

(4.39) xV,us xV,vs LV,Fi = F̄V,i


Fi ,o Fi ,o

because x(us ,vs ) ≡ 1 on the chart V. This implies the statement. 


Fk ,o Fk ,o

For any f ∈ R, we let deg̟ f be the degree of f considered as a polynomial in


̟-variables only.

Definition 4.19. Let f ∈ Bpre-q be a binomial relation of pre-quotient type. We say


f is a binomial relation of quotient type if deg̟ f = 0, that is, it does not contain
any ̟-variable. We let Bq be the set of all binomial relations of quotient type. Fix
q pre-q
a standard chart V as in Definition 4.16, we let BV,[k] ⊂ BV,[k] be the subset of all
the in-homogenized binomial relations of quotient type.
GRASSMANNIANS AND SINGULARITIES 45

We write
q q pre-q pre-q
BV := BV,[Υ] , BV := BV,[Υ] .

We let R̺ be the subring of R consisting of polynomials with ̺-variables only.


Then, binomial relations of quotient type belong to R̺ .
q
By Lemma 4.6, all binomials of Bq and BV are ̺-linear, in particular, they are
square-free.

Proposition 4.20. Let the notation be as in Proposition 4.18 for k = Υ. Then, the
scheme V ∩ V, as a closed subscheme of the chart V of RF , is defined by

(4.40) BV,(s,t) : xV,(us ,vs ) xV,ut xV,vt − xV,(ut ,vt ) xV,us xV,vs , ∀ s, t ∈ SF \sF ,

(4.41) BV,(sF ,s) : xV,(us ,vs ) xV,uF − xV,(m,uF ) xV,us xV,vs , ∀ s ∈ SF \sF ,
q
(4.42) BV ,
X
(4.43) LV,F : sgn(s)xV,(us ,vs )
s∈SF
P
for all F̄ ∈ Fm with F̄ being expressed as sgn(sF )xuF + s∈SF \sF sgn(s)xus xvs .
Here, we set
xV,m ≡ 1; xV,(us , vs ) ≡ 1, ∀ F̄ ∈ Fm ,
F,o F,o

q
Moreover, every binomial BV ∈ BV is linear in ̺-variables, in particular, square-free.

Proof. By Proposition 4.18 for k = Υ, the scheme V ∩ V, as a closed subscheme of


pre-q
the chart V of RF , is defined by relations in (4.40), (4.41) , (4.43), and BV .
pre-q q
It remains to reduce BV to BV .
pre-q q
We claim that any relation f of BV is generated by relations of BV .
We prove it by induction on deg̟ (f ).
When deg̟ (f ) = 0, the statement holds trivially.
Assume that statement holds for deg̟ < e for some e > 0.
Consider deg̟ (f ) = e.
By Lemma 4.6, we can write

f = xus xvs ns − xut xvt nt


46 YI HU

for some s, t ∈ SF and some F̄ ∈ Fm . Because on the chart V, we have

xus xvs − xus xvs x(us ,vs ) , xut xvt − xus xvs x(ut ,vt ) ,
F,o F,o F,o F,o

where sF,o is as in Definition 4.16 with x(us ,vs ) ≡ 1, we get


F,o F,o

f = xus xvs (x(us ,vs ) ns − x(ut ,vt ) nt ).


F,o F,o

pre-q
Observe that (x(us ,vs ) ns −x(ut ,vt ) nt ) ∈ BV . Since deg̟ (x(us ,vs ) ns −x(ut ,vt ) nt ) < e,
the statement then follows from the inductive assumption. 

mn res q
Definition 4.21. We let BV (respectively, BV , BV ) be the set of all binomial rela-
mn
tions of (4.41) (respectively, (4.40), (4.42)). We call relations of BV (respectively,
res q
BV , BV ) main (respectively, residual, of quotient type) binomial on the chart V.
We let
q
B = Bmn ⊔ Bres ⊔ Bq and BV = BV
mn res
∪ BV ∪ BV .

We let LV,Fm be the set of all linear equations of (4.43). We call relations of LV,Fm
linearized Plücker relations on the chart V.

5. ϑ-Blowups

We begin now the process of “removing” zero factors of main binomials by sequen-
tial smooth blowups. It is divided into three subsequences. The first are ϑ-blowups.
To start, it is useful to fix some terminology, used throughout.

5.1. Some conventions on blowups.


Let X be a scheme over the base field k. When we blow up the scheme X along
the ideal (the homogeneous ideal, respectively) I = hf0 , · · · , fm i, generated by some
e as the graph of the closure
elements f0 , · · · , fm , we will realize the blowup scheme X
of the rational map
f : X 99K Pm ,

x → [f0 (x), · · · , fm (x)].

Then, upon fixing the generators f0 , · · · , fm , we have a natural embedding

(5.1) e 
X / X × Pm .
GRASSMANNIANS AND SINGULARITIES 47

We let

(5.2) e −→ X
π:X

be the induced blowup morphism.


We will refer to the projective space Pm as the factor projective space of the blowup
corresponding to the generators f0 , · · · , fm . We let [ξ0 , · · · , ξm ] be the homogeneous
coordinates of the factor projective space Pm , corresponding to (f0 , · · · , fm ).
When X is smooth and the center of the blowup is also smooth, then, the scheme
e as a closed subscheme of X × Pm , is defined by the relations
X,

(5.3) fi ξj − fj ξi , for all 0 ≤ i 6= j ≤ m.

Definition 5.1. Suppose that the scheme X is covered by a set {V′} of open subsets,
called (standard) charts.
Fix any 0 ≤ i ≤ m. We let

(5.4) e
V = (V′ × (ξi 6= 0)) ∩ X.

We also often express this chart as

e
V = (V′ × (ξi ≡ 1)) ∩ X.

e and will be called a standard chart of X


It is an open subset of X, e lying over the
(standard) chart V′ of X. Note that every standard chart of X e lies over a unique
e is covered by the finitely many standard charts.
(standard) chart V′ of X. Clearly, X
In general, we let

ek −→ X
X ek−1 −→ · · · −→ X
e0 := X

fj → X
be a sequence of blowups such that every blowup X ej−1 is as in (5.1), j ∈ [k].

Consider any 0 ≤ j < k. Let V (resp. V′′ ) be a standard chart of X ek (resp. of


ej ). Let V′ be the unique standard chart V′ of X
X ek−1 such that V lies over V′ . Via
induction, we say V lies over V′′ if V′ equals to (when j = k − 1) or lies over V′′
(when j < k − 1).
48 YI HU

We keep the notation as above. Let Xe → X be a blowup as in (5.1); we let


e lying over a unique (standard) chart V′ of X; let
V be a standard chart of X,
πV,V′ : V −→ V′ be the induced projection.

Definition 5.2. Assume that V (resp. V′ ) is isomorphic to an affine space and


comes equipped with a set of coordinate variables VarV (resp. VarV′ ). Let y ∈ VarV
(resp. y ′ ∈ VarV′ ) be a coordinate variable of V (resp. V′ ). We say the coordinate
variable y is a proper transform of the coordinate variable y ′ if the divisor (y = 0)
on the chart V is the proper transform of the divisor (y ′ = 0) on the chart V′ .

Keep the notation and assumption as in Definition 5.2.


We assume in addition that the induced blowup morphism

π −1 (V′ ) −→ V′

corresponds to the blowup of V′ along the coordinate subspace of V′ defined by

Z = {y0′ = · · · = ym

= 0}

with {y0′ , · · · , ym

} ⊂ VarV′ . As earlier, we let Pm be the corresponding factor projec-
tive space with homogeneous coordinates [ξ0 , · · · , ξm ], corresponding to (y0′ , · · · , ym

).
Without loss of generality, we assume that the standard chart V corresponds to
(ξ0 ≡ 1), that is,
e
V = (V′ × (ξ0 ≡ 1)) ∩ X.
Then, we have that V, as a closed subscheme of V′ × (ξ0 ≡ 1), is defined

(5.5) yi′ − y0′ ξi , for all i ∈ [m].

The following proposition is standard and will be applied throughout.

Proposition 5.3. Keep the notation and assumption as above. In addition, we let
e → X.
E be the exceptional divisor of the blowup X
Then, the standard chart V comes equipped with a set of free variables

VarV = {ζ, y1, · · · , ym ; y := y ′ | y ′ ∈ VarV′ \{y0′ , · · · , ym }}

where ζ := y0′ , yi := ξi , i ∈ [m], and is isomorphic to the affine space with the
variables in VarV as its coordinate variables such that
GRASSMANNIANS AND SINGULARITIES 49

(1) E ∩ V = (ζ = 0); we call ζ the exceptional variable/parameter of E on V;


(2) yi ∈ VarV is a proper transform of yi′ ∈ VarV′ for all i ∈ [m];
(3) y ∈ VarV is a proper transform of y ′ ∈ VarV for all y ′ ∈ VarV′ \{y0′ , · · · , ym

}.

Proof. It is straightforward from (5.5). 

Let Al be the affine space of dimension l for some positive integer l with the set of
coordinate variables VarAl . Let m be a monomial in VarAl . Then, for every variable
x ∈ VarAl , we let degx m be the degree of x in m.

Definition 5.4. Keep the notation and assumption as in Proposition 5.3. In addi-
tion, we let
φ = {y0′ , · · · , ym

} ⊂ VarV′ .

Let BV′ = TV0 ′ − TV1 ′ be a binomial with variables in VarV′ . We let


m
X
mφ,T i ′ = degyj′ (TVi ′ ), i = 0, 1,
V
j=0

lφ,BV′ = min{mφ,T 0 ′ , mφ,T 1 ′ }.


V V

Applying (5.5), we substitute yi′ by y0′ ξi , for all i ∈ [m], into BV′ and switch y0′ by
∗ ′
ζ and ξi by yi with i ∈ [m] to obtain the pullback πV,V′ BV′ where πV,V′ : V −→ V

is the induced projection. We then let


∗ lφ,B ′
(5.6) BV = (πV,V′ BV′ )/ζ V .

We call BV , a binomial in VarV , the proper transform of BV′ .


In general, for any polynomial fV′ in VarV′ such that fV′ does not vanish identi-
cally along Z = (y0′ = · · · = ym
′ ∗
= 0), we let fV = πV,V′ fV′ . This is the pullback, but

for convenience, we also call fV the proper transform of fV′ . (We will only apply
this to linearized Plücker relations.)
∗ lψ,B ′ ∗
Moreover, suppose ζ appears in BV = (πV,V′ BV′ )/ζ V or in f
V = πV,V′ fV′ , and

is obtained through the substitution yi′ by y0′ ξi (note here that ζ := y0′ and i needs
not to be unique), then we say that the exceptional parameter ζ is acquired by yi′ . In
general, for sequential blowups, if ζ is acquired by y ′ and y ′ is acquired by y ′′ , then
we also say ζ is acquired by y ′′.
50 YI HU

Lemma 5.5. We keep the same assumption and notation as in Definition 5.4.
We let TV′ ,B (resp. TV,B ) be any fixed term of BV′ (resp. BV ). Consider any
y ∈ VarV \ζ and let y ′ ∈ VarV′ be such that y is the proper transform of y ′ . Then,
y b | TV,B if and only if y ′b | TV′ ,B for all integers b ≥ 0.

Proof. This is clear from (5.6). 

Definition 5.6. We keep the same assumption and notation as in Definition 5.4.
Consider an arbitrary binomial BV′ (resp. BV ) with variables in VarV′ (resp.
VarV ). Let z′ ∈ V′ (resp. z ∈ V) be any fixed closed point of the chart. We say BV′
(resp. BV ) terminates at z′ (resp. z) if (at least) one of the monomial terms of BV′
(resp. BV ), say, TV′ ,B (resp. TV,B ), does not vanish at z′ (resp. z). In such a case,
we also say TV′ ,B (resp. TV,B ) terminates at z′ (resp. z).

5.2. Main binomial equations: revisited.


Recall that we have chosen and fix the total order “ < ” on Fm and we have
listed it as

Fm = {F̄1 < · · · < F̄Υ }.

P
Now, fix any k ∈ [Υ] and consider Fk . We express Fk = s∈SFk sgn(s)pus pvs . Its
P
corresponding linearized Plücker equation can be expressed as s∈SFk sgn(s)x(us ,vs ) ,
denoted by LFk . We let sFk ∈ SFk be the index for the leading term of Fk , written
as sgn(sFk )pm puF . Correspondingly, the leading term of the in-homogenization
k

F̄k of Fk (resp. the linearized Plücker equation LFk ) is defined to be sgn(sFk )xuF
k

(resp. sgn(sFk )x(m,uF ) ). We then choose and fix an arbitrary order on the index
k

set SFk such that sFk is the smallest element. For preciseness, we choose and use the
following order: for any two s, t ∈ SFk \sFk with s 6= t, we say s < t if (us , vs ) <lex
(ut , vt ) (see Definition 3.12). As in Definition 3.1, we let (tFk + 1) be the number of
terms in Fk . Then, we can list SFk as

SFk = {sFk < s1 < · · · < stFk }.


GRASSMANNIANS AND SINGULARITIES 51
Q
By Corollary 4.11, the scheme V as a closed subscheme of RF = Um × F̄ ∈Fm PF
is defined by the following relations

(5.7) Bres , Bpre-q ,

(5.8) B(kτ ) : x(usτ ,vsτ ) xuF − x(m,uF ) xusτ xvsτ , ∀ sτ ∈ SFk \sFk , 1 ≤ τ ≤ tFk ,
k k
X
(5.9) LFk : sgn(s)x(us ,vs ) ,
s∈SFk
X
(5.10) F̄k = sgn(s)xus xvs
s∈SFk

for all k ∈ [Υ].

+
Definition 5.7. Given any binomial equation B(kτ ) as in (5.8), we let T(kτ ) =

x(usτ ,vsτ ) xuF , called the plus-term of B(kτ ) , and T(kτ ) = x(m,uFk ) xusτ xvsτ , called the
k
+ −
minus-term of B(kτ ) . Then, we have B(kτ ) = T(kτ ) − T(kτ ) .

We do not name any term of a binomial of Bres ∪Bpre-q a plus-term or a minus-term


since the two terms of such a binomial are indistinguishable.
In addition, we let BFmn
k
= {B(kτ ) | τ ∈ [tFk ]} for any k ∈ [Υ]. Then, we have
G
Bmn = BFmn
k
= {B(kτ ) | k ∈ [Υ], τ ∈ [tFk ]}.
k∈[Υ]

We let

(5.11) IndexBmn = {(kτ ) | k ∈ [Υ], τ ∈ [tFk ]}

be the index set of Bmn . Then, the set Bmn comes equipped with a total order “ < ”
induced by the lexicographic order on IndexBmn , that is,

B(kτ ) < B(k′ τ ′ ) ⇐⇒ (k, τ ) <lex (k ′ , τ ′ ).

5.3. ϑ-centers and ϑ-blowups.


Besides serving as a part of the process of “removing” zero factors of the main
binomial relations, the reason to perform ϑ-blowups first is to eliminate all residual
binomial relations by making them dependent on the main binomial relations.
52 YI HU

Recall that the scheme Reϑ[0] := RF comes equipped with two kinds of divisors:
̟-divisors Xw for all w ∈ Id,n \m (Definition 4.14) and ̺-divisors X(u,v) for all
(u, v) ∈ ΛFm (Definition 4.15).
m . We let
Definition 5.8. Fix any u ∈ Id,n

ϑu = (Xu , X(m,u) ).

We call it the ϑ-set with respect to u. We then call the scheme-theoretic intersection

Zϑu = Xu ∩ X(m,u)

the ϑ-center with respect to u.

We let
m },
Θ = {ϑu | u ∈ Id,n ZΘ = {Zϑu | u ∈ Id,n m }.
m
We let Θ, respectively, ZΘ , inherit the total order from Id,n . Thus, if we write

Im
d,n = {u1 < · · · < uΥ }

and also write ϑuk = ϑ[k] , Zϑu = Zϑ[k] , then, we can express
k

ZΘ = {Zϑ[1] < · · · < Zϑ[Υ] }.

We then blow up RF along Zϑ[k] , k ∈ [Υ], in the above order. More precisely,
we start by setting Reϑ[0] := RF . Suppose Reϑ[k−1] has been constructed for some
k ∈ [Υ]. We then let
Reϑ[k] −→ Reϑ[k−1]
be the blowup of Reϑ[k−1] along the proper transform of Zϑ[k] , and we call it the
ϑ-blowup in (ϑ[k] ).
The above gives rise to the following sequential ϑ-blowups

(5.12) Reϑ[Υ] → · · · → Reϑ[1] → Reϑ[0] := RF ,

Every blowup Reϑ[j] −→ Reϑ[j−1] comes equipped with an exceptional divisor, de-
noted by Eϑ[j] . Fix k ∈ [Υ]. For any j < k, we let Eϑ[k] ,j be the proper transform
of Eϑ[j] in Reϑ[k] . For notational consistency, we set Eϑ[k] = Eϑ[k] ,k . We call the
divisors Eϑ ,j , j ≤ k, the exceptional divisors on Reϑ . For every w ∈ Id,n \ m, we
[k] [k]
GRASSMANNIANS AND SINGULARITIES 53

let Xϑ[k] ,w be the proper transform of Xw in Reϑ[k] , still called ̟-divisor; for every
(u, v) ∈ ΛFm , we let Xϑ[k] ,(u,v) ∩ V be the proper transform of X(u,v) in Reϑ[k] , still
called ̺-divisor.

5.4. Properties of ϑ-blowups.


By Definition 5.1, the scheme Reϑ[k] is covered by a set of standard charts.

Proposition 5.9. Consider any standard chart V of Reϑ[k] , lying over a unique
chart V[0] of Reϑ[0] = RF . We suppose that the chart V[0] is indexed by ΛoFm =
{(vsF,o , vsF,o ) | F̄ ∈ Fm } (cf. (4.33)). As earlier, we have Λ⋆Fm = ΛFm \ΛoFm .
Then, the standard chart V comes equipped with

a subset eV ⊂ Id,n \m and a subset dV ⊂ Λ⋆Fm

such that every exceptional divisor of Reϑ[k] with Eϑ[k] ∩ V 6= ∅ is either labeled by a
unique element w ∈ eV or labeled by a unique element (u, v) ∈ dV . We let Eϑ[k] ,w
be the unique exceptional divisor on the chart V labeled by w ∈ eV ; we call it an ̟-
exceptional divisor. We let Eϑ[k] ,(u,v) be the unique exceptional divisor on the chart
V labeled by (u, v) ∈ dV ; we call it an ̺-exceptional divisor. (We note here that
being ̟-exceptional or ̺-exceptional is strictly relative to the given standard chart.)
Further, the standard chart V comes equipped with the set of free variables
 
 ε , δ w ∈ eV , (u, v) ∈ dV 
V,w V,(u,v)
(5.13) VarV := ,
 xV,w , xV,(u,v) w ∈ Id,n \m\eV , (u, v) ∈ Λ⋆F \dV 
m
such that it is canonically isomorphic to the affine space with the variables in (5.13)
as its coordinate variables. Moreover, on the standard chart V, we have

(1) the divisor Xϑ[k] ,w ∩ V is defined by (xV,w = 0) for every w ∈ Id,n \ m\eV ;
(2) the divisor Xϑ[k] ,(u,v) ∩ V is defined by (xV,(u,v) = 0) for every (u, v) ∈
Λ⋆Fm \dV ;
(3) the divisor Xϑ[k] ,w ∩ V does not intersect the chart for all w ∈ eV ;
(4) the divisor Xϑ[k] ,(u,v) does not intersect the chart for all (u, v) ∈ dV ;
(5) the ̟-exceptional divisor Eϑ[k] ,w ∩ V labeled by an element w ∈ eV is define
by (εV,w = 0) for all w ∈ eV ;
54 YI HU

(6) the ̺-exceptional divisor Eϑ[k] ,(u,v) ∩ V labeled by an element (u, v) ∈ dV is


define by (δV,(u,v) = 0) for all (u, v) ∈ dV ;
(7) any of the remaining exceptional divisor of Reϑ[k] other than those that are
labelled by some some w ∈ eV or (u, v) ∈ dV does not intersect the chart.

Proof. When k = 0, we have Reϑ[0] = RF . In this case, we set

eV = dV = ∅.

Then, the statement follows from Proposition 4.17 with k = Υ.


We now suppose that the statement holds for Reϑ[k−1] for some k ∈ [Υ].
We consider Reϑ . [k]

As in the statement, we let V be a standard chart of Reϑ[k] , lying over a (necessarily


unique) standard chart V′ of Reϑ . [k−1]

If (m, uk ) ∈ ΛoF[k] (cf. (4.33)), then V′ does not intersect the proper transform
of the blowup center Zϑk and V → V′ is an isomorphism. In this case, we let
VarV = VarV′ , eV′ = eV , and dV = dV′ . Then, the statements on V′ carry over to V.
/ ΛoF[k] .
In what follows, we assume (m, uk ) ∈
Consider the embedding

Reϑ[k] −→ Reϑ[k−1] × Pϑ[k]

where Pϑ[k] is the factor projective space with homogeneous coordinates [ξ0 , ξ1 ] cor-
responding to (Xuk , X(m,uk ) ). We let Eϑ[k] be the exceptional divisor created by the
blowup Reϑ → Reϑ
[k]
. [k−1]

First, we consider the case when

V = Reϑ[k] ∩ (V′ × (ξ0 ≡ 1).

We let Zϑ′ k be the proper transform of the ϑ-center Zϑk in Reϑ[k−1] . Then, in this
case, on the chart V′ , we have

Zϑ′ k ∩ V′ = {xV′ ,uk = xV′ ,(m,uk ) = 0}


GRASSMANNIANS AND SINGULARITIES 55

where xV′ ,uk (resp. xV′ ,(m,uk ) ) is the proper transform of xu (resp. x(m,uk ) ) on the
chart V′ . Then, V as a closed subset of V′ × (ξ0 ≡ 1) is defined by

xV′ ,(m,uk ) = xV′ ,uk ξ1 .

We let

eV = uk ⊔ eV′ , dV = dV′ , and

εV,uk = xV′ ,uk , xV,(m,uk ) = ξ1 ; yV = yV′ , ∀ yV′ ∈ VarV′ \{xV′ ,uk , xV′ ,(m,uk ) }.

Observe that Eϑ[k] ∩ V = (εV,uk = 0) and xV,(m,uk ) = ξ1 is the proper transform of


xV′ ,(m,uk ) . By the inductive assumption on the chart V′ , one verifies directly that
(1) - (7) of the proposition hold (cf. Proposition 5.3).
Next, we consider the case when

V = Reϑ[k] ∩ (V′ × (ξ1 ≡ 1).

Then, V as a closed subset of V′ × (ξ1 ≡ 1) is defined by

xV′ ,uk = xV′ ,(m,uk ) ξ0 .

We let

eV = eV′ , dV = {(m, uk )} ⊔ dV′ , and

δV,(m,uk ) = xV′ ,(m,uk ) , xV,uk = ξ0 ; yV = yV′ , ∀ yV′ ∈ VarV′ \{xV′ ,uk , xV′ ,(m,uk ) }.

Observe that Eϑ[k] ∩ V = (δV,(m,uk ) = 0) and xV,uk = ξ0 is the proper transform


of xV′ ,uk . By the inductive assumption on the chart V′ , like in the above case, one
checks directly that (1) - (7) of the proposition hold.
This proves the proposition. 

Observe here that xV,u with u ∈ eV and xV,(u,v) with (u, v) ∈ δV are not variables
in VarV . For notational convenience, to be used throughout, we make a convention:

(5.14) • xV,u = 1 if u ∈ eV ; • xV,(u,v) = 1 if (u, v) ∈ dV .


56 YI HU

For any k ∈ [Υ], the ϑ-blowup in (ϑ[k] ) gives rise to

Veϑ[k]  Reϑ[k]

(5.15) /

 

V  / RF ,

where Veϑ[k] is the proper transform of V in Reϑ[k] .


Alternatively, we can set Veϑ[0] := VF . Suppose Veϑ[k−1] has been constructed for
some k ∈ [Υ]. We then let Veϑ ⊂ Reϑ be the proper transform of Veϑ
[k] [k]
.
[k−1]

Definition 5.10. Fix any standard chart V of Reϑ[k] lying over a unique standard
chart V′ of Reϑ[k−1] for any k ∈ [Υ]. When k = 0, we let BV and LV,F be as in
Proposition 4.20 for any B ∈ Bmn ∪ Bres ∪ Bq and F̄ ∈ Fm . Consider any fixed
general k ∈ [Υ]. Suppose BV′ and LV′ ,F have been constructed over V′ . Applying
Definition 5.4, we obtain the proper transforms on the chart V

BV , ∀ B ∈ Bmn ∪ Bres ∪ Bq ; LV,F , ∀ F̄ ∈ Fm .

We need the following notations.


mn res
Fix any k ∈ [Υ]. We let B<k (resp. B<k or L<k ) be the set of all main (resp.
mn
residual or linear Plücker) relations corresponding to F < Fk . Similarly, we let B>k
res
(resp. B>k , L>k ) be the set of all main (resp. residual or linear Plücker) relations
corresponding to F > Fk . Likewise, replacing < by ≤ or > by ≥, we can introduce
mn res mn res
B≤k , B≤k , and L≤k or B≥k , B≥k , and L≥k . Then, upon restricting the above to a
mn res
fixed standard chart V, we obtain BV,<k , BV,<k , LV,<k , etc..
Recall from the above proof, we have

Reϑ[k] ⊂ Reϑ[k−1] × Pϑk

where Pϑk be the factor projective space of the blowup Reϑ[k] −→ Reϑ[k−1] . We write
Pϑk = P[ξ0 ,ξ1 ] such that [ξ0 , ξ1] corresponds to (Xuk , X(m,uk ) ).

Definition 5.11. (cf. Definition 5.1) Let V′ be any standard chart on Reϑ[k−1] . Then,
we call
V = Reϑ[k] ∩ (V′ × (ξ0 ≡ 1))
GRASSMANNIANS AND SINGULARITIES 57

a ̟-standard chart of Reϑ[k] ; we call

V = Reϑ[k] ∩ (V′ × (ξ1 ≡ 1))

a ̺-standard chart of Reϑ[k] .

Proposition 5.12. We keep the notation and assumptions in Proposition 5.9.


Suppose (m, uk ) ∈ ΛoF[k] or V is a ̺-standard chart. Then, we have that the
scheme Veϑ[k] ∩ V, as a closed subscheme of the chart V of Reϑ[k] , is defined by
q mn
(5.16) BV , BV,<k , LV,<k ,

(5.17) BV,(sFk ,s) : xV,(us ,vs ) xV,uk − x̃V,us x̃V,vs , ∀ s ∈ SFk \sFk ,
X
(5.18) LV,Fk : sgn(sF )δV,(m,uk ) + sgn(s)xV,(us ,vs ) ,
s∈SF \sF
mn res
(5.19) BV,>k , BV,>k , LV,>k ,

where x̃V,us and x̃V,vs are some monomials in VarV .


Suppose (m, uk ) ∈/ ΛoF[k] and V is a ̟-standard chart. Then, we have that the
scheme Veϑ[k] ∩ V, as a closed subscheme of the chart V of Reϑ[k] , is defined by
q mn
(5.20) BV , BV,<k , LV,<k ,

(5.21) BV,(sFk ,s) : xV,(us ,vs ) − xV,(m,uk ) x̃V,us x̃V,vs , ∀ s ∈ SFk \sFk ,
X
(5.22) LV,Fk : sgn(sF )εV,uk xV,(m,uk ) + sgn(s)xV,(us ,vs ) ,
s∈SF \sF
mn res
(5.23) BV,>k , BV,>k , LV,>k ,

where x̃V,us and x̃V,vs are some monomials in VarV .


Moreover, for any binomial B ∈ Bmn ⊔ B>k
res
, BV is ̺-linear and square-free.
Furthermore, consider an arbitrary binomial B ∈ Bq and its proper transform BV
on the chart V. Let TV,B be any fixed term of BV . Then, TV,B is ̺-linear and admits
at most one ϑ-exceptional parameter in the form of δ(m,u) for some (m, u) ∈ dV or
εu x(m,u) for some u ∈ eV . In particular, it is square-free.

Proof. We follow the notation as in the proof of Proposition 5.9.


When k = 0, we have (Veϑ[0] ⊂ Reϑ[0] ) = (VF ⊂ RF ). Then, the statement follows
from Proposition 4.20.
58 YI HU

Suppose that the statement holds for (Veϑ[k−1] ⊂ Reϑ[k−1] ) for some k ∈ [Υ].
We now consider (Veϑ ⊂ Reϑ ).
[k] [k]

As in the proof of Proposition 5.9, we let V be a standard chart of Reϑ[k] lying


over a (necessarily unique) standard chart V′ of Reϑ[k−1] . Also, V lies over a unique
standard chart V[0] of of Reϑ . We let πV,V : V → V[0] be the induced projection.
[0] [0]

To prove the statement about the defining equations of Veϑ[k] ∩V in V, by applying


the inductive assumption to V′ , it suffices to prove that the proper transform of any
residual binomial of Fk depends on the main binomials on the chart V.
For that purpose, we take any two s, t ∈ SFk \sFk and consider the residual bino-
mial BFk ,(s,t) (cf. (4.27)).
Suppose (m, uk ) ∈ ΛoF[k] , hence xV,(m,uk ) ≡ 1 on the chart V. In this case, the
blowup along (the proper transform of) Zϑ[k] does not affect the chart V′ of Reϑ[k−1] .
Likewise, suppose V is a ̺-standard chart. Then, (m, uk ) ∈ dV , hence xV,(m,uk ) = 1
on the chart V by (5.14). In any case, one calculates and finds that we have the
following two main binomials

BV,(sFk ,s) : xV,(us ,vs ) xV,uk − x̃V,us x̃V,vs ,

BV,(sFk ,t) : xV,(ut ,vt ) xV,uk − x̃V,ut x̃V,vt ,



where x̃V,w = πV,V [0]
xV[0] ,w denoted the pullback for any w ∈ Id,n \m. Similarly, one
calculates and finds that we have

BV,(s,t) : xV,(us ,vs ) x̃V,ut x̃V,vt − xV,(ut ,vt ) x̃V,us x̃V,vs .

Then, one verifies directly that we have

BV,(s,t) = xV,(us ,vs ) BV,(sFk ,t) − xV,(ut ,vt ) BV,(sFk ,s) .

This proves the statement about the defining equations of Veϑ[k] ∩ V in V.


Moreover, consider any B ∈ Bmn with respect to Fj . Observe that x(m,uk )
uniquely appears in the main binomials with respect to Fk ; xuk only appears in
the main binomials with respect to Fk and the minus terms of certain main binomi-
als of Fj with j > k. It follows that BV is ̺-linear and square-free.
GRASSMANNIANS AND SINGULARITIES 59

Likewise, consider any B ∈ Bres with respect to Fj with j > k. It is of the form

B(s,t) : x(us ,vs ) xut xvt − x(ut ,vt ) xus xvs

for some s 6= t ∈ SFj . Observe here that B does not contain any ̺-variable of the
form x(m,u) and the ̟-variables in B are identical to those of the minus terms of
the corresponding main binomials. Hence, the same line of the proof above for main
binomials implies that BV is ̺-linear and square-free.
Further, consider any B ∈ Bq . If BV′ does not contain xV′ ,(m,uk ) or (m, uk ) ∈ ΛoF[k] ,
then the form of BV′ remains unchanged (except for the meanings of its variables).
/ ΛoF[k] . Note that the
Suppose next that BV′ contains xV′ ,(m,uk ) and (m, uk ) ∈
proper transform of the ϑ-center ϑ[k] on the chart V′ equals to (xV′ ,uk , xV′ ,(m,uk ) ).
Thus, from the chart V′ to the ̺-standard chart V, we have that xV′ ,(m,uk ) be-
comes δV,(m,uk ) in BV . By Lemma 4.6 (2), applied to the variable pm (before
in-homogenization), we see that any fixed term TB of B contains at most one ̺-
m . Hence, one sees that the last statement
variables of the form x(m,u) with u ∈ Id,n
on BV holds, in this case.
Thus, this proves the statement of the proposition when (m, uk ) ∈ ΛoF[k] or when
V is a ̺-standard chart.
/ ΛoF[k] and V is a ̟-standard chart.
Next, we consider the case when (m, uk ) ∈
Again, to prove the statement about the defining equations of Veϑ[k] ∩ V in V, it
suffices to prove that the proper transform of any residual binomial of Fk depends
on the main binomials on the chart V.
To show this, we again take any two s, t ∈ SFk \sFk .
On the chart V, we have the following two the main binomials

BV,(sFk ,s) : xV,(us ,vs ) − xV,(m,uk ) x̃V,us x̃V,vs ,

BV,(sFk ,t) : xV,(ut ,vt ) − xV,(m,uk ) x̃V,ut x̃V,vt .

We also have the following residual binomial

BV,(s,t) : xV,(us ,vs ) x̃V,ut x̃V,vt − xV,(ut ,vt ) x̃V,us x̃V,vs .


60 YI HU

Then, we have

BV,(s,t) = x̃V,ut x̃V,vt BV,(sFk ,s) − x̃V,us x̃V,vs BV,(sFk ,t) .

Thus, the statement of the proposition about the equations of Veϑ[k] ∩ V follows.
Next, consider any B ∈ Bmn . The fact that BV is ̺-linear and square-free follows
from the same line of proof in the previous case.
Finally, consider any B ∈ Bq . If BV′ does not contain xV′ ,(m,uk ) , then the form
of BV′ remains unchanged. Suppose next that BV′ contains xV′ ,(m,uk ) . Again, the
proper transform of the ϑ-center ϑ[k] on the chart V′ equals to (xV′ ,uk , xV′ ,(m,uk ) ).
Hence, from the chart V′ to the ̟-standard chart V, we have that xV′ ,(m,uk ) turns
into εV,uk xV,(m,uk ) in BV . Then, again, by applying Lemma 4.6 (2), applied to
the variable pm (before in-homogenization), we have that any fixed term TB of B
m
contains at most one ̺-variables of the form x(m,u) with u ∈ Id,n . Hence, one sees
that the last statement on BV holds.
This completes the proof of the proposition. 

We need the final case of ϑ-blowups.


We set Reϑ := Reϑ[Υ] . Veϑ := Veϑ[Υ] .

Corollary 5.13. Let the notation be as in Proposition 4.18 for k = Υ. Then, the
scheme Veϑ ∩ V, as a closed subscheme of the chart V of Reϑ = Reϑ[Υ] , is defined by
q mn
(5.24) BV , BV , LV,Fm .
mn q
Further, for any binomial BV ∈ BV ∪ BV , it is ̺-linear and square-free.

Corollary 5.14. Let Xϑ,(m,uk ) be the proper transform of X(m,uk ) in Reϑ . Then

Veϑ ∩ Xϑ,(m,uk ) = ∅, ∀ k ∈ [Υ].

Further, on any chart V of Reϑ , either x(m,uk ) = 1 or the variable xV,(m,uk ) exists
in VarV and is invertible along Veϑ ∩ V for all k ∈ [Υ].

Proof. Fix any standard chart V.


If V lies over the chart (x(m,uk ) ≡ 1) of RF , that is, (m, uk ) ∈ ΛoF[k] , then the
statement follows from the definition.
GRASSMANNIANS AND SINGULARITIES 61

If V lies over a ̺-standard chart of Reϑ[k] , then the fact that Veϑ ∩ Xϑ,(m,uk ) = ∅
follows from Proposition 5.9 (4); x(m,uk ) = 1 by the convention (5.14).
Suppose V lies over a ̟-standard chart of Reϑ[k] . Then, in this case, we have the
following main binomial relation

(5.25) BV,(sFk ,sF,o) : 1 − xV,(m,uk ) xV,us xV,vs


F,o F,o

because xV,(us ,us ) ≡ 1 with (usF,o , usF,o ) ∈ ΛoF[k] and xV,uk = 1 by (5.14). This
F,o F,o

implies that xV,(m,uk ) is nowhere vanishing along Veϑ ∩ V. 

6. ℘-Blowups

Besides serving as the necessary part of the process of “removing” zero factors of
the main binomial relations, another purpose of ℘-blowups is to help to control the
proper transforms of the binomial relations of quotient type.

6.1. The initial setup: (℘(11) r0 ).


Our initial scheme in (℘(11) r0 ) is Re(℘(11) r0 ) := Reϑ .
On the scheme Reϑ , we have three kinds of divisors:
• the proper transforms Xϑ,w of ̟-divisors Xw for all w ∈ Id,n \m;
• the proper transforms Xϑ,(u,v) of ̺-divisors X(u,v) for all (u, v) ∈ ΛFm ;
• the proper transforms Eϑ,[k] ⊂ Reϑ of the ϑ-exceptional divisors Eϑ ⊂ Reϑ [k] [k]
for
all k ∈ [Υ].
The set of all ϑ-exceptional divisors of Re(℘(11) r0 ) := Reϑ can be expressed as

E(℘(11) r0 ) = {E(℘(11) r0 ),(11)0h = Eϑ,[h] | h ∈ [σ(11)0 ]}, where σ(11)0 := Υ.

As a set of the initial data, we need to introduce the instrumental notion: “association”
with multiplicity as follows.

Definition 6.1. Consider any main binomial relation B ∈ Bmn written as

B = BV,(sFk ,s) = TB+ − TB− : xV,(us ,vs ) xV,uk − x(m,uk ) xV,us xV,vs

for some k ∈ [Υ] and s ∈ SFk , where uk = uFk . Consider any ̟-divisor, ̺-divisor,
or ϑ-exceptional divisor Y on Reϑ .
62 YI HU

Let Y = Xϑ,u be any ̟ divisor for some u ∈ Id,n . We set


(
1, if u = uk
mY,T + =
B
0, otherwise.
(
1, if u = us or u = vs ,
mY,T − =
B
0, otherwise.
Let Y = Xϑ,(u,v) be any ̺ divisor. We set
(
1, if (u, v) = (us , vs )
mY,T + =
B
0, otherwise.

Due to Corollary 5.14, we do not associate X(m,uk ) with TB− . Hence, we set

mY,T − = 0.
B

Let Y = Eϑ,[j] be any ϑ-exceptional divisor for some j ∈ [Υ]. If k = j, we set

mY,T + = mY,T − = 0.
B B

Suppose now k 6= j. We set


mY,T + = 0,
B

mY,T − = mXϑ,u −.
,TB
B j

We call the number mY,T ± the multiplicity of Y associated with the term TB± . We
B

say Y is associated with TB± if mY,T ± is positive. We do not say Y is associated with
B

TB± if the multiplicity mY,T ± is zero.


B

Definition 6.2. Consider any linearized Plücker relation


X
LF = sgn(s)x(us ,vs ) .
s∈SF

for some F ∈ Fm . Fix any s ∈ SF . Consider any ̟-divisor, ̺-divisor, or ϑ-


exceptional divisor Y on Reϑ .
Let Y = Xϑ,w be any ̟ divisor for some w ∈ Id,n . We set mY,s = 0.
Let Y = Xϑ,(u,v) be any ̺ divisor. We set
(
1, if (u, v) = (us , vs )
mY,s =
0, otherwise.
GRASSMANNIANS AND SINGULARITIES 63

Let Y = Eϑ,[k] be any ϑ-exceptional divisor for some k ∈ [Υ]. We let

mY,s = mXϑ,(m,u ,s .
k)

We call the number mY,s the multiplicity of Y associated with s ∈ SF . We say Y


is associated with s if mY,s is positive. We do not say Y is associated with s if the
multiplicity mY,s is zero.

6.2. ℘-centers and ℘-blowups in (℘(kτ ) rµ ).


We proceed by applying induction on the set

{(kτ )µ | k ∈ [Υ], τ ∈ [tFk ], µ ∈ [ρ(kτ ) ]},

ordered lexicographically on (k, τ, µ), where ρ(kτ ) is a to-be-defined finite positive


integer depending on (kτ ) ∈ IndexBmn (cf. (5.11)).
The initial case is ℘(11) r0 and the initial scheme is R(℘(11) r0 ) := Reϑ .
We suppose that the following package in (℘(kτ ) rµ−1 ) has been introduced for some
integer µ ∈ [ρ(kτ ) ], where 1 ≤ ρ(kτ ) ≤ ∞ is an integer depending on (kτ ) ∈ IndexBmn .
(It will be proved to be finite.) Here, to reconcile notations, we make the convention:

(℘(kτ ) r0 ) := (℘(k(τ −1)) rρ(k(τ −1)) ), ∀ 1 ≤ k ≤ Υ, 2 ≤ τ ≤ tFk ,

(℘(k1) r0 ) := (℘((k−1)tFk−1 ) rρ((k−1)tF )


), ∀ 2 ≤ k ≤ Υ,
k−1

provided that ρ(k(τ −1)) and ρ((k−1)tFk−1 ) are (proved to be) finite.

• The inductive assumption. The scheme Re(℘(kτ ) rµ−1 ) has been constructed; it comes
equipped with the set of ̟-divisors,

D(℘(kτ ) rµ−1 ),̟ : X(℘(kτ ) rµ−1 ),w , w ∈ Id,n \m,

the set of ̺-divisors,

D(℘(kτ ) rµ−1 ),̺ : X(℘(kτ ) rµ−1 ),(u,v) , (u, v) ∈ ΛFm ,

and the set of the exceptional divisors,

E(℘(kτ ) rµ−1 ) : E(℘(kτ ) rµ−1 ),(k′ τ ′ )µ′ h′ , (11)0 ≤ (k ′ τ ′ )µ′ ≤ (kτ )(µ − 1), h′ ∈ [σ(k′ τ ′ )µ′ ]

for some finite positive integer σ(k′ τ ′ )µ′ depending on (k ′ τ ′ )µ′ .


64 YI HU

We let
D(℘(kτ ) rµ−1 ) = D(℘(kτ ) rµ−1 ),̟ ⊔ D(℘(kτ ) rµ−1 ),̺ ⊔ E(℘(kτ ) rµ−1 )
be the set of all the aforelisted divisors.
Fix any Y ∈ D(℘(kτ ) rµ−1 ) . Consider any B ∈ Bq ∪ Bmn and let TB be any fixed
term of B. Then, we have that Y is associated with TB with the multiplicity mY,TB ,
a nonnegative integer. In what follows, we say Y is associated with TB if mY,TB > 0;
we do not say Y is associated with TB if mY,TB = 0.
P
Likewise, for any term of Ts = sgn(s)x(us ,vs ) of LF = s∈SF sgn(s)x(us ,vs ) , Y is
associated with Ts with the multiplicity mY,s , a nonnegative integer. We say Y is
associated with Ts if mY,s > 0; we do not say Y is associated with Ts if mY,s = 0.

We are now to construct the scheme Re(℘(kτ ) rµ ) . The process consists of a finite
steps of blowups; the scheme Re(℘ r ) is the one obtained in the final step.
(kτ ) µ

As before, fix any k ∈ [Υ], we write BFmn


k
= {B(kτ ) | τ ∈ [tFk ]}. For every B(kτ ) ∈
BFmn
k
, we have the expression
+ −
B(kτ ) = T(kτ ) − T(kτ ) = x(us ,vs ) xuk − xus xvs x(m,uk )

where s ∈ SFk \sFk corresponds to τ and xuk is the leading variable of F̄k for some
u ∈ Im .
k d,n

Definition 6.3. A pre-℘-set φ in (℘(kτ ) rµ ), written as

φ = {Y + , Y − },

consists of exactly two divisors of the scheme Re(℘(kτ ) rµ−1 ) such that

Y + 6= X(℘(kτ ) rµ−1 ),(us ,vs ) , Y − 6= X(℘(kτ ) rµ−1 ),(m,uk ) ,


±
and Y ± is associated with T(kτ ).

Given the above pre-℘-set φ, we let

Zφ = Y + ∩ Y −

be the scheme-theoretic intersection. The pre-℘-set φ (resp. Zφ ) is called a ℘-set


(resp. ℘-center) in (℘(kτ ) rµ ) if

Zφ ∩ Ve(℘(kτ ) rµ−1 ) 6= ∅.
GRASSMANNIANS AND SINGULARITIES 65

As there are only finitely many ̟-, ̺-, and exceptional divisors on the scheme
Re(℘(kτ ) rµ−1 ) , that is, the set D(℘(kτ ) rµ−1 ) is finite, one sees that there are only finitely
many ℘-sets in (℘(kτ ) rµ ). We let Φ℘(kτ ) rµ be the finite set of all ℘-sets in (℘(kτ ) rµ );
we let Z℘(kτ ) rµ be the finite set of all corresponding ℘-centers in (℘(kτ ) rµ ). We need
a total ordering on the set Φ℘(kτ ) rµ to produce a canonical sequence of blowups.

Definition 6.4. Consider the set D(℘(kτ ) rµ−1 ) = D(℘(kτ ) rµ−1 ),̟ ⊔D(℘(kτ ) rµ−1 ),̺ ⊔E(℘(kτ ) rµ−1 ) .
We introduce a total order “ <♭ ” on the set as follows. Consider any two distinct
elements, Y1 , Y2 ∈ D(℘(kτ ) rµ−1 ) .

(1) Suppose Y1 ∈ E(℘(kτ ) rµ−1 ) and Y2 ∈


/ E(℘(kτ ) rµ−1 ) . Then, Y1 <♭ Y2 .
(2) Suppose Y1 ∈ D(℘(kτ ) rµ−1 ),̺ and Y2 ∈ D(℘(kτ ) rµ−1 ),̟ . Then, Y1 <♭ Y2 .
(3) Suppose Y1 , Y2 ∈ E(℘(kτ ) rµ−1 ) . Then, we can write Yi = E(℘(kτ ) rµ−1 ),(ki τi )µi hi for
some (ki τi )µi ≤ (kτ )(µ − 1) and some 1 ≤ hi ≤ σ(ki τi )µi , with i = 1, 2. Then,
we say Y1 <♭ Y2 if (k1 , τ1 , µ1 , h1 ) <invlex (k2 , τ2 , µ2 , h2 ).
(4) Suppose Y1 , Y2 ∈ D(℘(kτ ) rµ−1 ),̟ . Write Y1 = X(℘(kτ ) rµ−1 ),u and Y2 = X(℘(kτ ) rµ−1 ),v .
Then, Y1 <♭ Y2 if u <lex v.
(5) Suppose Y1 , Y2 ∈ D(℘(kτ ) rµ−1 ),̺ . Write Yi = X(℘(kτ ) rµ−1 ),(ui ,vi ) with i = 1, 2.
Then, Y1 <♭ Y2 if (u1 , v1 ) <lex (u2 , v2 ).

Using the order “ <♭ ” of Definition 6.4, and applying Definition 3.12, one sees
that the set Φ℘(kτ ) rµ now comes equipped with the total order “ <:=<♭,lex ” which
is the lexicographic order induced by “ <=<♭ ”. Thus, we can list Φ℘(kτ ) rµ as

Φ℘(kτ ) rµ = {φ(kτ )µ1 < · · · < φ(kτ )µσ(kτ )µ }

for some finite positive integer σ(kτ )µ depending on (kτ )µ. We then let the set
Z℘(kτ ) rµ of the corresponding ℘-centers inherit the total order from that of Φ℘(kτ ) rµ .
Then, we can express

Z℘(kτ ) rµ = {Zφ(kτ )µ1 < · · · < Zφ(kτ )µσ }.


(kτ )µ

We let Re(℘(kτ ) rµ s1 ) −→ Re(℘(kτ ) rµ−1 ) be the blowup of Re(℘(kτ ) rµ−1 ) along the ℘-center
Zφ(kτ )µ1 . Inductively, we assume that Re(℘(kτ ) rµ s(h−1) ) has been constructed for some
66 YI HU

h ∈ [σ(kτ )µ ]. We then let

Re(℘(kτ ) rµ sh ) −→ Re(℘(kτ ) rµ sh−1 )

be the blowup of Re(℘(kτ ) rµ sh−1 ) along (the proper transform of) the ℘-center Zφ(kτ )µh .
Here, to reconcile notation, we set

Re(℘(kτ ) rµ s0 ) := Re(℘(kτ ) rµ−1 ) := Re(℘(kτ ) rµ−1 sσ(kτ )(µ−1) ) .

All of these can be summarized as the sequence

Re(℘(kτ ) rµ ) := Re(℘(kτ ) rµ sσ(kτ )µ ) −→ · · · −→ Re(℘(kτ ) rµ s1 ) −→ Re(℘(kτ ) rµ−1 ) .

Given h ∈ [σ(kτ )µ ], consider the induced morphism Re(℘(kτ ) rµ sh ) −→ Re(℘(kτ ) rµ−1 ) .


• We let X(℘ r s ),w be the proper transform of X(℘ r ),w in Re(℘ r s ) , for
(kτ ) µ h (kτ ) µ−1 (kτ ) µ h

all w ∈ Id,n \m. These are still called ̟-divisors. We denote the set of all ̟-divisors
on Re(℘ r s ) by D(℘ r s ),̟ .
(kτ ) µ h (kτ ) µ h

• We let X(℘(kτ ) rµ sh ),(u,v) be the proper transform of the ̺-divisor X(℘(kτ ) rµ−1 ),(u,v)
in Re(℘(kτ ) rµ sh ) , for all (u, v) ∈ ΛFm . These are still called ̺-divisors. We denote the
set of all ̺-divisors on Re(℘(kτ ) rµ sh ) by D(℘(kτ ) rµ sh ),̺ .
• We let E(℘ r s ),(k′ τ ′ )µ′ h′ be the proper transform of E(℘ r ),(k′ τ ′ )µ′ h′ in Re(℘ r
(kτ ) µ h (kτ ) µ−1 (kτ ) µ sh )
,
′ ′ ′ ′ ′ ′ ′
for all k ∈ [Υ], τ ∈ [tFk′ ], 0 ≤ µ ≤ ρ(k′ τ ′ ) and h ∈ [σ(k′ τ ′ )µ′ ] with (k τ )µ ≤
(kτ )(µ − 1). (Here, we allow (k ′ τ ′ )µ′ = (11)0. Recall that E(℘(kτ ) rµ−1 ),(11)0h with
h ∈ [σ(11)0 ] = [Υ] correspond to ϑ-divisors.). These are named as ℘-exceptional
or simply exceptional divisors. We denote the set of these exceptional divisors on
Re(℘ r s ) by E(℘ r s ),old .
(kτ ) µ h (kτ ) µ h

We let
D̄(℘(kτ ) rµ sh ) = D(℘(kτ ) rµ sh ),̟ ⊔ D(℘(kτ ) rµ sh ),̺ ⊔ E(℘(kτ ) rµ sh ),old

be the set of all of the aforementioned divisors on Re(℘(kτ ) rµ sh ) .


As each divisor in D̄(℘(kτ ) rµ sh ) is the proper transform of a unique divisor in
D(℘(kτ ) rµ−1 ) . The total order on D(℘(kτ ) rµ−1 ) carries over to provide an induced to-
tal order on the set D̄(℘(kτ ) rµ sh ) .
In addition to the proper transforms of the divisors from Re(℘(kτ ) rµ−1 ) , there are
the following new exceptional divisors.
GRASSMANNIANS AND SINGULARITIES 67

For any h ∈ [σ(kτ )µ ], we let E(℘(kτ ) rµ sh ) be the exceptional divisor of the blowup
Re(℘(kτ ) rµ sh ) −→ Re(℘(kτ ) rµ sh−1 ) . For any 1 ≤ h′ < h ≤ σ(kτ )µ , we let E(℘(kτ ) rµ sh ),(kτ )µh′
be the proper transform in Re(℘ r s ) of the exceptional divisor E(℘ r s ′ ) . To
(kτ ) µ h (kτ ) µ h

reconcile notation, we also set E(℘(kτ ) rµ sh ),(kτ )µh := E(℘(kτ ) rµ sh ) . We set

E(℘(kτ ) rµ sh ),new = {E(℘(kτ ) rµ sh ),(kτ )µh′ | 1 ≤ h′ ≤ h ≤ σ(kτ )µ }.

We then order the exceptional divisors of E(℘(kτ ) rµ sh ),new in the reverse order of oc-
currence, that is, E(℘(kτ ) rµ sh ),(kτ )µh′′ ≤ E(℘(kτ ) rµ sh ),(kτ )µh′ if h′′ ≥ h′ .
We then let
D(℘(kτ ) rµ sh ) = D̄(℘(kτ ) rµ sh ) ⊔ E(℘(kτ ) rµ sh ),new .
For any Y1 ∈ E(℘(kτ ) rµ sh ),new and any Y2 ∈ D̄(℘(kτ ) )rµ sh , we say Y1 <♭ Y2 . As each of
D̄(℘(kτ ) rµ sh ) and E(℘(kτ ) rµ sh ),new is a totally ordered set, this endows D(℘(kτ ) rµ sh ) with a
total order “ <♭ ”.
Finally, we set Re(℘(kτ ) )rµ := Re(℘(kτ ) rµ sσ(kτ )µ ) , and let

D(℘(kτ ) rµ ),̟ = D(℘(kτ ) rµ sσ(kτ )µ ),̟ , D(℘(kτ ) rµ ),̺ = D(℘(kτ ) rµ sσ(kτ )µ ),̺ ,

E℘(kτ ) rµ = E(℘(kτ ) rµ sσ(kτ )µ ),old ⊔ E(℘(kτ ) rµ sσ(kτ )µ ),new .


This can be summarized as

D(℘(kτ ) rµ ) := D(℘(kτ ) rµ ),̟ ⊔ D(℘(kτ ) rµ ),̺ ⊔ E(℘(kτ ) rµ ) .

This way, we have equipped the scheme Re(℘(kτ ) rµ ) with the set D(℘(kτ ) rµ ),̟ of ̟-
divisors, the set D(℘(kτ ) rµ ),̺ of ̟-divisors, and the set E℘(kτ ) rµ of exceptional divisors.
Now, we are ready to introduce the notion of “association” in (℘(kτ ) rµ ), as required
to carry on the process of induction.
We do it inductively on the set [σ(kτ )µ ].

Definition 6.5. Fix any B ∈ Bq ∪ Bmn . We let TB be any fixed term of the bino-
mial B. Meanwhile, we also consider any F̄ ∈ Fm and let Ts be the term of LF
corresponding to any fixed s ∈ SF .
We assume that the notion of “association” in (℘(kτ ) rµ sh−1 ) has been introduced.
That is, for every divisor Y ′ ∈ D(℘(kτ ) rµ sh−1 ) , the multiplicities mY ′ ,TB and mY ′ ,s have
been defined.
68 YI HU

Consider an arbitrary divisor Y ∈ D(℘(kτ ) rµ sh ) .


First, suppose Y 6= E(℘(kτ ) rµ sh ) . Then, it is the proper transform of a (unique)
divisor Y ′ ∈ D(℘(kτ ) rµ sh−1 ) . In this case, we set

mY,TB = mY ′ ,TB , mY,s = mY ′ ,s .

Next, we consider the exceptional Y = E(℘(kτ ) rµ sh ) .


We let φ = φ(kτ )µh . We have that

φ = {Y + , Y − } ⊂ D(℘(kτ ) rµ−1 ) .

For any B ∈ Bmn ∪ Bq , we write B = TB0 − TB1 . We let

mφ,TBi = mY + ,TBi + mY − ,TBi , i = 0, 1,

lφ,B = min{mφ,TB0 , mφ,TB1 }.


(For instance, by definition, lφ,B > 0 when B = B(kτ ) . In general, it can be zero.)
Then, we let
mE(℘ i
,TB = mφ,TBi − lφ,TBi .
(kτ ) rµ sh )

Likewise, for s ∈ SF with F ∈ Fm , we let

mE(℘(kτ ) rµ sh ) ,s = mY + ,s + mY − ,s .

We say Y is associated with TB (resp. Ts ) if its multiplicity mY,TB (resp. mY,s ) is


positive. We do not say Y is associated with TB (resp. Ts ) if its multiplicity mY,TB
(resp. mY,s ) equals to zero.

When h = σ(kτ )µ , we obtain all the desired data on Re(℘(kτ ) rµ ) = Re(℘(kτ ) rµ )sσ(kτ )µ .
Now, with all the aforedescribed data equipped for Re(℘ r ) , we obtain our in- (kτ ) µ

ductive package in (℘(kτ ) rµ ). This allows us to introduce the ℘-sets Φ℘(kτ ) rµ+1 and
℘-centers Z℘(kτ ) rµ+1 in (℘(kτ ) rµ+1 ) as in Definition 6.3, endow total orders on Φ℘(kτ ) rµ+1
and Z℘(kτ ) rµ+1 as in the paragraph immediately following Definition 6.4, and then
advance to the next round of the ℘-blowups. Here, to reconcile notations, we set

(℘(kτ ) rρ(kτ )+1 ) := (℘((k(τ +1)) r1 ), 1 ≤ τ < tFk ;

(℘(ktFk ) rρ(ktF )+1


) := (℘((k+1)1) r1 ), 1 ≤ k < Υ,
k
GRASSMANNIANS AND SINGULARITIES 69

provided that ρ(kτ ) and ρ(ktFk ) are (proved to be) finite.


Given any (℘(kτ ) rµ sh ), the ℘-blowup in (℘(kτ ) rµ sh ) gives rise to

Ve(℘(kτ ) rµ sh )  Re(℘(kτ ) rµ sh )

(6.1) /

 

V / RF ,

where Ve(℘(kτ ) rµ sh ) is the proper transform of V in Re(℘(kτ ) rµ sh ) .


We let Ve(℘(kτ ) rµ ) = Ve(℘(kτ ) rµ sσ(kτ )µ ) .

Definition 6.6. Fix any k ∈ [Υ], τ ∈ [tFk ]. Suppose there exists a finite integer µ
such that for any pre-℘-set φ in (℘(kτ ) rµ+1 ) (cf. Definition 6.3), we have

Zφ ∩ Ve(℘(kτ ) rµ ) = ∅.

Then, we let ρ(kτ ) be the smallest integer such that the above holds. Otherwise, we
let ρ(kτ ) = ∞.

It will be shown soon that ρ(kτ ) is finite for all k ∈ [Υ], τ ∈ [tFk ].
For later use, we let

(6.2) Φ = {φ(kτ )µh | k ∈ [Υ], τ ∈ [tFk ], 1 ≤ µ ≤ ρ(kτ ) , h ∈ [σ(kτ )µ ]},

IndexΦ = {(kτ )µh | k ∈ [Υ], τ ∈ [tFk ], 1 ≤ µ ≤ ρ(kτ ) , h ∈ [σ(kτ )µ ]}.

Upon proving that ρ(kτ ) is finite for all (kτ ) ∈ IndexBmn , we can summarize the
process of ℘-blowups as a single sequence of blowup morphisms:

(6.3) Re℘ −→ · · · −→ Re(℘(kτ ) rµ sh ) −→ Re(℘(kτ ) rµ sh−1 ) −→ · · · −→ Re(℘(11) r0 ) := Reϑ ,

where Re℘ := Re(℘(ΥtΥ ) rρΥt ) := Re(℘(ΥtΥ ) rρΥt sσ(Υt ) is the blowup scheme reached
Υ Υ Υ )ρΥtΥ

in the final step (℘(ΥtΥ ) rρΥtΥ sσ(ΥtΥ )ρΥt ) of all ℘-blowups.


Υ

Further, the end of all ℘-blowups gives rise to the following induced diagram

Ve℘  Re℘

(6.4) /

 

V  / RF ,
70 YI HU

where Ve℘ is the proper transform of V in Re℘ .

6.3. Properties of ℘-blowups.


Recall that we have set Re(℘ r (11) 1 s0 )
:= Reϑ := Reϑ[Υ] .
Now, fix and consider any (kτ )µh ∈ IndexΦ (cf. (6.2)).

Proposition 6.7. Suppose that the scheme Re(℘(kτ ) rµ sh ) has been constructed, covered
by a finite set of open subsets, called standard charts (see Definition 5.1).
Consider any standard chart V of Re(℘(kτ ) rµ sh ) , lying over a unique chart V[0] of
RF . We suppose that the chart V[0] is indexed by ΛoFm = {(vsF,o , vsF,o ) | F̄ ∈ Fm }.
As earlier, we have Λ⋆Fm = ΛFm \ΛoFm .
Then, the chart V comes equipped with

a subset eV ⊂ Id,n \m and a subset dV ⊂ Λ⋆Fm

such that every exceptional divisor E (i.e., not a ̟- nor a ̺-divisor) of Re(℘(kτ ) rµ sh )
with E ∩ V 6= ∅ is either labeled by a unique element w ∈ eV or labeled by a unique
element (u, v) ∈ dV . We let E(℘(kτ ) rµ sh ),w be the unique exceptional divisor on the
chart V labeled by w ∈ eV ; we call it an ̟-exceptional divisor. We let E(℘(kτ ) rµ sh ),(u,v)
be the unique exceptional divisor on the chart V labeled by (u, v) ∈ dV ; we call it
an ̺-exceptional divisor. (We note here that being ̟-exceptional or ̺-exceptional is
strictly relative to the given standard chart.)
Further, the chart V comes equipped with a set of free variables
 
 ε , δ w ∈ e , (u, v) ∈ d 
V,w V,(u,v) V V
(6.5) VarV := ,
 xV,w , xV,(u,v) w ∈ Id,n \m\eV , (u, v) ∈ Λ⋆F \dV 
m
such that it is canonically isomorphic to the affine space with the variables in (6.5)
as its coordinate variables. Moreover, on the standard chart V, we have

(1) the divisor X(℘(kτ ) rµ sh ),w ∩V is defined by (xV,w = 0) for every w ∈ Id,n \m\eV ;
(2) the divisor X(℘(kτ ) rµ sh ),(u,v) ∩ V is defined by (xV,(u,v) = 0) for every (u, v) ∈
Λ⋆Fm \dV ;
(3) the divisor X(℘(kτ ) rµ sh ),w does not intersect the chart for all w ∈ eV ;
(4) the divisor X(℘(kτ ) rµ sh ),(u,v) does not intersect the chart for all (u, v) ∈ dV ;
GRASSMANNIANS AND SINGULARITIES 71

(5) the ̟-exceptional divisor E(℘(kτ ) rµ sh ),w ∩ V labeled by an element w ∈ eV is


define by (εV,w = 0) for all w ∈ eV ;
(6) the ̺-exceptional divisor E(℘(kτ ) rµ sh ),(u,v) ∩V labeled by an element (u, v) ∈ dV
is define by (δV,(u,v) = 0) for all (u, v) ∈ dV ;
(7) any of the remaining exceptional divisor of Re(℘(kτ ) rµ sh ) other than those that
are labelled by some some w ∈ eV or (u, v) ∈ dV does not intersect the chart.

Proof. We prove by induction on (kτ )µh ∈ {(11)10} ⊔ IndexΦ .


For the initial case, the scheme is Re(℘ r s ) = Reϑ = Reϑ . Then, this proposition
(11) 1 0 [Υ]

is the same as Proposition 5.9 with k = Υ. Thus, it holds.


We suppose that the statement holds over Re(℘(kτ ) rµ sh−1 ) with (kτ )µh ∈ IndexΦ .
We now consider Re(℘ r s ) . We have the embedding
(kτ ) µ h

Re(℘(kτ ) rµ sh )  Re(℘(kτ ) rµ sh−1 ) × Pφ(kτ )µh ,


 /

where Pφ(kτ )µh is the factor projective space. We let φ′(kτ )µh = {Y0′ , Y1′ } be the proper
transforms in Re(℘(kτ ) rµ sh−1 ) of the two divisors of the ℘-set φ(kτ )µh = {Y + , Y − } with
±
Y ± being associated with T(kτ ) . In addition, we let [ξ0 , ξ1 ] be the homogenous
coordinates of Pφ(kτ )µh corresponding to {Y0′ , Y1′ }.
Then, the given standard chart V of Re(℘(kτ ) rµ sh ) lies over a unique standard chart
V′ of Re(℘(kτ ) rµ sh−1 ) such that V = (V′ × (ξi ≡ 1)) ∩ Re(℘(kτ ) rµ sh ) , for i = 0 or 1.
By assumption, the chart V′ comes equipped with a subset eV′ ⊂ Id,n \m, a subset
dV′ ⊂ Λ⋆Fm , and admits a set of coordinate variables
 
 ε ′ , δ ′ w ∈ eV , (u, v) ∈ dV 
V ,w V ,(u,v) ′ ′
(6.6) VarV′ := ,
 xV′ ,w , xV′ ,(u,v) w ∈ Id,n \m\eV′ , (u, v) ∈ Λ⋆F \dV′ 
m
verifying the properties (1)-(7) as in the proposition.
We now prove the statements for the chart V of Re(℘ .
(kτ ) rµ sh )

First, we suppose that the proper transform Zψ′ (kτ )µh in Re(℘(kτ ) rµ sh−1 ) of the ℘-
center Zψ(kτ )µh does not meet the chart V′ . Then, we let V inherit all the data from
those of V′ , that is, we set eV = eV′ , dV = dV′ , and VarV = VarV′ : changing the
subindex “ V′ ” for all the variables in VarV′ to “ V ”. As the ℘-blowup along
72 YI HU

the proper transform of Z(℘(kτ ) rµ sh ) does not affect the chart V′ , one sees that the
statements of the proposition hold for V.
Next, we suppose that Zψ′ (kτ )µh meets the chart V′ along a nonempty closed subset.
On the chart V′ , by the inductive assumption, we can suppose

(6.7) Y0′ ∩ V′ = (y0′ = 0), Y1′ ∩ V′ = (y1′ = 0), for some y0′ , y1′ ∈ VarV′ .

Then, the chart V = (V′ × (ξi ≡ 1)) ∩ Re(℘(kτ ) rµ sh ) of the scheme Re(℘(kτ ) rµ sh ) , as a
closed subscheme of V′ × (ξi ≡ 1), is defined by

(6.8) yj′ = yi′ ξj , with j ∈ {0, 1}\i.

There are three possibilities for Yi′ ∩ V′ according to the assumption on the chart
V′ . Based on every of such possibilities, we set

 ′
 eV = eV′ ⊔ w, dV = dV′ , if yi = xV′ ,w for some w ∈ Id,n \m\eV′

(6.9) eV = eV′ , dV = dV′ , if yi′ = εV′ ,w for some w ∈ eV′


 d = d ′, e = e ′, if yi′ = δV′ ,(u,v) for some (u, v) ∈ dV′ .
V V V V

Accordingly, we introduce


 εV,w = yi′ , xV,w = 1, if yi′ = xV′ ,w for some w ∈ eV \eV′

(6.10) εV,w = yi′ , if yi′ = εV′ ,w for some w ∈ eV′ = eV


 δV,(u,v) = yi′ , if yi′ = δV′ ,(u,v) for some (u, v) ∈ dV′ = dV .

To introduce the set VarV , for j ∈ {0, 1}\i, we then set



 if yj′ = xV′ ,a
 xV,a = ξj ,

(6.11) εV,a = ξj , if yj′ = εV′ ,a


 δ = ξj , if yj′ = δV′ ,(a,b) .
V,(a,b)

Thus, we have introduced yi′ , ξj ∈ VarV where yi′ (respectively, ξj ) is endowed with its
new name as in (6.10), respectively, in (6.11). Next, we define the set VarV \{yi′ , ξj }
to consist of the following variables:


 xV,w = xV′ ,w , ∀ w ∈ Id,n \m\eV and xV′ ,w 6= yj′



 x
V,(u,v) = xV′ ,(u,v) , ∀ (u, v) ∈ Λ⋆Fm \dV
(6.12)

 ∀ w ∈ eV and εV′ ,w 6= yj′ , yi′
 εV,w = εV′ ,w ,


 δ
V,(u,v) = δV′ ,(u,v) , ∀ (u, v) ∈ dV and δV′ ,(u,v) 6= yj′ , yi′ .
GRASSMANNIANS AND SINGULARITIES 73

Substituting (6.8), one checks that the chart V is isomorphic to the affine space
with the variables in (6.10), (6.11), and (6.12), as its coordinate variables (cf. Propo-
sition 5.3). Putting all together, the above matches description of VarV in (6.5),
Now, it remains to verity (1)-(7) of the proposition on the chart V.
First, consider the unique new exceptional divisor E(℘(kτ ) rµ sh ) created by the
blowup
Re(℘(kτ ) rµ sh ) −→ Re(℘(kτ ) rµ sh−1 ) .

Then, we have
E(℘(kτ ) rµ sh ) ∩ V = (yi′ = 0)

where yi′ is renamed as in (6.10) according to the three possibilities of its form in
VarV′ . This way, the new exceptional divisor E(℘(kτ ) rµ sh ) is labelled on the chart V.
In any case of the three situations, we have that the proper transform in Re(℘ r s ) (kτ ) µ h

of Yi′ does not meet the chart V, and if Yi′ is a ̟-exceptional divisor labeled by
some element of eV′ (resp, ̺-exceptional divisor labeled by some element of dV′ ),
then, on the chart V, its proper transform is no longer labelled by any element of
Id,n \m (resp. Λ⋆Fm ). This verifies any of (3)-(7) whenever the statement therein
involves the newly created exceptional divisor E(℘(kτ ) rµ sh ) and/or its corresponding
exceptional variable yi′ (which is renamed in (6.10) to match that of (6.5) in the
statement of the proposition). Observe that the statements (1) and (2) are not
related to the new exceptional divisor E(℘(kτ ) rµ sh ) or its corresponding exceptional
variable yi′ (again, renamed in (6.10)).
For any of the remaining ̟-, ̺-, and exceptional divisors on Re(℘(kτ ) rµ sh ) , it is
the proper transform of a unique corresponding ̟-, ̺-, and exceptional divisor on
Re(℘(kτ ) rµ sh−1 ) . Hence, applying the inductive assumption on V′ , one verifies directly
that every of the properties (1)-(7) of the proposition is satisfied whenever it applies
to such a divisor and/or its corresponding local variable in VarV .
This completes the proof. 

6.4. Proper transforms of defining equations in (℘(kτ ) rµ sh ).


Consider any fixed B ∈ Bmn ∪ Bq and F̄ ∈ Fm . Suppose BV′ and LV′ ,F have
been constructed over V′ . Applying Definition 5.4, we obtain the proper transforms
74 YI HU

on the chart V
BV , B ∈ Bmn ∪ Bq ; LV,F , F̄ ∈ Fm .

Definition 6.8. (cf. Definition 5.6) Consider any main binomial relation B ∈ Bmn .
We say B terminates on a given standard chart V of the scheme Re(℘ r s ) if BV (kτ ) µ h

terminates at all closed points of Ve(℘(kτ ) rµ sh ) ∩V. We say B terminates on Re(℘(kτ ) rµ sh )


if it terminates on all standard charts V of Re(℘ r s ) . (kτ ) µ h

In what follows, for any B = TB+ − TB− ∈ Bmn , we express BV = TV,B


+ −
− TV,B . If
+ −
B = B(kτ ) for some k ∈ [Υ] and τ ∈ [tFk ], we also write BV = TV,(kτ ) − TV,(kτ ) .

Below, we follow the notations of Proposition 6.7 as well as those in its proof.
In particular, we have that V is a standard chart of Re(℘ r s ) , lying over a stan-
(kτ ) µ h

dard chart V of Re(℘(kτ ) rµ sh−1 ) . We have that φ′(kτ )µh = {Y0′ , Y1′ } is the proper trans-

forms of φ(kτ )µh = {Y + , Y − } in Re(℘ r s ) with Y ± being associated with T ± .


(kτ ) µ h−1 (kτ )
Likewise, Zφ′ (kτ )µh is the proper transforms of the ℘-center Zφ(kτ )µh in Re(℘(kτ ) rµ sh−1 ) .
Also, assuming that Zφ′ (kτ )µh ∩ V′ 6= ∅, then, as in (6.7), we have

Y0′ ∩ V′ = (y0′ = 0), Y1′ ∩ V′ = (y1′ = 0), with y0′ , y1′ ∈ VarV′

Further, we have Pφ(kτ )µh = P[ξ0 ,ξ1 ] with the homogeneous coordinates [ξ0 , ξ1 ] corre-
sponding to (y0′ , y1′ ).

Proposition 6.9. Let the notation be as in Proposition 6.7 and be as in above.


Let V be any standard chart of Re(℘(kτ ) rµ sh ) . Then, the scheme V(℘(kτ ) rµ sh ) ∩ V, as
a closed subscheme of the chart V is defined by
mn q
BV , BV , LV,Fm .

Suppose that Zφ′ (kτ )µh ∩V′ 6= ∅ where Zφ′ (kτ )µh is the proper transform of the ℘-center
Zφ(kτ )µh in Re(℘(kτ ) rµ sh−1 ) . Further, we let ζ = ζV,(kτ )µh be the exceptional parameter
in VarV such that
E(℘(kτ ) rµ sh ) ∩ V = (ζ = 0).

Then, we have that the following hold.


(1) Suppose V = (V′ × (ξ0 ≡ 1)) ∩ Re(ð(kτ ) rµ sh ) . We let y1 ∈ VarV be the proper
transform of y1′ . Then, we have
GRASSMANNIANS AND SINGULARITIES 75

+ + +
(1a) The plus-term TV,(kτ ) is square-free and deg(TV,(kτ ) ) = deg(TV′ ,(kτ ) ) − 1.
We let degy1′ TV−′ ,(kτ ) = b for some integer b, positive by definition, then
− −
we have degζ TV,(kτ ) = b − 1. Consequently, either TV,(kτ ) is linear in y1

or else ζ | TV,(kτ ).
+
(1b) Let B ∈ BFmn
k
with B > B(kτ ) . Then, TV,B is square-free. Suppose
− − −
y1 | TV,B , then either TV,B is linear in y1 or ζ | TV,B .
+
(1c) Let B ∈ BFmn
k
with B < B(kτ ) or B ∈ Bmn \BFmn
k
. Then, TV,B is square-
+ − −
free. y1 ∤ TV,B . If y1 | TV,B , then ζ | TV,B .
(1d) Consider any B ∈ Bq . Then, BV is square-free.
(2) Suppose V = (V′ × (ξ1 ≡ 1)) ∩ Re(ð(kτ ) rµ sh ) . We let y0 ∈ VarV be the proper
transform of y0′ . Then, we have
+ − − −
(2a) TV,(kτ ) is square-free. y0 ∤ TV,(kτ ) . deg(TV,(kτ ) ) = deg(TV′ ,(kτ ) ) − 1. We
let degy1′ TV−′ ,(kτ ) = b for some integer b, positive by definition, then we
− ′
have degζ TV,(kτ ) = b − 1. Note here that y1 ∈ VarV becomes ζ ∈ VarV .

+
(2b) Let B ∈ BFmn
k
with B > B(kτ ) . Then, TV,B is square-free. Further,

y0 ∤ TV,B .
+
(2c) Let B ∈ BFmn
k
with B < B(kτ ) or B ∈ Bmn \BFmn
k
. Then, TV,B is square-
+ − −
free. y0 ∤ TV,B . If y0 | TV,B , then ζ | TV,B .
(2d) Consider any B ∈ Bq . Then, BV is square-free.
Consequently, we have
(3) ρ(kτ ) < ∞.
(4) Moreover, consider Re(℘(kτ ) rρ(kτ ) ) = Re(℘(kτ ) rρ(kτ ) sσ(kτ )ρ ). Let V be an ar-
(kτ )

bitrary standard chart of Re(℘(kτ ) rρ(kτ ) ) . For every B = B(k′ τ ′ ) ∈ Bmn for
some (k ′ τ ′ ) ∈ IndexBmn with (k ′ τ ′ ) ≤ (kτ ), we write B(k′ τ ′ ) = TB+ − TB− =
x(ut ,vt ) xuk′ − xut xvt x(m,uk′ ) where t ∈ SFk′ \sFk′ corresponds to τ ′ and xuk′
m +
is the leading variable of F̄k′ for some uk′ ∈ Id,n . Then, xV,(ut ,vt ) | TV,B and
TV,B /xV,(ut ,vt ) is invertible along V ∩ Ve(℘(kτ ) rρ(kτ ) ) . (Recall the convention:
+

xV,(u,v) = 1 if (u, v) ∈ dV .)

Proof. We continue to follow the notation in the proof of Proposition 6.7.


We prove the proposition by applying induction on (kτ )µh ∈ ((11)10) ⊔ IndexΦ .
76 YI HU

The initial case is (11)10 with Re(℘(11) r1 s0 ) = Reϑ . In this case, the statement about
defining equations of Re(ð r s ) ∩ V follows from Proposition 5.12 with k = Υ; the
(11) 1 0

remainder statements (1) - (4) are void.


Assume that the proposition holds for (℘(kτ ) rµ sh−1 ) with (kτ )µh ∈ IndexΦ .
Consider (℘(kτ ) rµ sh ).
Consider any standard chart V of Re(℘(kτ ) rµ sh ) , lying over a standard chart of V′
of Re℘(kτ ) rµ sh−1 ) . By assumption, all the desired statements of the proposition hold
over the chart V′ .
To begin with, we suppose that the proper transform Zφ′ (kτ )µh of the ℘-center
Zφ℘(kτ ) µh in Re(℘(kτ ) rµ sh−1 ) does not meet the chart V′ . Then, by the proof of Propo-
sition 6.7, we have that V retains all the data from those of V′ . As the ℘-blowup
along the proper transform of Zφ(kτ )µh does not affect the chart V′ , we have that the
statement of the proposition about the defining equations follows immediately from
the inductive assumption. The statements (1) and (2) are void.
In what follows, we suppose that the proper transform Zφ′ (kτ )µh meets the chart
V′ along a nonempty closed subset.
The statement of the proposition on the defining equations of Ve(℘(kτ ) rµ sh ) ∩ V
follows straightforwardly from the inductive assumption.
Proof of (1).
(1a). We may express

B(kτ ) = x(usτ ,vsτ ) xuk − xusτ xvsτ x(m,uk )

where xuk is the leading variable of F̄k , and sτ ∈ SFk \sFk corresponds to τ ∈ [tFk ].
Observe that none of ̺-divisors appear in any ℘-set. Observe in addition that xuk
does not appear in any B ∈ BFmn
j
with j < k. Thus, the fact that the plus-term
+
TV,(kτ ) is square-free is immediate if τ = 1. For a general τ ∈ [tFk ], it follows
from the inductive assumption on TV+′ ,(kτ ) . The remainder statements follow from
straightforward calculations. We omit the obvious details.
(1b). We can write B = B(kτ ′ ) with τ ′ ∈ [tFk ] and τ ′ > τ . We may express

B = B(kτ ′ ) = x(us ′ ,vs ′ ) xuk − xus ′ xvs ′ x(m,uk ) .


τ τ τ τ
GRASSMANNIANS AND SINGULARITIES 77

We have TB+ = x(us ′ ,vs ′ ) xuk and it retains this form prior to the ℘-blowups with
τ τ

respect to binomials of BFk . (Recall here the convention: xV,u = 1 if u ∈ eV ;


xV,(u,v) = 1 if (u, v) ∈ dV .) Note that the ̺-variable x(us ′ ,vs ′ ) does not appear in
τ τ

any ϑ-set. Thus, as none of ̺-divisors appear in any ℘-set, we see that all possible
exceptional variables from the earlier blowups appearing in TV+′ ,B are acquired (cf.
Definition 5.4) through the unique ̟-variable xuk of TB+ . Note further that xuk does
+
not appear in any B ∈ BFmn
j
with j < k. From here, one sees directly that TV,(kτ ′)


is square-free. Now, suppose y1 | TV,B . We can assume degy1′ (TV−′ ,B ) = b for some
integer b > 0. Since B ∈ BFk and B > B(kτ ) , then by the above discussion about
− −
TB+ , we must have y0′ | TV,B
+
. Thus, degy1 (TV,B ) = b and degζ (TV,B ) = b − 1. Hence,
− −
in such a case, either TV,B is linear in y1 , when b = 1, or ζ | TV,B when b > 1.
(1c). Consider B ∈ BFmn
k
with B < B(kτ ) or B ∈ Bmn \BFmn
k
. We can write
B = B(k′ τ ′ ) for some (k ′ τ ′ ) 6= (kτ ) and we may suppose τ ′ corresponds to some
index t ∈ SFk′ . We have TB+ = x(ut ,vt ) xuk′ . Assume (k ′ τ ′ ) < (kτ ). Then, by the fact
that the variable x(ut ,vt ) uniquely appears in TB+ among all main binomial equations
of Bmn and Proposition 6.9 (4) in the case of (℘(k′ τ ′ ) rρ(k′ τ ′ ) ) (which holds by the
inductive assumption since (k ′ τ ′ ) < (kτ )), we conclude that neither of the variables
{y0′ , y1′ } appears in TV+′ ,B . Since TV+′ ,B is square-free by the inductive assumption,
this implies all three of the statements of (1c) when (k ′ τ ′ ) < (kτ ). Assume k ′ > k.
We have TB+ = x(ut ,vt ) xuk′ . Because x(ut ,vt ) uniquely appears in TB+ among all main
binomial equations and the leading variable xuk′ of F̄k′ does not appear in any F̄i
with i < k ′ , we again conclude that neither of the variables {y0′ , y1′ } appears in TV+′ ,B .
Hence, again, all three of the statements of (1c) hold as well when k ′ > k.
(1d). Consider any B ∈ Bq and let TB be any fixed term of B. We let Vϑ be
the unique standard chart of Reϑ such that V lies over Vϑ . If TB does not contain
any ̺-variable of the form x(m,u) with u ∈ Id,n m , then we immediately have that

TV,B is square-free because the ̺-variables of TB never appear in any ϑ-set or ℘-set.
m
Now assume that x(m,u) | TB for some u ∈ Id,n . By Lemma 4.6 (2), applied to the
variable pm , such a ̺-variable is unique. By the last statement of Proposition 5.12
with k = Υ about Bq , either δVϑ ,(m,u) | TVϑ ,B or εVϑ ,u xVϑ ,(m,u) | TVϑ ,B . We can
write u = uFj for some j ∈ [Υ]. Suppose j < k. Then, y0′ can not appear in TV′ ,B .
78 YI HU

Suppose j = k. Then, y1′ can not appear in TV′ ,B . Suppose j > k. Then, neither
of {y0′ , y1′ } can appear in TV′ ,B . In any case, one sees that at most one variable of
{y0′ , y1′ } may appear in TV′ ,B . By the inductive assumption, TV′ ,B is square-free.
Hence, TV,B is square-free.
Proof of (2).
+
(2a). The proof of the fact that the plus-term TV,(kτ ) is square-free is totally
analogous to the corresponding part of (1a). The remainder statements follow from
straightforward calculations. We omit the obvious details.
+
(2b). The fact that TV,(kτ ) is square-free, again, follows straightforwardly, and its
proof is analogous to the corresponding proof of (1b). Since B ∈ BFk and B > B(kτ ) ,
+ − −
we must have y0′ | TV,B , thus y0′ ∤ TV,B . Hence, y0 ∤ TV,B .
(2c). This follows from the same line of arguments as in (1c).
(2d). This follows from the same line of arguments as in (1d).
Proof of (3).
First, we observe the following.
+
By the construction of ℘-centers, we always have xV,(usτ ,vsτ ) | TV,(kτ ).

The number of variables on any standard chart is a constant.


+ +
If (1) occurs, we have (1a): deg(TV,(kτ ) ) = deg(TV′ ,(kτ ) ) − 1.
− −
If (2) occurs, we have (2a): deg(TV,(kτ ) ) = deg(TV′ ,(kτ ) ) − 1.

Hence, if (1) keeps occurring, then, as µ increases, after finitely many rounds,
+ e
we will eventually have TV,(kτ ) /xV,(usτ ,vsτ ) is invertible over V ∩ V(℘(kτ ) rµ ) , for all V.
If (2) keeps happening, as µ increases, after finitely many rounds, using Corollary
5.14, we see that T − must terminate at all points of V ∩ Ve(℘ r ) , hence, BV,(kτ )
V,(kτ ) (kτ ) µ

terminates over V, for all V.


The above implies that the process of ℘-blowups in (℘(kτ ) ) must terminate after
finitely many rounds. That is, ρ(kτ ) < ∞.
Proof of (4).
When B = B(k′ τ ′ ) ∈ Bmn with (k ′ τ ′ ) < (kτ ), the statement follows from the corre-
sponding statement on Re℘ ′ ′ r by the inductive assumption in (℘(k′ τ ′ ) rρ ′ ′ ).
(k τ ) ρ(k′ τ ′ ) (k τ )
GRASSMANNIANS AND SINGULARITIES 79

Now consider B = B(kτ ) . As earlier, we express B(kτ ) = x(us ,vs ) xuk −xus xvs x(m,uk )
with s = sτ ∈ SFk \sFk corresponding to τ . Consider any standard chart V of the
final scheme Re℘(kτ ) rρ(kτ ) in (℘(kτ ) ). Suppose that the standard chart V is the result
after (1) occurs. Then, BV must be of the form

xV,(us ,vs ) − TV,(kτ ).

Suppose that V is the result after (2) occurs. Then, BV must be of the form
+
TV,(kτ ) − xV,(m,uk ) .

(Recall here the convention: xV,(u,v) = 1 if (u, v) ∈ dV .) The above implies the
statements, by applying Corollary 5.14 to the last case.
Therefore, by induction on (kτ )µh ∈ {(11)10}⊔IndexΦ , Proposition 6.9 is proved.


7. ð-Blowups

In this section, we finalize the process of “removing” all possible zero factors of all
the main binomial relations, building upon the already performed ϑ- and ℘-blowups.

7.1. The initial setup: (ð(11) r0 ).


Our initial scheme in (ð(11) r0 ) is Re(ð11 r0 ) := Re℘ . Recall that the scheme Re℘ comes
equipped with three kinds of divisors:
• the proper transforms X℘,w of ̟-divisors Xw for all w ∈ Id,n \m;
• the proper transforms X℘,(u,v) of ̺-divisors X(u,v) for all (u, v) ∈ ΛFm ;
• the proper transforms E℘,(11)0h of ϑ-exceptional divisors Eϑ,[h] ⊂ Reϑ with
h ∈ [Υ]; the proper transforms E℘,(kτ )µh of the ℘-exceptional divisors E(kτ )µh ⊂
Re(℘(kτ ) rµ sh ) with (kτ )µh ∈ IndexΦ .
The set of exceptional divisors on R℘ can be written as

E(ð11 r0 ) = {E℘,(kτ )µi | (kτ )µi ∈ {(11)0i | i ∈ [Υ]} ⊔ IndexΦ }.

This is an ordered set with the cardinality ς(11)0 := Υ + |Φ|. Thus, for notational
consistency later on, using the order of the set E(ð11 r0 ) , we can express it as

E(ð11 r0 ) = {E(ð11 r0 ),(11)0h | h ∈ [ς(11)0 ]}.


80 YI HU

7.2. ð-centers and ð-blowups in (ð(kτ ) rµ ).


We suppose that the following package in (ð(kτ ) rµ−1 ) has been introduced for some
integer µ ∈ [κ(kτ ) ], where κ(kτ ) is a to-be-defined finite positive integer depending
on (kτ ) ∈ IndexBmn . Here, to reconcile notations, we make the convention:

(ð(kτ ) r0 ) := (ð(k(τ −1)) rκ(k(τ −1) ), ∀ 1 ≤ k ≤ Υ, 2 ≤ τ ≤ tFk ,

(ð(k1) r0 ) := (ð((k−1)tFk−1 ) rκ((k−1)tF )


), ∀ 2 ≤ k ≤ Υ,
k−1

provided that κ(k(τ −1) and κ((k−1)tFk−1 ) are (proved to be) finite.

• The inductive assumption. The scheme Re(ð(kτ ) rµ−1 ) has been constructed; it comes
equipped with the set of ̟-divisors,

D(ð(kτ ) rµ−1 ),̟ : X(ð(kτ ) rµ−1 ),w , w ∈ Id,n \m,

the set of ̺-divisors,

D(ð(kτ ) rµ−1 ),̺ : X(ð(kτ ) rµ−1 ),(u,v) , (u, v) ∈ ΛFm ,

and the set of the exceptional ̺-divisors,

E(ð(kτ ) rµ−1 ) : E(ð(kτ ) rµ−1 ),(k′ τ ′ )µ′ h′ , (11)0 ≤ (k ′ τ ′ )µ′ ≤ (kτ )(µ − 1), h′ ∈ [ς(k′ τ ′ )µ′ ]

for some finite positive integer ς(k′ τ ′ )µ′ depending on (k ′ τ ′ )µ′ .


We let
D(ð(kτ ) rµ−1 ) = D(ð(kτ ) rµ−1 ),̟ ⊔ D(ð(kτ ) rµ−1 ),̺ ⊔ E(ð(kτ ) rµ−1 )

be the set of all the aforelisted divisors.


Fix any Y ∈ D(ð(kτ ) rµ−1 ) . Consider any B ∈ Bq ∪ Bmn and let TB be any fixed term
of B. Then, we have that Y is associated with TB with the multiplicity mY,TB , a
nonnegative integer. In what follows, we say Y is associated with TB if mY,TB > 0;
we do not say Y is associated with TB if mY,TB = 0.
P
Likewise, for any term of Ts = sgn(s)x(us ,vs ) of LF = s∈SF sgn(s)x(us ,vs ) , Y is
either associated with Ts with a positive multiplicity mY,s or not associated with Ts
in which case we set mY,s = 0.
We are to construct the scheme Re(ð(kτ ) )µ in the next round rµ . This round consists
of finitely many steps and the scheme Re(ð )µ is the one obtained in the final step.
(kτ )
GRASSMANNIANS AND SINGULARITIES 81

As before, fix any k ∈ [Υ], we write BFmn


k
= {B(kτ ) | τ ∈ [tFk ]}. For every main
binomial B(kτ ) of BFmn
k
, we have the expression
+ −
B(kτ ) = T(kτ ) − T(kτ ) = x(us ,vs ) xuk − xus xvs x(m,uk )

where s ∈ SFk \sFk corresponds to τ and xuk is the leading variable of F̄k .

Definition 7.1. Consider the above main binomial relation B(kτ ) ∈ BFmn
k
. A pre-ð-
set in (ð(kτ ) rµ )
ψ = {Y + = X(ð(kτ ) rµ−1 ),(us ,vs ) , Y − }
consists of the ̺-divisor Y + = X(ð(kτ ) rµ−1 ),(us ,vs ) and a ̟- or exceptional divisor
Y − on the scheme Re(ð r ) such that Y − is associated with T − . (Here, Y + is
(kτ ) µ−1 (kτ )
+
automatically associated with T(kτ ). Also observe that X(ð(kτ ) rµ−1 ),(m,uk ) is the only

̺-divisor associated with T(kτ ), but, we do not use it due to Corollary 5.14.) Given
+ −
a pre-ð-set ψ = {Y , Y }, we let Zψ be the scheme-theoretic intersection

Zψ = Y + ∩ Y − .

The pre-ð-set ψ (resp. Zψ ) is called an ð-set (resp. an ð-center) in (ð(kτ ) rµ ) if

Zψ ∩ Ve(ð(kτ ) rµ−1 ) 6= ∅.

As there are only finitely many ̟-, ̺-, and exceptional divisors on the scheme
Re(ð(kτ ) rµ−1 ) , that is, the set D(ð(kτ ) rµ−1 ) is finite, one sees that there are only finitely
many ð-sets in (ð(kτ ) rµ ). We let Ψ(kτ )µ be the finite set of all ð-sets in (ð(kτ ) rµ ). We
let Zð(kτ ) rµ be the finite set of the corresponding ð-centers in (ð(kτ ) rµ ). We need a
total ordering on the finite set Ψ(kτ )µ to proceed.

Definition 7.2. Consider the set D(ð(kτ ) rµ−1 ) = D(ð(kτ ) rµ−1 ),̟ ⊔D(ð(kτ ) rµ−1 ),̺ ⊔E(ð(kτ ) rµ−1 ) .
We introduce a total order “ <♭ ” on the set as follows. Consider any two distinct
elements, Y1 , Y2 ∈ D(ð(kτ ) rµ−1 ) .
(1) Suppose Y1 ∈ E(ð(kτ ) rµ−1 ) and Y2 ∈
/ E(ð(kτ ) rµ−1 ) . Then, Y1 <♭ Y2 .
(2) Suppose Y1 ∈ D(ð(kτ ) rµ−1 ),̺ and Y2 ∈ D(ð(kτ ) rµ−1 ),̟ . Then, Y1 <♭ Y2 .
(3) Suppose Y1 , Y2 ∈ E(ð(kτ ) rµ−1 ) . Then, we can write Yi = E(ð(kτ ) rµ−1 ),(ki′ τj )µj h′i for
some (kj τj )µj ≤ (kτ )(µ −1) and some 1 ≤ h′i ≤ ς(kj τj )µj , with i = 1, 2. Then,
we say Y1 <♭ Y2 if (k1′ , τ1′ , µ′1 , h′1 ) <invlex (k2′ , τ2′ , µ′2 , h′2 ).
82 YI HU

(4) Suppose Y1 , Y2 ∈ D(ð(kτ ) rµ−1 ),̟ . Write Y1 = X(ð(kτ ) rµ−1 ),u and Y2 = X(ð(kτ ) rµ−1 ),v .
Then, Y1 <♭ Y2 if u <lex v.
(5) Suppose Y1 , Y2 ∈ D(ð(kτ ) rµ−1 ),̺ . Write Yi = X(ð(kτ ) rµ−1 ),(ui ,vi ) with i = 1, 2.
Then, Y1 <♭ Y2 if (u1 , v1 ) <lex (u2 , v2 ).

Using the order “ <♭ ” of Definition 7.2, and applying Definition 3.12, one sees
that the set Ψ(kτ ) now comes equipped with the total order “ <:=<♭,lex ” which is
the lexicographic order induced by “ <♭ ”. Thus, we can list Ψ(kτ )µ as

Ψ(kτ )µ = {ψ(kτ )µ1 < · · · < ψ(kτ )µς(kτ )µ }

for some finite positive integer ς(kτ )µ depending on (kτ )µ. We then let the set Zð(kτ ) rµ
of the corresponding ð-centers inherit the total order from that of Ψ(kτ )µ . Then, we
can express
Zð(kτ ) rµ = {Zψ(kτ )µ1 < · · · < Zψ(kτ )µς(kτ )µ }.

We let Re(ð(kτ ) rµ s1 ) −→ Re(ð(kτ ) rµ−1 ) be the blowup of Re(ð(kτ ) rµ−1 ) along the ð-center
Zψ(kτ )µ1 . Inductively, we assume that Re(ð(kτ ) rµ s(h−1) ) has been constructed for some
h ∈ [ς(kτ )µ ]. We then let

Re(ð(kτ ) rµ sh ) −→ Re(ð(kτ ) rµ sh−1 )

be the blowup of Re(ð(kτ ) rµ sh−1 ) along (the proper transform of) the ð-center Zψ(kτ )µh .
Here, to reconcile notation, we set Re(ð(kτ ) rµ s0 ) := Re(ð(kτ ) rµ−1 ) .
All of these can be summarized as the sequence

Re(ð(kτ ) rµ ) := Re(ð(kτ ) rµ sς(kτ )µ ) −→ · · · −→ Re(ð(kτ ) rµ s1 ) −→ Re(ð(kτ ) rµ s0 ) := Re(ð(kτ ) rµ−1 ) .

For any h ∈ [ς(kτ )µ ], consider the induced morphism Re(ð(kτ ) rµ sh ) −→ Re(ð(kτ ) rµ−1 ) .
• We let X(ð r s ),w be the proper transform of X(ð r ),w in Re(ð r s ) , for
(kτ ) µ h (kτ ) µ−1 (kτ ) µ h

all w ∈ Id,n \m. These are still called ̟-divisors. We denote the set of all ̟-divisors
on Re(ð r s ) by D(ð r s ),̟ .
(kτ ) µ h (kτ ) µ h

• We let X(ð(kτ ) rµ sh ),(u,v) be the proper transform of the ̺-divisor X(ð(kτ ) rµ−1 ),(u,v)
in Re(ð(kτ ) rµ sh ) , for all (u, v) ∈ ΛFm . These are still called ̺-divisors. We denote the
set of all ̺-divisors on Re(ð r s ) by D(ð r s ),̺ .
(kτ ) µ h (kτ ) µ h
GRASSMANNIANS AND SINGULARITIES 83

• We let E(ð(kτ ) rµ sh ),(k′ τ ′ )µ′ h′ be the proper transform of E(ð(kτ ) rµ−1 ),(k′ τ ′ )µ′ h′ in Re(ð(kτ ) rµ sh ) ,
for all (11)0 ≤ (k ′ τ ′ )µ′ ≤ (kτ )(µ − 1) and h′ ∈ [ς(k′ τ ′ )µ′ ]. (Here, E(ð(kτ ) rµ−1 ),(11)0h with
h ∈ [ς(11)0 ] correspond to ℘-divisors, where ς(11)0 = Υ + |Φ|.). These are named as
ð-exceptional or simply exceptional divisors. We denote the set of these exceptional
divisors on Re(ð r s ) by E(ð r s ),old .
(kτ ) µ h (kτ ) µ h

We let

D̄(ð(kτ ) rµ sh ) = D(ð(kτ ) rµ sh ),̟ ⊔ D(ð(kτ ) rµ sh ),̺ ⊔ E(ð(kτ ) rµ sh ),old

be the set of all of the aforementioned divisors on Re(ð(kτ ) rµ sh ) . As each divisor in


D̄(ð(kτ ) rµ sh ) is the proper transform of a unique divisor in D(ð(kτ ) rµ−1 ) . The total order
on D(ð(kτ ) rµ−1 ) carries over to provide an induced total order on the set D̄(ð(kτ ) rµ sh ) .
In addition to the proper transforms of the divisors from Re(ð(kτ ) rµ−1 ) , there are the
following new exceptional divisors.
For any h ∈ [ς(kτ )µ ], we let E(ð(kτ ) rµ sh ) be the exceptional divisor of the blowup
Re(ð(kτ ) rµ sh ) −→ Re(ð(kτ ) rµ sh−1 ) . For any 1 ≤ h′ < h ≤ ς(kτ )µ , we let E(ð(kτ ) rµ sh ),(kτ )µh′
be the proper transform in Re(ð r s ) of the exceptional divisor E(ð r s ′ ) . To
(kτ ) µ h (kτ ) µ h

reconcile notation, we also set E(ð(kτ ) rµ sh ),(kτ )µh := E(ð(kτ ) rµ sh ) . We set

E(ð(kτ ) rµ sh ),new = {E(ð(kτ ) rµ sh ),(kτ )µh′ | 1 ≤ h′ ≤ h ≤ ς(kτ )µ }.

We then order the exceptional divisors of E(ð(kτ ) rµ sh ),new in the reverse order of oc-
currence, that is, E(ð(kτ ) rµ sh ),(kτ )µh′′ ≤ E(ð(kτ ) rµ sh ),(kτ )µh′ if h′′ ≥ h.
We then let

D(ð(kτ ) rµ sh ) = D̄(ð(kτ ) rµ sh ) ⊔ E(ð(kτ ) rµ sh ),new .

For any Y1 ∈ E(ð(kτ ) rµ sh ),new and any Y2 ∈ D̄(ð(kτ ) )rµ sh , we say Y1 <♭ Y2 . As each of
E(ð(kτ ) ),new and D̄(ð(kτ ) )sh is a totally ordered set, this endows D(ð(kτ ) rµ sh ) with a total
order “ <♭ ”.
Finally, we set Re(ð(kτ ) )rµ := Re(ð(kτ ) rµ sς(kτ )µ ) , and let

D(ð(kτ ) rµ ),̟ = D(ð(kτ ) rµ sς(kτ )µ ),̟ , D(ð(kτ ) rµ ),̺ = D(ð(kτ ) rµ sς(kτ )µ ),̺ ,

Eð(kτ ) rµ = E(ð(kτ ) rµ sς(kτ )µ ),old ⊔ E(ð(kτ ) rµ sς(kτ )µ ),new


84 YI HU

This can be summarized as

D(ð(kτ ) rµ ) = D(ð(kτ ) rµ ),̟ ⊔ D(ð(kτ ) rµ ),̺ ⊔ E(ð(kτ ) rµ ) .

This way, we have equipped the scheme Re(ð(kτ ) rµ ) in (ð(kτ ) rµ ) with the set D(ð(kτ ) rµ ),̟
of ̟-divisors, the set D(ð(kτ ) rµ ),̺ of ̟-divisors, and the set Eð(kτ ) rµ of exceptional
divisors, as required by induction.
Now, we are ready to introduce the notion of “association” in (ð(kτ ) rµ ), as required
to carry on the process of induction.
We do it inductively on the set [ς(kτ )µ ]

Definition 7.3. Fix any B ∈ Bq ∪ Bmn . We let TB be any fixed term of the bino-
mial B. Meanwhile, we also fix any F̄ ∈ Fm and let Ts be the fixed term of LF
corresponding to some s ∈ SF .
We assume that the notion of “association” in (ð(kτ ) rµ sh−1 ) has been introduced.
That is, for every divisor Y ′ ∈ D(ð(kτ ) rµ sh−1 ) , the multiplicities mY ′ ,TB and mY ′ ,s have
been defined.
Consider an arbitrary divisor Y ∈ D(ð(kτ ) rµ sh ) .
First, suppose Y 6= E(ð(kτ ) rµ sh ) . Then, it is the proper transform of a (unique)
divisor Y ′ ∈ D(ð(kτ ) rµ sh−1 ) . In this case, we set

mY,TB = mY ′ ,TB , mY,s = mY ′ ,s .

Next, we consider the exceptional Y = E(ð(kτ ) rµ sh ) .


We let ψ = ψ(kτ )µh . We have that

ψ = {Y + , Y − } ⊂ D(ð(kτ ) rµ−1 ) .

For any B ∈ Bq ∪ Bmn , we write B = TB0 − TB1 . We let

mψ,TBi = mY + ,TBi + mY − ,TBi , i = 0, 1, and

lψ,B = min{mψ,TB0 , mψ,TB1 }.

(For instance, by definition, lψ,B > 0, when B = B(kτ ) .) Then, we let

mE(ð i
,TB = mψ,TBi − lψ,TBi .
(kτ ) rµ sh )
GRASSMANNIANS AND SINGULARITIES 85

Likewise, for s ∈ SF with F ∈ Fm , we let

mE(ð ,s = mY + ,s + mY − ,s .
(kτ ) rµ sh )

We say Y is associated with TB (resp Ts ) if its multiplicity mY,TB (resp. mY,s ) is


positive. We do not say Y is associated with TB (resp Ts ) if its multiplicity mY,TB
(resp. mY,s ) equals to zero.

When h = ς(kτ )µ , we obtain all the desired data on Re(ð(kτ ) rµ ) = Re(ð(kτ ) rµ )sς(kτ )µ .
Now, with all the aforedescribed data equipped for Re(ð(kτ ) rµ ) , we obtain our in-
ductive package in (ð(kτ ) rµ ). This allows us to introduce the ð-sets Ψð(kτ ) rµ+1 and
ð-centers Zð(kτ ) rµ+1 in (ð(kτ ) rµ+1 ) as in Definition 7.1, endow total orders on Ψð(kτ ) rµ+1
and Zð(kτ ) rµ+1 as in the paragraph immediately following Definition 7.2, and then
advance to the next round of the ℘-blowups. Here, to reconcile notations, we set

(ð(kτ ) rκ(kτ )+1 ) := (ð((k(τ +1)) r1 ), 1 ≤ τ < tFk ;

(ð(ktFk ) rκ(ktF )+1


) := (ð((k+1)1) r1 ), 1 ≤ k < Υ,
k

provided that κ(kτ ) and κ(ktFk ) are (proved to be) finite.


For any (kτ )µh ∈ IndexΨ , the ð-blowup in (ð(kτ ) rµ sh ) gives rise to

Ve(ð(kτ ) rµ sh )  Re(ð(kτ ) rµ sh )

(7.1) /

 

V  / RF ,

where Ve(ð(kτ ) rµ sh ) is the proper transform of V in Re(ð(kτ ) rµ sh ) .


We let Ve(ð(kτ ) rµ ) = Ve(ð(kτ ) rµ sς(kτ )µ ) .

Definition 7.4. Fix any k ∈ [Υ], τ ∈ [tFk ]. Suppose there exists a finite integer µ
such that for any pre-ð-set ψ in (ð(kτ ) rµ+1 ) (cf. Definition 7.1), we have

Zψ ∩ Ve(ð(kτ ) rµ ) = ∅.

Then, we let κ(kτ ) be the smallest integer such that the above holds. Otherwise, we
let κ(kτ ) = ∞.
86 YI HU

It will be shown soon that κ(kτ ) is finite for all k ∈ [Υ], τ ∈ [tFk ].
For later use, we let

(7.2) Ψ = {ψ(kτ )µh | (kτ ) ∈ IndexBmn , 1 ≤ µ ≤ κ(kτ ) , h ∈ [ς(kτ )µ ]},

IndexΨ = {(kτ )µh | (kτ ) ∈ IndexBmn , 1 ≤ µ ≤ κ(kτ ) , h ∈ [ς(kτ )µ ]}.

Upon proving that κ(kτ ) is finite for all for all (kτ ) ∈ IndexBmn later on, we can
summarize the process of ð-blowups as a single sequence of blowup morphisms:

(7.3) Reð −→ · · · −→ Re(ð(kτ ) rµ sh ) −→ Re(ð(kτ ) rµ sh−1 ) −→ · · · −→ Re(ð(11) r0 ) := R℘ ,

where Reð := Re(ð(ΥtΥ ) rκΥt ) := Re(ð(ΥtΥ ) rκΥt sς(Υt ) is the blowup scheme reached
Υ Υ Υ )κΥtΥ

in the final step (ð(ΥtΥ ) rκΥtΥ sς(ΥtΥ )κΥt ) of all ð-blowups.


Υ

The end of all ð-blowups gives rise to the following induced diagram

Veð  Reð

(7.4) /

 

V  / RF ,

where Veð is the proper transform of V in Reð .

7.3. Properties of ð-blowups.


To begin with, recall that we have set Re(ð11 r1 s0 ) := Re℘ .
Fix and consider any (kτ )µh ∈ IndexΨ (see (7.2)).

Proposition 7.5. Suppose that the scheme Re(ð(kτ ) rµ sh ) has been constructed.
Consider any standard chart V of Re(ð(kτ ) rµ sh ) , lying over a unique chart V[0] of
RF . We suppose that the chart V[0] is indexed by ΛoFm = {(vsF,o , vsF,o ) | F̄ ∈ Fm }.
As earlier, we have Λ⋆Fm = ΛFm \ΛoFm .
Then, the chart V comes equipped with

a subset eV ⊂ Id,n \m and a subset dV ⊂ Λ⋆Fm

such that every exceptional divisor E (i.e., not a ̟- nor a ̺-divisor) of Re(ð(kτ ) rµ sh )
with E ∩ V 6= ∅ is either labeled by a unique element w ∈ eV or labeled by a unique
element (u, v) ∈ dV . We let E(ð(kτ ) rµ sh ),w be the unique exceptional divisor on the
GRASSMANNIANS AND SINGULARITIES 87

chart V labeled by w ∈ eV ; we call it an ̟-exceptional divisor. We let E(ð(kτ ) rµ sh ),(u,v)


be the unique exceptional divisor on the chart V labeled by (u, v) ∈ dV ; we call it
an ̺-exceptional divisor. (We note here that being ̟-exceptional or ̺-exceptional is
strictly relative to the given standard chart.)
Further, the chart V admits a set of free variables
 
 ε , δ w ∈ eV , (u, v) ∈ dV 
V,w V,(u,v)
(7.5) VarV := ,
 xV,w , xV,(u,v) w ∈ Id,n \m\eV , (u, v) ∈ Λ⋆F \dV 
m

such that it is canonically isomorphic to the affine space with the variables in (7.5)
as its coordinate variables. Moreover, on the standard chart V, we have

(1) the divisor X(ð(kτ ) rµ sh ),w ∩V is defined by (xV,w = 0) for every w ∈ Id,n \m\eV ;
(2) the divisor X(ð(kτ ) rµ sh ),(u,v) ∩ V is defined by (xV,(u,v) = 0) for every (u, v) ∈
Λ⋆Fm \dV ;
(3) the divisor X(ð(kτ ) rµ sh ),w ∩ V does not intersect the chart for all w ∈ eV ;
(4) the divisor X(ð(kτ ) rµ sh ),(u,v) does not intersect the chart for all (u, v) ∈ dV ;
(5) the ̟-exceptional divisor E(ð(kτ ) rµ sh ),w ∩ V labeled by an element w ∈ eV is
define by (εV,w = 0) for all w ∈ eV ;
(6) the ̺-exceptional divisor E(ð(kτ ) rµ sh ),(u,v) ∩V labeled by an element (u, v) ∈ dV
is define by (δV,(u,v) = 0) for all (u, v) ∈ dV ;
(7) any of the remaining exceptional divisor of Re(ð(kτ ) rµ sh ) other than those that
are labelled by some some w ∈ eV or (u, v) ∈ dV does not intersect the chart.

Proof. We prove by induction on (kτ )µh ∈ {(11)10}⊔IndexΨ . (The proof is basically


parallel to that of Proposition 6.7. We provide details for completeness.)
For the initial case, the scheme is Re(ð11 r1 s0 ) := R℘ , the scheme obtained in the
final step of ℘-blowups. Then, this proposition is the same as Proposition 6.7 for
R℘ . Thus, it holds.
We suppose that the statement holds over Re(ð(kτ ) rµ sh−1 ) with (kτ )µh ∈ IndexΨ .
We now consider Re(ð r s ) . We have the embedding
(kτ ) µ h

Re(ð(kτ ) rµ sh )  Re(ð(kτ ) rµ sh−1 ) × Pψ(kτ )µh ,


 /
88 YI HU

where Pψ(kτ )µh (∼


= P1 ) is the factor projective space. The ð-set ψ(kτ )µh consists of
two divisors {Y + , Y − } on Re(ð(kτ ) rµ−1 ) = Re(ð(kτ ) rµ−1 sς(kτ )(µ−1) ) . We let Y0′ and Y1′ be the
proper transforms in Re(ð r s ) of the two divisors in the ð-set ψ(kτ )µh = {Y + , Y − }.
(kτ ) µ h−1

In addition, we let [ξ0 , ξ1 ] be the homogenous coordinates of Pψ(kτ )µh , corresponding


to (Y0′ , Y1′ ).
Fix any standard chart V′ of Re(ð(kτ ) rµ sh−1 ) , let i be any of 0 or 1 and set (ξi ≡ 1).
Then, we obtain a standard chart, V = (V′ × (ξi ≡ 1)) ∩ Re(ð r s ) , of Re(ð r s ) .
(kτ ) µ h (kτ ) µ h

Any standard chart of Re(ð(kτ ) rµ sh ) is such a form. As V is assumed to lie over V[0] , so
does V′ . By assumption, the chart V′ comes equipped with a subset eV′ ⊂ Id,n \m,
a subset dV′ ⊂ Λ⋆Fm , and admits a set of coordinate variables
 
 ε ′ , δ ′ w ∈ e , (u, v) ∈ d 
V ,w V ,(u,v) V′ V′
(7.6) VarV′ := ,
 xV′ ,w , xV′ ,(u,v) w ∈ Id,n \m\eV′ , (u, v) ∈ Λ⋆F \dV′ 
m

verifying the properties (1)-(7) as in the proposition.


We now prove the statements for the chart V of Re(ð (kτ ) rµ sh )
.
First, we suppose that the proper transform Zψ′ (kτ )µh in Re(ð(kτ ) rµ sh−1 ) of the ð-center
Zψ(kτ )µh does not meet the chart V′ . Then, we let V inherit all the data from those
of V′ , that is, we set eV = eV′ , dV = dV′ , and VarV = VarV′ : changing the subindex
“ V′ ” for all the variables in VarV′ to “ V ”. As the ð-blowup along the proper
transform Zψ′ (kτ )µh does not affect the chart V′ , one sees that the statements of the
proposition hold for V.
Next, we suppose that the proper transform Zψ′ (kτ )µh meets the chart V′ along a
nonempty closed subset. That is, we have Y0′ ∩ Y1′ ∩ V′ 6= ∅. On the chart V′ , by
the inductive assumption, we can suppose

Yj′ ∩ V′ = (yj′ = 0), for some yj′ ∈ VarV′ with j = 0, 1.

Fix and consider the chart (ξi ≡ 1) for some i = 0 or 1. We let j ∈ {0, 1}\i. Then,
the chart V = (V′ × (ξi ≡ 1)) ∩ Re(ð(kτ ) rµ sh ) of the scheme Re(ð(kτ ) rµ sh ) , as a closed
subscheme of V′ × (ξi ≡ 1), is defined by

(7.7) yj′ = yi′ ξj .


GRASSMANNIANS AND SINGULARITIES 89

There are 4 possibilities for Yi′ ∩ V′ according to the assumption on the chart V′ .
Based on every of such possibilities, we set


 eV = eV′ ⊔ {w}, dV = dV′ , if yi′ = xV′ ,w for some w ∈ Id,n \m\eV′



 e = e ′ , d = d ′ ⊔ {(u, v)}, if y ′ = x ′ ⋆
V V V V i V ,(u,v) with (u, v) ∈ ΛFm \dV′
(7.8)

 eV = eV′ , dV = dV′ , if yi′ = εV′ ,w for some w ∈ eV′



 d = d ′, e = e ′,
V V V V if yi′ = δV′ ,(u,v) for some (u, v) ∈ dV′ .
Accordingly, we introduce


 εV,w = yi′ , xV,w = 1, if yi′ = xV′ ,w for some w ∈ eV \eV′



 δ ′
V,(u,v) = yi , xV,(u,v) = 1, if yi′ = xV′ ,(u,v) for some (u, v) ∈ dV \dV′
(7.9)

 εV,w = yi′ , if yi′ = εV′ ,w for some w ∈ eV′ = eV



 δ ′
V,(u,v) = yi , if yi′ = δV′ ,(u,v) for some (u, v) ∈ dV′ = dV .

To introduce the set VarV , we then set




 xV,a = ξj , if yj′ = xV′ ,a



 x
V,(a,b)
= ξj , if yj′ = xV′ ,(a,b)
(7.10)

 εV,a = ξj , if yj′ = εV′ ,a



 δ
V,(a,b)
= ξj , if yj′ = δV′ ,(a,b) .
Thus, we have introduced yi′ , ξj ∈ VarV where yi′ , respectively, ξj , is endowed with
its new name as in (7.9), respectively, in (7.10). Next, we define the set VarV \yi′ \ξj
to consist of the following variables:

 ∀ w ∈ Id,n \m\eV and xV′ ,w 6= yj′
 xV,w = xV′ ,w ,



 x
V,(u,v) = xV′ ,(u,v) , ∀ (u, v) ∈ Λ⋆Fm \dV and xV′ ,(u,v) 6= yj′
(7.11)

 εV,w = εV′ ,w , ∀ w ∈ eV and εV′ ,w 6= yj′ , yi′



 δ
V,(u,v) = δV′ ,(u,v) , ∀ (u, v) ∈ dV and δV′ ,(u,v) 6= yj′ , yi′ .

Substituting (7.7), one checks that the chart V is isomorphic to the affine space
with the variables in (7.9), (7.10), and (7.11), as its coordinate variables. Putting
all together, the above matches description of VarV in (7.5),
Now, it remains to verity (1)-(7) of the proposition on the chart V.
First, consider the unique new exceptional divisor E(ð(kτ ) rµ sh ) created by the blowup

Re(ð(kτ ) rµ sh ) −→ Re(ð(kτ ) rµ sh−1 ) .


90 YI HU

Then, we have
E(ð(kτ ) rµ sh ) ∩ V = (yi′ = 0)

where yi′ is renamed as in (7.9) according to the four possibilities of its form in
VarV′ . This way, the new exceptional divisor E(ð(kτ ) rµ sh ) is labelled on the chart V.
In any case of the four situations, we have that the proper transform in Re(ð r s ) (kτ ) µ h

of Yi′ does not meet the chart V, and if Yi′ is a ̟-exceptional divisor labeled by
some element of eV′ (resp, ̺-exceptional divisor labeled by some element of dV′ ),
then, on the chart V, its proper transform is no longer labelled by any element of
Id,n \m (resp. Λ⋆Fm ). This verifies any of (3)-(7) whenever the statement therein
involves the newly created exceptional divisor E(ð(kτ ) rµ sh ) and/or its corresponding
exceptional variable yi′ (which is renamed in (7.9) to match that of (7.5) in the
statement of the proposition). Observe that the statements (1) and (2) are not
related to the new exceptional divisor E(ð(kτ ) rµ sh ) or its corresponding exceptional
variable yi′ (again, renamed in (7.9)).
For any of the remaining ̟-, ̺-, and exceptional divisors on Re(ð(kτ ) rµ sh ) , it is
the proper transform of a unique corresponding ̟-, ̺-, and exceptional divisor on
Re(ð(kτ ) rµ sh−1 ) . Hence, applying the inductive assumption on V′ , one verifies directly
that every of the properties (1)-(7) of the proposition is satisfied whenever it applies
to such a divisor and/or its corresponding local variable in VarV .
This completes the proof. 

7.4. Combining all ϑ-blowups, ℘-blowups, and ð-blowups.


For the convenience of narration, we combine all ϑ-blowups, ℘-blowups, and ð-
blowups into a single grand sequence.

Definition 7.6. We introduce


( )
ℏ equals to ϑ with k ∈ [Υ], or (℘ r s ) with
[k] (k ′ τ ′ ) µ′ h ′
(7.12) a := ℏ .
(k ′ τ ′ )µ′ h′ ∈ IndexΦ , or (ð(kτ ) rµ sh ) with (kτ )µh ∈ IndexΨ

The set a comes equipped with a partial order induced from the total orders of the
sets [Υ], IndexΦ , and IndexΨ . We then let

ϑ[k] < (℘(k′ τ ′ ) rµ′ sh′ ) < (ð(kτ ) rµ sh )


GRASSMANNIANS AND SINGULARITIES 91

for all ϑ[k] with k ∈ [Υ], (℘(k′ τ ′ ) rµ′ sh′ ) with (k ′ τ ′ )µ′ h′ ∈ IndexΦ , and (ð(kτ ) rµ sh ) with
(kτ )µh ∈ IndexΨ . This way, a is endowed with a total order compatible with the
orders of ϑ-, ℘-, and ð-blowups.

Definition 7.7. (cf. Definition 5.6) Consider an arbitrary standard chart of V of


Reℏ with ℏ ∈ a. We let Veℏ ⊂ Reℏ be the proper transform of V in Reℏ .
Let z ∈ V be any fixed closed point of the chart. Consider any main binomial
relation B ∈ Bmn . We say that BV
mn
terminates at a closed point z ∈ V if (at least)
one of its two terms does not vanish at z.
mn
We say BV terminates on the chart V if it terminates at all closed points of
Veℏ ∩ V. We say B mn terminates on the scheme Reℏ if BV mn
terminates over all
standard charts V of Reℏ .

In a similar vein, for all F̄ ∈ Fm , the linearized Plücker equation LV,F automat-
ically terminates on the chart V because LV,F is the pullback of LV[0] ,F , where V[0]
is the unique standard chart of RF such that V lies over V[0] , and at any closed
point of V[0] , at least one of the terms of LV[0] ,F does not vanish.

Lemma 7.8. Consider any two indexes ℏ′ , ℏ ∈ a with ℏ′ < ℏ. Let V be a standard
chart of Reℏ , lying over a unique standard chart V′ of Reℏ′ . Fix and consider any

main binomial relation B ∈ Bmn . Assume that BV
mn mn
′ terminates on V . Then, BV

terminates on V.

Proof. Under the assumption, we have


mn ∗ mn ∗ + ∗ −
BV = πV,V′ BV′ = πV,V′ TV′ − πV,V′ TV′ ,

where πV,V′ : V −→ V′ is the (induced) projection. Then, one sees that the state-
ment follows immediately. 

Fix and consider any of ϑ-blowups, ℘-blowups, and ð-blowups, that is, we consider
Reϑ → Reϑ
[k]
for some k ∈ [Υ], or Re(℘ r s ) → Re(℘ r s ) for some (kτ )µh ∈
[k−1] (kτ ) µ h (kτ ) µ h−1

IndexΦ , or Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 ) for some (kτ )µh ∈ IndexΨ . We let

Reℏ → Reℏ′

denote the above fixed blowup.


92 YI HU

We let Eℏ be the exceptional divisor created by the above fixed blowup.


Consider an arbitrary standard chart V of Reℏ , lying over a (unique) standard
chart V′ of Reℏ′ . Suppose the chart V′ intersects the center of the blowup Reℏ → Reℏ′
along a nonempty closed subset. Then, on the chart V′ , the corresponding ϑ- or ℘-
or ð-center, corresponds to two variables

{y0′ , y1′ } ⊂ VarV′ .

We let P1[ξ0 ,ξ1 ] be the corresponding factor projective space with [ξ0 , ξ1] corresponding
to (y0′ , y1′ ). We suppose that the chart V corresponds to (ξi ≡ 1) for one of i ∈ {0, 1}.
We let j ∈ {0, 1}\i and let yj ∈ VarV be the proper transform of yj′ . We let ζV ∈ VarV
be the exceptional parameter of Eℏ on the chart V (it corresponds to yi′ ).
We point out here that in the above, if ℏ = ϑ[k] , then it corresponds to the fixed
Fk ; if ℏ = (℘(kτ ) rµ sh ) or (ð(kτ ) rµ sh ), then it corresponds to the fixed B(kτ ) .

Definition 7.9. Consider B(kτ ) ∈ Bmn . (Here, if ℏ = ϑ[k] , then we consider any
τ ∈ [tFk ].) Suppose B(kτ ) does not terminate on the chart V′ , but terminates on
the chart V. Then, we call the proper transform yj ∈ VarV of yj′ ∈ VarV′ the
terminating central variable for the binomial equation B(kτ ) on the chart V. We call
ζV the terminating exceptional parameter on the chart V.
Moreover, for any standard chart V e of some fixed ϑ- or ℘- or ð-blowup scheme,
lying over the above chart V, we also say B e e we also
terminates on the chart V;
V,(kτ )

e ∈ VarV
call the proper transform yV,j e of yj the terminating central variable for the
e
binomial equation B(kτ ) on the chart V.
Likewise, we call the proper transform ζVe ∈ VarVe of ζV ∈ VarV the terminating
e
exceptional parameter for the binomial equation B(kτ ) on the chart V.
In particular, the above applies to the standard charts of the final scheme Reð .

7.5. Proper transforms of defining equations in (ð(kτ ) rµ sh ).


We keep the notations of Proposition 7.5 as well as those in its proof.
In particular, we have that V is a standard chart of Re(ð(kτ ) rµ sh ) , lying over a stan-
dard chart V′ of Re(ð(kτ ) rµ sh−1 ) . We suppose that B(kτ ) corresponds to s ∈ SFk \sFk .
GRASSMANNIANS AND SINGULARITIES 93

We then let
′ ′ ′
ψ(kτ )µh = {Y0 = X(ð(kτ ) rµ sh−1 ),(us ,vs ) , Y1 }

be the proper transforms of ð-set ψ(kτ )µh in Re(ð(kτ ) rµ sh−1 ) ; we let Zψ′ (kτ )µh be proper
transform of the ð-center Zψ in Re(ð r s ) . In addition, we assume that
(kτ )µh (kτ ) µ h−1

Zψ′ (kτ )µh ∩ V 6= ∅. For convenience, we let y0′ = xV′ ,(us ,vs ) . We then have

(7.13) Y0′ ∩ V′ = (y0′ = 0)

Y1′ ∩ V′ = (y1′ = 0), for some y1′ ∈ VarV′ .

Note here that the first identity holds according to Proposition 7.5 (2). Furthermore,
we let [ξ0 , ξ1 ] be the homogenous coordinates of the factor projective space Pψ(kτ )µh
corresponding to (y0′ , y1′ ).
Consider any fixed B ∈ Bmn ∪ Bq and F̄ ∈ Fm . Suppose BV′ and LV′ ,F have
been constructed over V′ . Applying Definition 5.4, we obtain the proper transforms
on the chart V
BV , B ∈ Bmn ∪ Bq ; LV,F , F̄ ∈ Fm .

In what follows, for any B = TB+ − TB− ∈ Bmn , we express BV = TV,B


+ −
− TV,B . If
+ −
B = B(kτ ) for some k ∈ [Υ] and τ ∈ [tFk ], we also write BV = TV,(kτ ) − TV,(kτ ) .

Proposition 7.10. Let the notation be as in Proposition 7.5 and be as in above.


Let V be any standard chart of Re(ð r s ) . Then, the scheme Ve(ð r s ) ∩ V, as
(kτ ) µ h (kτ ) µ h

a closed subscheme of the chart V, is defined by

mn q
BV , BV , LV,Fm .

Suppose that Zψ′ (kτ )µh ∩V′ 6= ∅ where Zψ′ (kτ )µh is the proper transform of the ð-center
Zψ in Re(ð r s ) . Further, we let ζ = ζV,(kτ )µh be the exceptional parameter
(kτ )µh (kτ ) µ h−1

in VarV such that


E(ð(kτ ) rµ sh ) ∩ V = (ζ = 0).

Then, the following hold.

(1) Suppose V = (V′ × (ξ0 ≡ 1)) ∩ Re(ð(kτ ) rµ sh ) . We let y1 ∈ VarV be the proper
transform of y1′ ∈ VarV′ . Then, we have
94 YI HU

+ +
(1a) The plus-term TV,(kτ ) terminates on the chart V. y1 ∤ TV,(kτ ) . Suppose
degy1′ TV−′ ,(kτ ) = b for some integer b, positive by definition, then we have
− −
degζ TV,(kτ ) = b − 1. Consequently, either TV,(kτ ) is linear in y1 or else
− ′
ζ | TV,(kτ ) . (Note here that y0 ∈ VarV becomes ζ ∈ VarV .)

+ − −
(1b) Let B ∈ Bmn \{B(kτ ) }. Then, y1 ∤ TV,B ; if y1 | TV,B , then ζ | TV,B .
(1c) Consider any fixed term TB of any given B ∈ Bq . Suppose y1 | TV,B ,
then either TV,B is linear in y1 or ζ | TV,B .
(2) Suppose V = (V′ × (ξ1 ≡ 1)) ∩ Re(ð(kτ ) rµ sh ) . We let y0 = xV,(us ,vs ) ∈ VarV be
the proper transform of y0′ = xV′ ,(us ,vs ) . Then, we have
+ − − −
(2a) TV,(kτ ) is linear in y0 . y0 ∤ TV,(kτ ) . deg(TV,(kτ ) ) = deg(TV′ ,(kτ ) ) − 1. In-
deed, if we let degy1′ TV−′ ,(kτ ) = b for some integer b, positive by definition,
− ′
then we have degζ TV,(kτ ) = b − 1. Note here that y1 ∈ VarV becomes

the exceptional variable ζ ∈ VarV .


±
(2b) Let B ∈ Bmn \{B(kτ ) }. Then, y0 ∤ TV,B .
(2c) Consider any fixed term TB of any given B ∈ Bq . Suppose y0 | TV,B ,
then TV,B is linear in y0 = xV,(us ,vs ) .
Consequently, we have
(3) κ(kτ ) < ∞;
(4) Moreover, consider Re(ð(kτ ) rκ(kτ ) ) = Re(ð(kτ ) rκ(kτ ) sς(kτ )κ ). For every B ∈ Bmn
(kτ )

with B ≤ B(kτ ) , we have that B terminates on Re(ð(kτ ) rκ(kτ ) ) .

Proof. We continue to follow the notation in the proof of Proposition 7.5. (In many
ways, the proofs are similar to those of Proposition 6.9.)
We prove the proposition by applying induction on (kτ )µh ∈ ((11)10) ⊔ IndexΨ .
The initial case is ((11)10) with Re(ð(11) r1 s0 ) = Re℘ . In this case, the statement
about defining equations of Re(ð r s ) ∩ V follows from Proposition 6.9 for Re℘ ; the
(11) 1 0

remainder statements (1) - (4) are void.


Assume that the proposition holds for (ð(kτ ) rµ sh−1 ) with (kτ )µh ∈ IndexΨ .
Consider (ð(kτ ) rµ sh ). Consider a standard chart V of Re(ð(kτ ) rµ sh ) , lying over a
unique standard chart of V′ over Re(ð r s ) . By assumption, all the desired state-
(kτ ) µ h−1

ments of the proposition hold over V′ .


GRASSMANNIANS AND SINGULARITIES 95

To begin with, we suppose that the proper transform Zψ′ (kτ )µh of the ð-center
Zψ in Re(ð r s ) does not meet the chart V′ . Then, by the proof of Proposition
(kτ )µh (kτ ) µ h−1

7.5, we have that V retains all the data from those of V′ . As the ð-blowup along the
proper transform of Zψ(kτ )µh does not affect the chart V′ , we have that the statement
of the proposition on the defining equations follows immediately from the inductive
assumption. The statements (1) and (2) are void.
In what follows, we suppose that the proper transform Zψ′ (kτ )µh meets the chart
V′ along a nonempty closed subset.
The statement of the proposition on the defining equations of Ve(ð(kτ ) rµ sh ) ∩ V
follows straightforwardly from the inductive assumption.
Proof of (1).
+
(1a). The fact that TV,(kτ ) terminates on the chart V follows from the form of the
ð-set ψ(kτ )µh on the chart V′ (see (7.13)) and Proposition 6.9 (4). The remainder
follows from a direct check.
(1b). Consider any B = B(k′ τ ′ ) ∈ Bmn with (k ′ τ ′ ) 6= (kτ ). We can express
TB+ = x(ut ,vt ) xuk′ for some t ∈ SFk′ and t 6= s (if k ′ = k). Suppose (k ′ τ ′ ) < (kτ ),
by Proposition 7.10 (4) on Re(ð(k′ τ ′ ) rκ(k′ τ ′ ) ) , which holds by the inductive assumption,
TV+′ ,B terminates on V′ . Hence, we have y0′ , y1′ ∤ TV+′ ,B . Suppose (k ′ τ ′ ) > (kτ ),
by Proposition 6.9 (4), we we have y0′ , y1′ ∤ TV+′ ,B . In either case, this implies the
statements of (1b).
(1c). Consider any B ∈ Bq . Fix and consider any term TV′ ,B of BV′ . We assume
y1 | TV,B . If none of the variables {y0′ , y1′ } of (7.13) appears in the other term TV′ ′ ,B
of BV′ , then the statement follows immediately. Suppose one of the variables of
(7.13) appears in TV′ ′ ,B . Note that y0′ = xV′ ,(us ,us ) , a ̺-variable, is linear in BV′ (see
Proposition 4.20), and it becomes the exceptional variable ζ ∈ VarV . We may write
y1′ = ζy1. As degy1′ TV−′ ,B = b for some integer b > 0, one sees that ζ b−1 y1b | TV,B .
This implies the statement.
Proof of (2).
In this case, y1′ = ζ, that is, y1′ is renamed as the exceptional parameter in VarV ,
and we have y0 = xV,(us ,vs ) is the proper transform of y0′ .
96 YI HU

+
(2a) The fact that TV,(kτ ) is linear in y0 = xV,(us ,vs ) follows immediately from the
+
expression T(kτ ) = x(us ,vs ) xuk . The remainder follows from the form of the ð-set
ψ(kτ )µh on the chart V′ (see (7.13)) and a straightforward calculation.
(2b) follows immediately because x(us ,vs ) uniquely appears in B(kτ ) among all
main binomial equations of Bmn .
(2c) follows immediately because BV is linear in ̺-variables (see Proposition 4.20).
Proof of (3). (The proof is analogous to that of Proposition 6.9 (3).)
+
If (1) occurs, then by (1a), TV,(kτ ) terminates, thus BV,(kτ ) terminates.

Assuming (1) does not occur and (2) keeps happening, then, by (2a), as µ in-

creases, after finitely many rounds, TV,(kτ ) must terminate, hence, BV,(kτ ) terminates,
for any standard chart V.
This implies that the ð-blowups in (ð(kτ ) ) must terminate after finitely many
rounds. That is, κ(kτ ) < ∞.
Proof of (4).
Suppose B ∈ Bmn with B < B(kτ ) . We write B = B(k′ τ ′ ) for some (k ′ τ ′ ) ∈
IndexBmn with (k ′ τ ′ ) < (kτ ). Then, B terminates on Re(ð r ) by applying Propo-
(kτ ) κ(kτ )

sition 7.10 (4) in (ð(k′ τ ′ ) ) which holds by the inductive assumption, and Lemma 7.8.
For B = B(kτ ) , it follows immediately from the proof of (3).
Therefore, by induction on (kτ )µh ∈ ((11)10)⊔IndexΨ , Proposition 7.10 is proved.


The following is the special case of Proposition 7.10.

Corollary 7.11. The binomial B terminates on Reð for all B ∈ Bmn .

Corollary 7.12. Let the notation be as in Definition 7.9.


Consider B = B(kτ ) ∈ Bmn with (kτ ) ∈ IndexBmn . Suppose BV,(kτ ) terminates on
a standard chart V of Reℏ for some ℏ ∈ a. Suppose the minus term TV,(kτ

) of BV,(kτ )
is divisible by the terminating exceptional parameter ζV . Then, we have

Eℏ ∩ Veℏ ∩ V = ∅.

In particular, suppose Eℏ ∩ Veℏ ∩ V 6= ∅, then, the binomial equation BV,(kτ ) is linear


in the (unique) terminating central variable y ∈ VarV .
GRASSMANNIANS AND SINGULARITIES 97


Proof. First, observe that we have ζ(z) = 0 for any z ∈ Eℏ ∩V. Hence, if ζV | TV,(kτ ),
we must have Eℏ ∩ Veℏ ∩ V = ∅, because BV,(kτ ) terminates.
Suppose now Eℏ ∩ Veℏ ∩ V 6= ∅.
If V is a standard chart of ϑ-blowup scheme Reϑ[k] for some k ∈ [Υ], then the
statements follow immediately from Proposition 5.12.
If V is a standard chart of ℘-blowup scheme Re℘(kτ ) rµ sh for some (kτ )µh ∈ IndexΦ ,
then the statements follow immediately from Proposition 6.9 (1a) and (2a).
If V is a standard chart of ð-blowup scheme Reð r s for some (kτ )µh ∈ IndexΨ ,
(kτ ) µ h

then the statements follow immediately from Proposition 7.10 (1a) and (2a). 

8. Γ-schemes and Their Transforms

8.1. Γ-schemes.
Here, we return to the initial affine chart Um ⊂ P(∧d E).

Definition 8.1. Let Γ be an arbitrary subset of VarUm = {xu | u ∈ Id,n \m}. We


let IΓ be the ideal of k[xu ]u∈Id,n \m generated by all the elements xu in Γ, and,

I℘,Γ = hxu , F̄ | xu ∈ Γ, F̄ ∈ Fm i

be the ideal of k[xu ]u∈Id,n \m generated by IΓ together with all the in-homogenized
m-primary Plücker relations of Grd,E . We let ZΓ (⊂ Grd,E ∩ Um ) be the closed
subscheme of the affine space Um defined by the ideal I℘,Γ. The subscheme ZΓ is
called the Γ-scheme of Um . Note that ZΓ 6= ∅ since 0 ∈ ZΓ .

(Thus, a Γ-scheme is an intersection of certain Schubert divisors with the chart


Um . But, in this article, we do not investigate Γ-schemes in any Schubert way.)
Take Γ = ∅. Then, I℘,∅ is the ideal generated by all the in-homogenized m-primary
Plücker relations. Thus, Z∅ = Um ∩ Grd,E .
Let Γ be any fixed subset of VarUm . We let Um,Γ be the coordinate subspace of
Um defined by IΓ . That is,

Um,Γ = {(xu = 0)xu ∈Γ } ⊂ Um .



This is a coordinate subspace of dimension nd − 1 − |Γ| where |Γ| is the cardinality
of Γ. Then, ZΓ is the scheme-theoretic intersection of Grd,E with the coordinate
98 YI HU

subspace Um,Γ . For any m-primary Plücker equation F̄ ∈ Fm , we let F̄ |Γ be the


induced polynomial obtained from the in-homogeneous polynomial F̄ by setting
xu = 0 for all xu ∈ Γ. Then, F̄ |Γ becomes a polynomial on the affine subspace
Um,Γ . We point out that F̄ |Γ can be identically zero on Um,Γ .

Definition 8.2. Let Γ be any fixed subset of VarUm . Let (F̄ ) F be any fixed (in-
homogenized) m-primary Plücker relation. We say (F̄ ) F is Γ-irrelevant if every
term of F̄ belongs to the ideal IΓ . Otherwise, we say (F̄ ) F is Γ-relevant. We let
rel
Fm ,Γ be the set of all Γ-relevant in-homogenized m-primary Plücker relations. We
irr
let Fm ,Γ be the set of all Γ-irrelevant in-homogenized m-primary Plücker relations.

If F̄ is Γ-irrelevant, then F̄ |Γ is identically zero along Um,Γ . Indeed, F̄ is Γ-


irrelevant if and only if every term of F̄ contains a member of Γ. The sufficiency
direction is clear. To see the necessary direction, we suppose a term xu xv ∈ IΓ , then
as IΓ is prime (the coordinate subspace Um,Γ is integral), we have xu or xv ∈ Γ.

8.2. F -transforms of Γ-schemes in VF[k] .


In what follows, we keep notation of Proposition 4.18.
Recall that for any F̄ ∈ Fm , ΛF = {(us , vs ) | s ∈ SF }.

Lemma 8.3. Fix any subset Γ of Um . Assume that ZΓ is integral.


Consider Fk ∈ Fm for any fixed k ∈ [Υ].
Then, we have the following:
• there exists a closed subscheme ZF[k] ,Γ of VF[k] with an induced morphism
ZF[k] ,Γ → ZΓ ;

• ZF[k] ,Γ comes equipped with an irreducible component ZF [k] ,Γ
with the induced

morphism ZF [k] ,Γ
→ ZΓ ;
• for any standard chart V of RF[k] such that ZF[k] ,Γ ∩ V 6= ∅, there are two
subsets, possibly empty,

e=0 ⊂ VarV ,
Γ e =1 ⊂ VarV .
Γ
V V

Further, consider any given standard chart V of RF[k] with ZF[k] ,Γ ∩V 6= ∅. Then,

there exists a subset F[k];V,Γ ⊂ Fm such that the following hold.
GRASSMANNIANS AND SINGULARITIES 99

(1) The scheme ZF[k] ,Γ ∩ V, as a closed subscheme of the chart V, is defined by


the following relations

(8.1) e=0 ,
y, y ∈ Γ V

e=1 ,
y − 1, y ∈ Γ V
pre-q
(8.2) BV,[k] ,

BV,(s,t) : xV,(us ,vs ) xV,ut xV,vt − xV,(ut ,vt ) xV,us xV,vs , s, t ∈ SFi i ∈ [k],
X

LV,Fi : sgn(s)xV,(us ,vs ) , F̄i ∈ F[k];V,Γ , i ∈ [k],
s∈SFi
X
F̄V,j : sgn(s)xV,us xV,vs , k < j ≤ Υ.
s∈SFj

e=0 ⊂ VarV to be the maximal subset (under inclusion)


Further, we take Γ V

among all those subsets that satisfy the above.



(2) The induced morphism ZF [k] ,Γ
→ ZΓ is birational.

(3) Fix any variable y = xV,u or y = xV,(u,v) ∈ VarV , ZF [k] ,Γ
∩ V ⊂ (y = 0) if
and only if ZF[k] ,Γ ∩ V ⊂ (y = 0). (We remark here that this property is not
used within this lemma, but will be used as the initial case of Lemma 8.5.)

Proof. We prove the statement by induction on k with k ∈ {0} ∪ [Υ].


When k = 0, we have RF[0] := Um , VF[0] := Um ∩ Grd,E . There exists a unique
chart V = Um . In this case, we set

ZF[0] ,Γ = ZF [0] ,Γ
:= ZΓ

Further, we let
e=0 = Γ, Γ
Γ e=0 = ∅.
V V

Then, the statement holds trivially.


Inductively, we suppose that Lemma 8.3 holds for VF[k−1] ⊂ RF[k−1] .
We now consider VF[k] ⊂ RF[k] .
Recall from (4.19), we have the natural birational morphsim

ρF[k] : VF[k] −→ VF[k−1] ,

induced from the forgetful map RF[k] −→ RF[k−1] .


100 YI HU

First, we suppose Fk is Γ-relevant.


In this case, we set

(8.3) Λ=0 =1
Fk ,Γ := {x(u,v) ∈ ΛFk | xu or xv ∈ Γ}, ΛFk ,Γ := ∅.

(Here, recall the convention of (4.2): x(u,v) = x(v,u) .)


We then let ρ−1
F[k] (ZF[k−1] ,Γ ) be the scheme-theoretic pre-image and define ZF[k] ,Γ

to be the scheme-theoretic intersection

(8.4) ZF[k] ,Γ = ρ−1 =0


F[k] (ZF[k−1] ,Γ ) ∩ (x(u,v) = 0 | (u, v) ∈ ΛFk ),


Next, because Fk is Γ-relevant and ZF [k−1] ,Γ
is birational to ZΓ , one checks that

ZF [k−1] ,Γ
is not contained in the exceptional locus of the birational morphism ρF[k] .
†◦ †
Thus, there exists a Zariski open subset ZF [k−1] ,Γ
of ZF [k−1] ,Γ
such that
†◦ †◦
ρ−1
F[k] (ZF[k−1] ,Γ ) −→ ZF[k−1] ,Γ

is an isomorphism.
We claim
†◦
(8.5) ρ−1 −1 =0
F[k] (ZF[k−1] ,Γ ) ⊂ ZF[k] ,Γ = ρF[k] (ZF[k−1] ,Γ ) ∩ (x(u,v) = 0 | (u, v) ∈ ΛFk ).

To see this, note that since F̄k is Γ-relevant, there exists a term xus xvs of F̄k for some

s ∈ SFk such that it does not vanish generically along ZF [k−1] ,Γ
(which is birational
to ZΓ ). Then, we consider the binomial relation of VF[k] in RF[k]

(8.6) x(u,v) xus xvs − xu xv x(us ,vs ) ,

for any (u, v) ∈ ΛFk . It follows that x(u,v) vanishes identically along ρ−1 †◦ ∼
F[k] (ZF[k−1] ,Γ ) =
†◦
ZF [k−1] ,Γ
if xu or xv ∈ Γ. Hence, (8.5) holds.
† †◦ †
We then let ZF [k] ,Γ
be the closure of ρ−1
[k] (ZF[k−1] ,Γ ) in ZF[k] ,Γ . Since ZF[k] ,Γ is
†◦
closed in ZF[k] ,Γ and contains the Zariski open subset ρ−1
[k] (ZF[k−1] ,Γ ) of ZF[k] ,Γ , it is
an irreducible component of ZF[k] ,Γ .
Further, consider any standard chart V of RF[k] , lying over a unique standard
chart V′ of RF[k−1] , such that ZF[k] ,Γ ∩ V 6= ∅. We set

(8.7) e=0 = Γ
Γ e=0′ ⊔ {x(u,v) ∈ ΛF | xu or xv ∈ Γ}; e=1 = Γ
Γ e=1′ .
V V k V V
GRASSMANNIANS AND SINGULARITIES 101

We are now ready to prove Lemma 8.3 (1), (2) and (3) in the case of RF[k] .
(1). Note that scheme-theoretically, we have

ρ−1 −1
F[k] (ZF[k−1] ,Γ ) ∩ V = πF[k] ,F[k−1] (ZF[k−1] ,Γ ) ∩ VF[k] ∩ V

where πF[k] ,F[k−1] : RF[k] → RF[k−1] is the projection. We can apply Lemma 8.3 (1)
−1
in the case of RF[k−1] to ZF[k−1] ,Γ and πF[k] ,F[k−1]
(ZF[k−1] ,Γ ), apply Proposition 4.18
to VF[k] ∩ V, and use the construction (8.4) of ZF[k] ,Γ (cf. (8.3) and (8.7)), we then
obtain that ZF[k] ,Γ ∩ V, as a closed subscheme of V, is defined by

e=0 ; y − 1, y ∈ Γ
e=1 ; pre-q
y, y ∈ Γ V V B[k] ;

BV,(s,t) , s, t ∈ SFi with all i ∈ [k]


LV,Fi , with F̄i ∈ F[k−1];V′ ,Γ ; LV,Fk ; F̄V,j , k < j ≤ Υ.

⋆ ⋆
Now, we let F[k];V,Γ = F[k−1];V′ ,Γ ∪ {F̄k }. Then, the above implies Lemma 8.3 (1) in

the case of RF[k] .


(2). By construction, we have that the composition ZeF

[k] ,Γ
→ ZeF

[k−1] ,Γ
→ ZΓ is
birational. This proves Lemma 8.3 (2) in the case of RF[k] .

(3). It suffices to prove that if ZF [k] ,Γ
∩ V ⊂ (y = 0), then ZF[k] ,Γ ∩ V ⊂ (y = 0).

If y = xV,u (= xu , cf. the proof of Proposition 4.17), then xu ∈ Γ because ZF [k] ,Γ

is birational to ZΓ . Therefore, ZF[k] ,Γ ∩ V ⊂ (y = 0) by (8.1), which holds by (the


just proved) Lemma 8.3 (1) for RF[k] .
Now assume y = xV,(u,v) . Here, xV,(u,v) is the in-homogenization of x(u,v) (cf. the
proof of Proposition 4.17). Below, upon setting x(us ,vs ) ≡ 1 for all i ∈ [k] (cf.
Fi ,o Fi ,o

Definition 4.16), we can write xV,(u,v) = xV′ ,(u,v) = x(u,v) .



Suppose (u, v) ∈ ΛFi with i ∈ [k−1]. By taking the images of ZF [k] ,Γ
∩V ⊂ (y = 0)

under ρF[k] , we obtain ZF [k−1] ,Γ
∩ V′ ⊂ (x(u,v) = 0). Hence, we have ZF[k−1] ,Γ ∩ V′ ⊂
(x(u,v) = 0) by Lemma 8.3 (3) for RF e=0′ by the maximality
. Therefore, x(u,v) ∈ Γ
[k−1] V
e=0′ . Then, by (8.7), ZF ,Γ ∩ V ⊂ (x(u,v) = 0).
of the subset Γ V [k]

Now suppose (u, v) ∈ ΛFk . Consider the relations

xV,u xV,v − xV,(u,v) xus xvs .


Fk ,o F k ,o
102 YI HU

Here, we have used x(us ,vs ) ≡ 1. Then, we have xV,u xV,v vanishes identically
F k ,o Fk ,o
† †
along ZF [k] ,Γ
, hence, so does one of xV,u and xV,v , that is, xu or xv ∈ Γ, since ZF [k] ,Γ

(birational to ZΓ ) is integral. In either case, it implies that ZF[k] ,Γ ⊂ (x(u,v) = 0) by


(8.3) and (8.4).
This proves the lemma when Fk is Γ-relevant.

Next, we suppose Fk is Γ-irrelevant.


Take an arbitrary standard chart V of RF[k] lying over a unique standard chart V′
of RF[k−1] such that ρ−1 ′
F[k] (ZF[k−1] ,Γ ) ∩ V 6= ∅ (equivalently, ZF[k−1] ∩ V 6= ∅). Then,

we have that
(ρ−1 ′
F[k] (ZF[k−1] ,Γ ) ∩ V)/(ZF[k−1] ,Γ ∩ V )
pre-q
is defined by the set of equations of LV,Fk and BV,[k] , all regarded as relations in
̺-variables of Fk . Thus, all these relations are linear in ̺-variables of Fk , according
pre-q
to Lemma 4.6. Putting them together, we call {LV,Fk , BV,[k] } a linear system in
̺-variables of Fk on the chart V/V′ .
We can let Λdet
Fk be the subset of ΛFk such that the minor corresponding to variables

{xV,(u,v) | (u, v) ∈ Λdet


Fk }

pre-q
achieves the maximal rank of the linear system {LV,Fk , BV,[k] } at any point of some
†◦ †
fixed Zariski open subset ZF [k−1] ,Γ
of ZF [k−1] ,Γ
. By shrinking if necessary, we may
†◦
assume that ZF [k−1] ,Γ
is contained in the intersection of all standard charts V′ with
ZF[k−1] ∩ V′ 6= ∅.
Then, we let

(8.8) Λ=1 det


Fk = ΛFk \ΛFk .

We then set and plug

(8.9) xV,(u,v) = 1, ∀ (u, v) ∈ Λ=1


Fk

pre-q
into the linear system {LV,Fk , BV,[k] } to obtain an induced linear system of full rank
†◦
over ZF [k−1] ,Γ
. This induced linear system can be solved over the Zariski open subset
†◦
ZF [k−1] ,Γ
such that all variables

{xV,(u,v) | (u, v) ∈ Λdet


Fk }
GRASSMANNIANS AND SINGULARITIES 103

are explicitly determined by the coefficients of the induced linear system.


We then let

(8.10) Λ=0 det


Fk ⊂ {x(u,v) | (u, v) ∈ ΛFk }

†◦
be the subset consisting of all the variables x(u,v) such that x(u,v) ≡ 0 along ZF [k−1] ,Γ
.
Observe here that we immediately obtain that for any (u, v) ∈ ΛFk ,
†◦
(8.11) x(u,v) vanishes identically along ZF [k−1] ,Γ
if and only if (u, v) ∈ Λ=0
Fk .

We let ZF[k] ,Γ be the scheme-theoretic intersection

(8.12) ρ−1 =0 =1
F[k] (ZF[k−1] ,Γ ) ∩ (x(u,v) = 0, (u, v) ∈ ΛFk ; x(u,v) = 1, (u, v) ∈ ΛFk ).

Further, for the above standard chart V of RF[k] , lying over the standard chart
V′ of RF[k−1] with ZeF[k] ,Γ ∩ V 6= ∅, we set

(8.13) e =0 = Γ
Γ e=0′ ⊔ Λ=0 ; e=1 = Γ
Γ e=1′ ⊔ Λ=1 .
V V Fk V V Fk

We are now ready to prove Lemma 8.3 (1), (2) and (3) in the case of RF[k] .
Similar to the proof of Lemma 8.3 (1) for the previous case when F̄k is Γ-relevant,
by Lemma 8.3 (1) in the case of RF[k−1] applied to ZF[k−1] ,Γ and ρ−1
F[k] (ZF[k−1] ,Γ ),

applying Proposition 4.18 to VF[k] ∩ V, and using (8.12) and (8.13), we obtain that
ZF[k] ,Γ ∩ V, as a closed subscheme of V, is defined by
e=1 ; Bpre-q ;
e =0 ; y − 1 y ∈ Γ
y, y ∈ Γ V V [k]

BV,(s,t) , s, t ∈ SFi with all i ∈ [k]



LV,Fi , with F̄i ∈ F[k−1];V′ ,Γ , LV,Fk ; F̄V,j , k < j ≤ Υ.

Then, we say (

F̄k ∈ F[k];V,Γ , if ΛFk 6= Λ=0 =1
Fk ∪ ΛFk ;

F̄k ∈
/ F[k]; m,Γ , if ΛFk = Λ=0 =1
Fk ∪ ΛFk .

Put it equivalently, F̄k ∈ F[k];V,Γ e=0 and y = 1 for all
if upon setting y = 0 for all y ∈ Γ V
e =1
y ∈ Γ , and plugging them into LV,F , we have that LV,F still contains a nontrivial
V k k

⋆ ⋆ ⋆
̺-variable of VarV ; F̄k ∈ / F[k];V,Γ , otherwise. We then set F[k];V,Γ = F[k−1];V′ ,Γ ∪{F̄k }

⋆ ⋆ ⋆ ⋆
if F̄k ∈ F[k];V,Γ ; F[k];V,Γ = F[k−1];V ′ ,Γ if F̄k ∈
/ F[k];V,Γ . Then, the above implies that
Lemma 8.3 (1) holds on RF[k] .
104 YI HU

Next, by construction, the induced morphism


†◦ †◦
ρ−1 =0 =1
F[k] (ZF[k] ,Γ ) ∩ (x(u,v) = 0, (u, v) ∈ ΛFk ; x(u,v) = 1, (u, v) ∈ ΛFk ) −→ ZF[k−1]


is an isomorphism. We let ZF [k] ,Γ
be the closure of
†◦
ρ−1 =0 =1
F[k] (ZF[k] ,Γ ) ∩ (x(u,v) = 0, (u, v) ∈ ΛFk ; x(u,v) = 1, (u, v) ∈ ΛFk )

in ZF[k] ,Γ . Then, it is closed in ZF[k] ,Γ and contains an open subset of ZF[k] ,Γ , hence,
is an irreducible component of ZF[k] ,Γ . It follows that the composition
† †
ZF [k] ,Γ
→ ZF [k−1]
→ ZΓ

is birational. This proves Lemma 8.3 (2) on RF[k] .



Finally, we are to prove Lemma 8.3 (3) on RF[k] . Suppose ZF [k] ,Γ
∩ V ⊂ (y = 0)
for some y ∈ VarV . If y = xV,u or y = xV,(u,v) with (u, v) ∈ ΛFi with i ∈ [k −1], then
the identical proof in the previous case carries over here without changes. We now

suppose ZF [k] ,Γ
∩ V ⊂ (xV,(u,v) = 0) with (u, v) ∈ ΛFk , then by (8.11), (u, v) ∈ Λ=0
Fk .

Thus, by (8.12), ZF[k] ,Γ ⊂ (x(u,v) = 0). This proves Lemma 8.3 (3) on RF[k] .
By induction, Lemma 8.3 is proved. 

Definition 8.4. We call ZF[k] ,Γ the F -transform of ZΓ in VF[k] for any k ∈ [Υ].

We keep the notation of Lemma 8.3. We set


⋆ ⋆ ⋆
(8.14) FV,Γ = F[Υ];V,Γ , LV,F[k];V,Γ
⋆ = {LV,F | F̄ ∈ F[k];V,Γ }, LV,Fm


= LV,F[Υ];V,Γ
⋆ .

8.3. ϑ-transforms of Γ-schemes in Veϑ[k] .


We now construct the ϑ-transform of ZΓ in Veϑ[k] ⊂ Reϑ[k] .

Lemma 8.5. Fix any subset Γ of Um . Assume that ZΓ is integral.


Fix any k ∈ [Υ].
Then, we have the following:
• there exists a closed subscheme Zeϑ[k] ,Γ of Veϑ[k] with an induced morphism
Zeϑ ,Γ → ZΓ ;
[k]

• Zeϑ[k] ,Γ comes equipped with an irreducible component Zeϑ† [k],Γ with the induced
morphism Ze† → ZΓ ;
ϑ[k] ,Γ
GRASSMANNIANS AND SINGULARITIES 105

• for any standard chart V of Reϑ[k] such that Zeϑ[k] ,Γ ∩ V 6= ∅, there are two
subsets, possibly empty,

e=0 ⊂ VarV ,
Γ e =1 ⊂ VarV .
Γ
V V

Further, consider any given standard chart V of Reϑ[k] with Zeϑ[k] ,Γ ∩ V 6= ∅. Then,
the following hold:

(1) the scheme Zeϑ[k] ,Γ ∩ V, as a closed subscheme of the chart V, is defined by


the following relations

(8.15) e =0 ,
y, y ∈ Γ V

y − 1, e =1
y∈Γ V ,

mn res q
BV , BV,>k , BV , LV,Fm


,

where LV,Fm
⋆ e=0 ⊂ VarV to be the maximal
is as in (8.14); further, we take Γ
,Γ V

subset (under inclusion) among all those subsets that satisfy the above;
(2) the induced morphism Zeϑ† [k] ,Γ → ZΓ is birational;
(3) for any variable y ∈ VarV , Ze† ∩ V ⊂ (y = 0) if and only if Zeϑ ,Γ ∩ V ⊂
ϑ[k] ,Γ [k]

(y = 0). Consequently, Zeϑ† [k] ,Γ ∩ V ⊂ Zeϑ[k+1] ∩ V if and only if Zeϑ[k] ,Γ ∩ V ⊂


Zeϑ ∩V where Zeϑ is the proper transform of the ϑ-center Zϑ
[k+1] [k]
in Reϑ .
[k+1] [k]

Proof. We prove by induction on k ∈ {0} ∪ [Υ].


The initial case is k = 0. In this case, we have

Reϑ[0] := RF , Veϑ[0] := VF , Zeϑ[0] ,Γ := ZF[Υ] ,Γ , Zeϑ† [0] ,Γ := ZF



[Υ] ,Γ
.

Then, in this case, Lemma 8.5 is Lemma 8.3 for k = Υ.


We now suppose that Lemma 8.5 holds over Reϑ[k−1] for some k ∈ [Υ].
We then consider the case of Reϑ . [k]

Suppose that Zeϑ[k−1] ,Γ , or equivalently Zeϑ† [k−1] ,Γ , by Lemma 8.5 (3) in the case of
Reϑ[k−1] , is not contained in Zϑ′ [k] where Zϑ′ [k] is the proper transform in Reϑ[k−1] of
the ϑ-center Zϑ (of Reϑ ). We then let Zeϑ ,Γ (respectively, Ze† ) be the proper
[k] [0] [k] ϑ[k] ,Γ

transform of Zeϑ[k−1] ,Γ (respectively, Zeϑ† [k−1] ,Γ ) in Veϑ[k] . As Zeϑ† [k] ,Γ is closed in Zeϑ[k] ,Γ
and contains a Zariski open subset of Zeϑ ,Γ, it is an irreducible component of Zeϑ ,Γ .
[k] [k]
106 YI HU

Further, consider any standard chart V of Reϑ[k] , lying over a unique standard
chart V′ of Reϑ , such that Zeϑ ,Γ ∩ V 6= ∅. We set
[k−1] [k]

e=0
Γ e=0
V = {yV | yV is the proper transform of some yV′ ∈ ΓV′ };

e=1 = {yV | yV is the proper transform of some yV′ ∈ Γ


Γ e=1′ }.
V V

We now prove Lemma 8.5 (1), (2) and (3) in the case of Reϑ[k] .
We can apply Lemma 8.5 (1) in the case of Reϑ[k−1] to Zeϑ[k−1] ,Γ to obtain the
defining equations of Zeϑ ′
,Γ ∩ V as stated in the lemma; we note here that these
[k−1]
res
equations include BV ′ ,≥k . We then take the proper transforms of these equations

in V to obtain the corresponding equations in V, and then apply (the proof of)
res
Proposition 5.12 to reduce BV,≥k res
to BV,>k . Because Zeϑ[k] Γ is the proper transform
of Zeϑ e
,Γ , this implies Lemma 8.5 (1) in the case of Rϑ .
[k−1] [k]

By construction, we have that the composition Zeϑ† [k] ,Γ → Zeϑ† [k−1] ,Γ → ZΓ is bira-
tional. This proves Lemma 8.5 (2) in the case of Reϑ[k] .
To show Lemma 8.5 (3) in Reϑ[k] , we fix any y ∈ VarV . It suffices to show that
if Ze† ∩ V ⊂ (y = 0), then Zeϑ ,Γ ∩ V ⊂ (y = 0). By construction, y 6= ζV,ϑ ,
ϑ[k] ,Γ [k] [k]

the exceptional variable in VarV corresponding to the ϑ-center Zϑ[k] . Hence, y is the
proper transform of some y ′ ∈ VarV′ . Then, by taking the images of Zeϑ† [k] ,Γ ∩ V ⊂
(y = 0) under the morphism ρϑ : Veϑ → Veϑ [k]
(which is induced from the
[k] [k−1]

blowup morphism πϑ[k] : Reϑ[k] → Reϑ[k−1] ), we obtain Zeϑ† [k−1] ,Γ ∩ V′ ⊂ (y ′ = 0), hence,
Zeϑ
[k−1]
′ ′ e
,Γ ∩ V ⊂ (y = 0) by the inductive assumption. Then, as Zϑ ,Γ is the proper [k]

transform of Zeϑ[k−1] ,Γ , we obtain Zeϑ[k] ,Γ ∩ V ⊂ (y = 0).


The last statement Lemma 8.5 (3) follows from the above because Zeϑ[k+1] ∩ V =
(y0 = y1 = 0) for some y0 , y1 ∈ VarV .

We now suppose that Zeϑ[k−1] ,Γ, or equivalently Zeϑ† [k−1] ,Γ , by Lemma 8.5 (3) in
Reϑ[k−1] , is contained in the proper transform Zϑ′ [k] of the ϑ-center Zϑ[k] .
Consider any standard chart V of Reϑ[k] , lying over a unique standard chart V′ of
Reϑ , such that Zeϑ ,Γ ∩ V 6= ∅.
[k−1] [k]
GRASSMANNIANS AND SINGULARITIES 107

We let ϑ′[k] be the proper transform in the chart V′ of the ϑ-set ϑ[k] . Then, ϑ′[k]
consists of two variables
ϑ′[k] = {y0′ , y1′ } ⊂ VarV′ .

We let P1[ξ0 ,ξ1 ] be the factor projective space for the ϑ-blowup Reϑ[k] → Reϑ[k−1] with
[ξ0 , ξ1 ] corresponding to (y0′ , y1′ ). Without loss of generality, we can assume that the
open chart V is given by

(V′ × (ξ0 ≡ 1)) ∩ Reϑ[k] ⊂ V′ × P1[ξ0 ,ξ1 ] .

We let ζV := ζV,ϑ[k] ∈ VarV be such that Eϑ[k] ∩ V = (ζV = 0) where Eϑ[k] is


the exceptional divisor of the blowup Reϑ[k] → Reϑ[k−1] . Note here that according
to the proof of Proposition 5.9, the variable y0′ corresponds to (or turns into) the
exceptional ζV on the chart V. We then let y1 (= ξ1 ) ∈ VarV be the proper transform
of y1′ ∈ VarV′ on the chart V.
In addition, we observe that

e=0′
ϑ′[k] = {y0′ , y1′ } ⊂ Γ V

because Zeϑ[k] ,Γ is contained in the proper transform Zϑ′ [k] of the ϑ-center Zϑ[k] .
We set,
=0
(8.16) e=0′ \ϑ′ },
ΓV = {ζV , yV | yV is the proper transform of some yV′ ∈ Γ V [k]

=1
(8.17) e=1′ }.
ΓV = { yV | yV is the proper transform of some yV′ ∈ Γ V

Consider the scheme-theoretic pre-image ρ−1 e e e


ϑ[k] (Zϑ[k−1] ,Γ ) where ρϑ[k] : Vϑ[k] → Vϑ[k−1]

is induced from the blowup morphism πϑ[k] : Reϑ[k] → Reϑ[k−1] .


Note that scheme-theoretically, we have,

ρ−1 e −1 e e
ϑ[k] (Zϑ[k−1] ,Γ ) ∩ V = πϑ[k] (Zϑ[k−1] ,Γ ) ∩ Vϑ[k] ∩ V.

Applying Lemma 8.5 (1) in Reϑ[k−1] to Zeϑ[k−1] ,Γ and ρ−1 e


ϑ[k] (Zϑ[k−1] ,Γ ), and applying

Proposition 5.12 to Veϑ[k] ∩V, we obtain that ρ−1 e


ϑ[k] (Zϑ[k−1] ,Γ )∩V, as a closed subscheme

of V, is defined by
=0 mn =1res q
(8.18) yV ∈ ΓV ; yV − 1, yV ∈ ΓV ; BV ; BV,>k ; BV ; LV,Fm


.
108 YI HU

=0
(Observe here that ζV ∈ ΓV .)
=0 =1
Thus, by setting yV = 0 for all yV ∈ ΓV and yV = 1 for all yV ∈ ΓV in
mn res q
BV , BV,>k , BV , LV,Fm


of the above, we obtain
mn res q
(8.19) B̃V , B̃V,>k , B̃V , L̃V,Fm


.

Note that for any F̄ ∈ FV,Γ (cf. (8.14)), if LV,F contains y1 , then it contains ζV ,
hence L̃V,F does not contain y1 . We keep those equations of (8.19) that contain the
variable y1 and obtain
mn res q
(8.20) B̂V , B̂V,>k , B̂V ,

viewed as a system of equations in y1 . By Proposition 5.12 (the last two statements),


one sees that (8.20) is a linear system of equations in y1 . Furthermore, we have that

(ρ−1 e e ′
ϑ[k] (Zϑ[k−1] ,Γ ) ∩ V)/(Zϑ[k−1] ,Γ ∩ V )

is defined by the linear system (8.20).


There are the following two cases for (8.20):
(⋆a) the rank of the linear system (8.20) equals one over general points of Zeϑ† [k−1] ,Γ .
(⋆b) the rank of the linear system (8.20) equals zero at general points of Ze† ,
ϑ[k−1] ,Γ

hence at all points of Zeϑ† [k−1] ,Γ .

Proof of Lemma 8.5 for Reϑ[k] under the condition (⋆a).

By the condition (⋆a), there exists a Zariski open subset Zeϑ†◦[k−1] ,Γ of Zeϑ† [k−1] ,Γ such
that the rank of the linear system (8.20) equals one at any point of Ze†◦ . By
ϑ[k−1] ,Γ

solving y1 from the linear system (8.20) over Zeϑ†◦[k−1] ,Γ , we obtain that the induced
morphism
ρ−1 e†◦ e◦
ϑ[k] (Zϑ[k−1] ,Γ ) −→ Zϑ[k−1] ,Γ

is an isomorphism.
First, we suppose y1 is identically zero along ρ−1 e†◦
ϑ[k] (Zϑ[k−1] ,Γ ). We then set,

(8.21) e=0 = {y1 } ∪ Γ=0


Γ V V

=0
where ΓV is as in (8.16). In this case, we let

(8.22) Zeϑ[k] ,Γ = ρ−1 e


ϑ[k] (Zϑ[k−1] ,Γ ) ∩ Dy1
GRASSMANNIANS AND SINGULARITIES 109

scheme-theoretically, where Dy1 is the closure of (y1 = 0) in Reϑ[k] . We remark here


that Dy1 does not depend on the choice of the chart V.
Next, suppose y1 is not identically zero along ρ−1 (Ze†◦ ϑ[k] ϑ[k−1] ,Γ ). We then set,
=0
(8.23) e=0
Γ V = ΓV

=0
where ΓV is as in (8.16). In this case, we let

(8.24) Zeϑ[k] ,Γ = ρ−1 e


ϑ[k] (Zϑ[k−1] ,Γ ).

We always set (under the condition (⋆a))

(8.25) e=1 = Γ=1


Γ V V

=1
where ΓV is as in (8.17).
In each case, by construction, we have

ρ−1 e†◦ e
ϑ[k] (Zϑ[k−1] ,Γ ) ⊂ Zϑ[k] Γ ,

and we let Zeϑ† [k] ,Γ be the closure of ρ−1 e†◦ e


(ϑ[k] (Zϑ[k−1] ,Γ ) in Zϑ[k] ,Γ . It is an irreducible
component of Zeϑ[k] ,Γ because Zeϑ† [k] ,Γ is closed in Zeϑ[k] ,Γ and contains the Zariski open
subset ρ−1 (Ze†◦
ϑ[k] ) of Zeϑ ,Γ . Then, we obtain that the composition
ϑ[k−1] ,Γ [k]

Zeϑ† [k] ,Γ −→ Zeϑ† [k−1] ,Γ −→ ZΓ

is birational. This proves Lemma 8.5 (2) over Reϑ[k] .


In each case of the above (i.e., (8.21) and (8.23)), by the paragraph of (8.18),
one sees that Zeϑ ,Γ ∩ V, as a closed subscheme of V, is defined by the equations as
[k]

stated in the Lemma. This proves Lemma 8.5 (1) over Reϑ[k] .
It remains to prove Lemma 8.5 (3) over Reϑ[k] .
Fix any y ∈ VarV , it suffices to show that if Zeϑ† [k] ,Γ ∩V ⊂ (y = 0), then Ze(ϑ[k] ,Γ ∩V ⊂
(y = 0). If y 6= ζV , y1, then y is the proper transform of some variable y ′ ∈ VarV′ .
Hence, by taking the images under ρϑ[k] , we have Zeϑ† [k−1] ,Γ ∩V′ ⊂ (y ′ = 0); by Lemma
8.5 (3) in Reϑ , we obtain Zeϑ
[k−1]

[k−1]
′ ′ e=0
,Γ ∩V ⊂ (y = 0), thus y ∈ Γ ′ by the maximality V
e=0′ . Therefore, Zeϑ ,Γ ∩ V ⊂ (y = 0), by (the already-proved) Lemma
of the subset Γ V [k]

e
8.5 (1) for Rϑ (cf. (8.16) and (8.21) or (8.23)). Next, suppose y = y1 (if it occurs).
[k]
110 YI HU

Then, by construction, Zeϑ[k] ,Γ ∩ V ⊂ (y = 0). Finally, we let y = ζV . Again, by


construction, Zeϑ ,Γ ∩ V ⊂ (ζV = 0).
[k]

As earlier, the last statement Lemma 8.5 (3) follows from the above.

Proof of Lemma 8.5 over Reϑ[k] under the condition (⋆b).

Under the condition (⋆b), we have that

ρ−1 e† e†
ϑ[k] (Zϑ[k−1] ,Γ ) −→ Zϑ[k−1] ,Γ

can be canonically identified with the trivial P[ξ0 ,ξ1 ] -bundle:

ρ−1 e† e†
ϑ[k] (Z(ϑ[k−1] ,Γ ) = Zϑ[k−1] ,Γ × P[ξ0 ,ξ1 ] .

In this case, we define

Zeϑ[k] ,Γ = ρ−1 e
ϑ[k] (Zϑ[k−1] ,Γ ) ∩ ((ξ0 , ξ1 ) = (1, 1)),

Zeϑ† [k] ,Γ = ρ−1 e†


ϑ[k] (Zϑ[k−1] ,Γ ) ∩ ((ξ0 , ξ1 ) = (1, 1)),

both scheme-theoretically. The induced morphism Zeϑ† [k] ,Γ −→ Zeϑ† [k−1] ,Γ is an isomor-
phism. Again, one sees that Zeϑ† [k] ,Γ is an irreducible component of Zeϑ[k] ,Γ . Therefore,

Zeϑ† [k] ,Γ −→ Zeϑ† [k−1] ,Γ −→ ZΓ

is birational. This proves Lemma 8.7 (2) over Reϑ[k] .


Further, under the condition (⋆b), we set
e =0 = Γ=0 ,
Γ e=1 = {y1 } ∪ Γ=1 .
Γ
V V V V

Then, again, by the paragraph of (8.18), one sees that Zeϑ[k],Γ ∩ V, as a closed
subscheme of V, is defined by the equations as stated in the Lemma. This proves
Lemma 8.5 (1) over Reϑ . [k]

It remains to prove Lemma 8.5 (3) in over Reϑ[k] .


Fix any y ∈ VarV , it suffices to show that if Zeϑ† [k] ,Γ ∩V ⊂ (y = 0), then Zeϑ[k],Γ ∩V ⊂
(y = 0). By construction, y 6= y1 . Then, the corresponding proof of Lemma 8.5
(3) for Reϑ[k] under the condition (⋆a) goes through here without change. The last
statement Lemma 8.5 (3) follows from the above. This proves Lemma 8.5 (3) in
Reϑ under the condition (⋆b).
[k]

This completes the proof of Lemma 8.5. 


GRASSMANNIANS AND SINGULARITIES 111

Definition 8.6. We call Zeϑ[k] ,Γ the ϑ-transform of ZΓ in Veϑ[k] for any k ∈ [Υ].

We need the final case of Lemma 8.5. We set

Zeϑ,Γ := Zeϑ[Υ] ,Γ , Zeϑ,Γ



:= Zeϑ† [Υ],Γ .

8.4. ℘-transforms of Γ-schemes in Ve(℘(kτ ) rµ sh ) .


We now construct the ℘-transform of ZΓ in Ve(℘ (kτ ) rµ sh )
⊂ Re(℘(kτ ) rµ sh ) .

Lemma 8.7. Fix any subset Γ of Um . Assume that ZΓ is integral.


Consider (kτ )µh ∈ IndexΦ .
Then, we have the following:

• there exists a closed subscheme Ze(℘(kτ ) rµ sh ),Γ of Ve(℘(kτ ) rµ sh ) with an induced


morphism Ze(℘(kτ ) rµ sh ),Γ → ZΓ ;
• Ze(℘(kτ ) rµ sh ),Γ comes equipped with an irreducible component Ze(℘ †
(kτ ) rµ sh ),Γ
with
the induced morphism Ze† → ZΓ ;
(℘(kτ ) rµ sh ),Γ

• for any standard chart V of Re(℘(kτ ) rµ sh ) such that Ze(℘(kτ ) rµ sh ),Γ ∩ V 6= ∅, there
come equipped with two subsets, possibly empty,

e=0 ⊂ VarV ,
Γ e =1 ⊂ VarV .
Γ
V V

Further, consider any given chart V of Re(℘(kτ ) rµ sh ) with Ze(℘(kτ ) rµ sh ),Γ ∩ V 6= ∅.


Then, the following hold:

(1) the scheme Ze(℘(kτ ) rµ sh ),Γ ∩ V, as a closed subscheme of the chart V, is defined
by the following relations

(8.26) e=0 ,
y, y ∈ Γ V

e=1 ,
y − 1, y ∈ Γ V

mn q
BV , BV , LV,Fm


;

e=0 ⊂ VarV to be the maximal subset (under inclusion)


further, we take Γ V

among all those subsets that satisfy the above.


(2) the induced morphism Ze(℘

(kτ ) rµ sh ),Γ
→ ZΓ is birational;
112 YI HU

(3) for any variable y ∈ VarV , Ze(℘ †


(kτ ) rµ sh ),Γ
∩ V ⊂ (y = 0) if and only if
Ze(℘(kτ ) rµ sh ),Γ ∩ V ⊂ (y = 0). Consequently, Ze(℘ †
(kτ ) rµ sh ),Γ
∩ V ⊂ Zeφ(kτ )µ(h+1) ∩ V
if and only if Ze(℘ rµ s ),Γ ∩V ⊂ Zeφ
(kτ ) h
∩V where Zeφ
(kτ )µ(h+1)
is the proper
(kτ )µ(h+1)

transform of the ℘-center Zφ(kτ )µ(h+1) in Re(℘(kτ ) rµ sh ) .

Proof. We prove by induction on (kτ )µh ∈ {(11)10)} ⊔ IndexΦ (cf. (6.2)).


The initial case is (11)10. In this case, we have

Re(℘(11) r1 s0 ) := Reϑ , Ve(℘(11) r1 s0 ) := Veϑ , Ze(℘(11) r1 s0 ),Γ := Zeϑ,Γ , Ze(℘



(11) r1 s0 ),Γ
:= Zeϑ,Γ

.

Then, in this case, Lemma 8.7 is Lemma 8.5 for k = Υ.


We now suppose that Lemma 8.7 holds for (kτ )µ(h−1) for some (kτ )µh ∈ IndexΦ .
We then consider the case of (kτ )µh.
We let

(8.27) ρ(℘(kτ ) rµ sh ) : Ve(℘(kτ ) rµ sh ) −→ Ve(℘(kτ ) rµ sh−1 )

be the morphism induced from π(℘(kτ ) rµ sh ) : Re(℘(kτ ) rµ sh ) → Re(℘(kτ ) rµ sh−1 ) .


Suppose that Ze(℘(kτ ) rµ sh−1 ),Γ , or equivalently Ze(℘

(kτ ) rµ sh−1 ),Γ
, by Lemma 8.7 (3) in
(℘(kτ ) rµ s(h−1) ), is not contained in Zφ′ (kτ )µh where Zφ′ (kτ )µh is the proper transform
in Re(℘(kτ ) rµ sh−1 ) of the ℘-center Zφ(kτ )µh (of Re(℘(kτ ) rµ−1 ) ). We then let Ze(℘(kτ ) rµ sh ),Γ
(resp. Ze(℘ †
(kτ ) rµ sh ),Γ
) be the proper transform of Ze(℘(kτ ) rµ sh−1 ),Γ (resp. Ze(℘

(kτ ) rµ sh−1 ),Γ
)
in V(℘ rµ s ) . As Ze†
(kτ ) h
is closed in Ze(℘ rµ s ),Γ and contains a Zariski open
(℘(kτ ) rµ sh ),Γ (kτ ) h

subset of Ze(℘(kτ ) rµ sh ),Γ , it is an irreducible component of Ze(℘(kτ ) rµ sh ),Γ .


Further, consider any standard chart V of Re(℘(kτ ) rµ sh ) , lying over a unique standard
chart V′ of Re(℘ rµ s ) , such that Ze(℘ rµ s ),Γ ∩ V 6= ∅. We set
(kτ ) h−1 (kτ ) h

e=0 = {yV | yV is the proper transform of some yV′ ∈ Γ


Γ e=0′ };
V V

e=1 = {yV | yV is the proper transform of some yV′ ∈ Γ


Γ e=1′ }.
V V

We now prove Lemma 8.7 (1), (2) and (3) in (℘(kτ ) rµ sh ).


Lemma 8.7 (1) in (℘(kτ ) rµ sh ) follows from Lemma 8.7 (1) in (℘(kτ ) rµ sh−1 ) because
Ze(℘ rµ s ),Γ is the proper transform of Ze(℘ rµ s ),Γ .
(kτ ) h (kτ ) h−1

By construction, we have that Ze(℘



(kτ ) rµ sh ),Γ
→ Ze(℘

(kτ ) rµ sh−1 ),Γ
→ ZΓ is birational.
This proves Lemma 8.7 (2) in (℘(kτ ) rµ sh ).
GRASSMANNIANS AND SINGULARITIES 113

To show Lemma 8.7 (3) in (℘(kτ ) rµ sh ), we fix any y ∈ VarV . It suffices to show that
if Ze† ∩ V ⊂ (y = 0), then Ze(℘ rµ s ),Γ ∩ V ⊂ (y = 0). By construction, y 6=
(℘(kτ ) rµ sh ),Γ (kτ ) h

ζV,(kτ )µh , the exceptional variable in VarV corresponding to the ℘-set φ(kτ )µh . Hence,
y is the proper transform of some y ′ ∈ VarV′ . Then, by taking the images under
the morphism ρ(℘(kτ ) rµ sh ) of (8.27), we obtain Ze(℘

(kτ ) rµ sh−1 ),Γ
∩ V′ ⊂ (y ′ = 0), hence,
Ze(℘ rµ s ),Γ ∩ V′ ⊂ (y ′ = 0) by the inductive assumption. Then, as Ze(℘ rµ s ),Γ
(kτ ) h−1 (kτ ) h

is the proper transform of Ze(℘(kτ ) rµ sh−1 ),Γ , we obtain Ze(℘(kτ ) sh ),Γ ∩ V ⊂ (y = 0). The
last statement of Lemma 8.7 (3) follows from the above because Zeφ(kτ )µ(h+1) ∩ V =
(y0 = y1 = 0) for some y0 , y1 ∈ VarV .

We now suppose that Ze(℘(kτ ) rµ sh−1 ),Γ , or equivalently Ze(℘



(kτ ) rµ sh−1 ),Γ
, by Lemma 8.7
(3) in (℘(kτ ) rµ s(h−1) ), is contained in the proper transform Zφ′ (kτ )µh of Zφ(kτ )µh .
Consider any standard chart V of Re(℘(kτ ) rµ sh ) , lying over a unique standard chart
V′ of Re(℘ rµ s ) , such that Ze(℘ rµ s ),Γ ∩ V′ 6= ∅.
(kτ ) h−1 (kτ ) h−1

We let φ′(kτ )µh be the proper transform in the chart V′ of the ℘-set φ(kτ )µh . Then,
φ′(kτ )µh consists of two variables such that

′ ′ ′
ψ(kτ )µh = {y0 , y1 } ⊂ VarV′ .

In addition, we let ζV ∈ VarV be such that E(℘(kτ ) rµ sh ) ∩ V = (ζV = 0) where


E(℘(kτ ) rµ sh ) is the exceptional divisor of the blowup Re(℘(kτ ) rµ sh ) → Re(℘(kτ ) rµ sh−1 ) . With-
out loss of generality, we may assume that y0′ corresponds to the exceptional variable
ζV on the chart V. We then let y1 ∈ VarV be the proper transform of y1′ .
Now, we observe that
e=0′
φ′(kτ )µh ⊂ Γ V

because Ze(℘(kτ ) rµ sh−1 ),Γ is contained in the proper transform Zφ′ (kτ )µh .
We set,
=0
e=0′ \φ′
ΓV = {ζV , yV | yV is the proper transform of some yV′ ∈ Γ
(8.28) V (kτ )µh },

=1
(8.29) e=1′ }.
ΓV = { yV | yV is the proper transform of some yV′ ∈ Γ V

Consider the scheme-theoretic pre-image ρ−1 e


(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ).
114 YI HU

Note that scheme-theoretically, we have

ρ−1 e −1 e e
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ∩ V = π(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ∩ V(℘(kτ ) rµ sh ) ∩ V.

Applying Lemma 8.7 (1) in (℘(kτ ) rµ sh−1 ) to Ze(℘(kτ ) rµ sh−1 ),Γ and ρ−1 e
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ),

and applying Proposition 6.9 to Ve(℘ r s ) ∩ V, we obtain that the pre-image


(kτ ) µ h

ρ−1 e ∩ V, as a closed subscheme of V, is defined by


(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ )

=0 mn q =1
(8.30) yV ∈ ΓV ; yV − 1, yV ∈ ΓV ; BV ; BV ; LV,Fm


.
=0
(Observe here that ζV ∈ ΓV .)
Thus, by setting
=0 =1
yV = 0 for all yV ∈ ΓV and yV = 1 for all yV ∈ ΓV

mn q
in BV , BV , LV,Fm


of the above, we obtain

mn q
(8.31) B̃V , B̃V , L̃V,Fm


.


Note that for any F̄ ∈ Fm ,Γ , if a term of LV,F contains y1 ∈ VarV , then it contains

ζV y1 , hence L̃V,F does not contain y1 . We keep those equations of (8.31) such that
they contain the variable y1 ∈ VarV and obtain

mn q
(8.32) B̂V , B̂V ,

viewed as a system of equations in y1 . Then, by Proposition 6.9 (1) and (2), (8.32)
is a linear system of equations in y1 . (We point out that y1 here can correspond to
either y0 or y1 as in Proposition 6.9.) Furthermore, one sees that

(ρ−1 e e ′
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ∩ V)/(Z(℘(kτ ) rµ sh−1 ),Γ ∩ V )

is defined by the linear system (8.32).


There are the following two cases for (8.32):
(⋆a) the ranks of the linear system (8.32) equal one at general points of Ze(℘

(kτ ) rµ sh−1 ),Γ
.
(⋆b) the ranks of the linear system (8.32) equal zero at general points of Ze(℘

(kτ ) rµ sh−1 ),Γ
,
hence at all points of Ze †
.
(℘(kτ ) rµ sh−1 ),Γ

Proof of Lemma 8.7 in (℘(kτ ) rµ sh ) under the condition (⋆a).


GRASSMANNIANS AND SINGULARITIES 115

By the condition (⋆a), there exists a Zariski open subset Ze(℘


†◦
(kτ ) rµ sh−1 ),Γ
of Ze(℘

(kτ ) rµ sh−1 ),Γ

such that the rank of the linear system (8.32) equals one at any point of Ze †◦
. (℘(kτ ) rµ sh−1 ),Γ
By solving y1 from the linear system (8.32) over Ze(℘
†◦
(kτ ) rµ sh−1 ),Γ
, we obtain that the
induced morphism

ρ−1 e†◦ e◦
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) −→ Z(℘(kτ ) rµ sh−1 ),Γ

is an isomorphism.
Suppose y1 is identically zero along ρ−1 e†◦
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ). We then set,

(8.33) e=0 = {y1 } ∪ Γ=0


Γ V V

=0
where ΓV is as in (8.28). In this case, we let

(8.34) Ze(℘(kτ ) rµ sh ),Γ = ρ−1 e


(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ∩ Dy1

scheme-theoretically, where Dy1 is the closure of (y1 = 0) in Re(℘(kτ ) rµ sh ) . We remark


here that Dy1 does not depend on the choice of the chart V.
Suppose y1 is not identically zero along ρ−1 e†◦
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ). We then set,

(8.35) e=0 = Γ=0


Γ V V

=0
where ΓV is as in (8.28) . In this case, we let

(8.36) Ze(℘(kτ ) rµ sh ),Γ = ρ−1 e


(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ).

We always set (under the condition (⋆a))

(8.37) e=1 = Γ=1


Γ V V

=1
where ΓV is as in (8.29).
In each case, we have

ρ−1 e†◦ e
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ⊂ Z(℘(kτ ) rµ sh ),Γ ,

and we let Ze(℘



(kτ ) rµ sh ),Γ
be the closure of ρ−1 e†◦ e
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) in Z(℘(kτ ) rµ sh ),Γ . It
is an irreducible component of Ze(℘ rµ s ),Γ because Ze†
(kτ ) h
is closed in Ze(℘ rµ s ),Γ
(℘(kτ ) rµ sh ),Γ (kτ ) h
116 YI HU

and contains the Zariski open subset ρ−1 e†◦ e


(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) of Z(℘(kτ ) rµ sh ),Γ . Then,
it follows that the composition

Ze(℘

(kτ ) rµ sh ),Γ
−→ Ze(℘

(kτ ) rµ sh−1 ),Γ
−→ ZΓ

is birational. This proves Lemma 8.7 (2) in (℘(kτ ) sh ).


In each case of (8.33) and (8.35), by the paragraph of (8.30), we have that
Ze(℘(kτ ) rµ sh ),Γ ∩ V as a closed subscheme of V is defined by the equations as stated in
the Lemma. This proves Lemma 8.7 (1) in (℘(kτ ) rµ sh ).
It remains to prove Lemma 8.7 (3) in (℘(kτ ) rµ sh ).
Fix any y ∈ VarV , it suffices to show that if Ze† (℘(kτ ) rµ sh ),Γ ∩ V ⊂ (y = 0), then
Ze(℘(kτ ) rµ sh ),Γ ∩ V ⊂ (y = 0). If y 6= ζV , y1 , then y is the proper transform of
some variable y ′ ∈ VarV′ . Hence, by taking the images under ρ(℘(kτ ) rµ sh ) , we ob-
tain Ze(℘ †
(kτ ) rµ sh−1 ),Γ
∩ V′ ⊂ (y ′ = 0), and then, by Lemma 8.7 (3) in (℘(kτ ) rµ sh−1 ),
Ze(℘(kτ ) rµ sh−1 ),Γ ∩ V′ ⊂ (y ′ = 0), thus y ′ ∈ Γ e=0′ by the maximality of Γ
V
e=0′ . Therefore,
V
e
Z(℘ rµ s ),Γ ∩ V ⊂ (y = 0), by (the already-proved) Lemma 8.7 (1) in (℘(kτ ) rµ sh ).
(kτ ) h

Next, suppose y = y1 (if it occurs). Then, by construction, Ze(℘(kτ ) sh ),Γ ∩V ⊂ (y = 0).


Finally, we let y = ζV. Again, by construction, Ze(℘ rµ s ),Γ ∩ V ⊂ (ζV = 0). As in
(kτ ) h

the previous case, the last statement of Lemma 8.7 (3) follows from the above.

Proof of Lemma 8.7 in (℘(kτ ) rµ sh ) under the condition (⋆b).

Under the condition (⋆b), we have that

ρ−1 e† e†
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) −→ Z(℘(kτ ) rµ sh−1 ),Γ

can be canonically identified with the trivial P[ξ0 ,ξ1 ] -bundle:

ρ−1 e† e†
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) = Z(℘(kτ ) rµ sh−1 ),Γ × P[ξ0 ,ξ1 ] .

In this case, we define

Ze(℘(kτ ) rµ sh ),Γ = ρ−1 e


(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ∩ ((ξ0 , ξ1 ) = (1, 1)),

Ze(℘

(kτ ) rµ sh ),Γ
:= ρ−1 e†
(℘(kτ ) rµ sh ) (Z(℘(kτ ) rµ sh−1 ),Γ ) ∩ ((ξ0 , ξ1 ) = (1, 1)),

both scheme-theoretically.
GRASSMANNIANS AND SINGULARITIES 117

The induced morphism Ze(℘ †


(kτ ) rµ sh ),Γ
−→ Ze(℘

(kτ ) rµ sh−1 ),Γ
is an isomorphism. Again,
one sees that Ze(℘

(kτ ) rµ sh ),Γ
is an irreducible component of Ze(℘(kτ ) rµ sh ),Γ . Therefore,

Ze(℘

(kτ ) rµ sh ),Γ
−→ Ze(℘

(kτ ) rµ sh−1 ),Γ
−→ ZΓ

is birational. This proves Lemma 8.7 (2) in (℘(kτ ) rµ sh ).


Further, under the condition (⋆b), we set

e =0 = Γ=0 ,
Γ e=1 = {y1 } ∪ Γ=1 .
Γ
V V V V

Then, by the paragraph of (8.30), we have that Ze(℘(kτ ) rµ sh ),Γ ∩ V, as a closed


subscheme of V, is defined by the equations as stated in the Lemma. This proves
Lemma 8.7 (1) in (℘(kτ ) rµ sh ).
It remains to prove Lemma 8.7 (3) in (℘(kτ ) rµ sh ).
Fix any y ∈ VarV , it suffices to show that if Ze(℘

(kτ ) sh ),Γ
∩ V ⊂ (y = 0), then
Ze(℘(kτ ) sh ),Γ ∩ V ⊂ (y = 0). By construction, y 6= y1 . Then, the corresponding proof
of Lemma 8.7 (3) in (℘(kτ ) rµ sh ) under the condition (⋆a) goes through here without
change. As earlier, the last statement of Lemma 8.7 (3) follows from the above.
This proves Lemma 8.7 (3) in (℘(kτ ) rµ sh ) under the condition (⋆b).
Putting all together, this completes the proof of Lemma 8.7. 

Definition 8.8. We call Ze(℘(kτ ) rµ sh ),Γ the ℘-transform of ZΓ in Ve(℘(kτ ) rµ sh ) for any
(kτ )µh ∈ IndexΦ .

We need the final case of Lemma 8.7. We set

Ze℘,Γ := Ze(℘(kτ ) rµ sh ),Γ , Ze℘,Γ



:= Ze(℘

(kτ ) rµ sh ),Γ

where k = Υ, τ = tFΥ , µ = ρΥtFΥ , h = σ(ΥtFΥ )ρΥtF . That is,


Υ

Ze℘,Γ = Ze(℘(ΥtF ) rρΥt sσ(Υt ),Γ , Ze℘,Γ



= Ze(℘

r sσ(Υt ),Γ .
Υ FΥ FΥ )ρΥt

(Υt FΥ ) ρΥtF FΥ )ρΥtF
Υ Υ

8.5. ð-transforms of Γ-schemes in Ve(ð(kτ ) rµ sh ) .


We now construct the ð-transform of ZΓ in Ve(ð (kτ ) rµ sh )
⊂ Re(ð(kτ ) rµ sh ) . (This sub-
section may be combined with the previous subsection to save some space, using the
notation of §7.4. But, we choose to exhibit it separately for clarity.)
118 YI HU

Lemma 8.9. Fix any subset Γ of Um . Assume that ZΓ is integral.


Consider (kτ )µh ∈ IndexΨ .
Then, we have the following:
• there exists a closed subscheme Ze(ð(kτ ) rµ sh ),Γ of Ve(ð(kτ ) rµ sh ) with an induced
morphism Ze(ð(kτ ) rµ sh ),Γ → ZΓ ;
• Ze(ð(kτ ) rµ sh ),Γ comes equipped with an irreducible component Ze(ð †
(kτ ) rµ sh ),Γ
with
the induced morphism Ze† → ZΓ ;
(ð(kτ ) rµ sh ),Γ

• for any standard chart V of Re(ð(kτ ) rµ sh ) such that Ze(ð(kτ ) rµ sh ),Γ ∩ V 6= ∅, there
are two subsets, possibly empty,
e=0 ⊂ VarV ,
Γ e =1 ⊂ VarV .
Γ
V V

Further, consider any given chart V of Re(ð(kτ ) rµ sh ) with Ze(ð(kτ ) rµ sh ),Γ ∩ V 6= ∅.


Then, the following hold:
(1) the scheme Ze(ð(kτ ) rµ sh ),Γ ∩ V, as a closed subscheme of the chart V, is defined
by the following relations

(8.38) e=0 ,
y, y ∈ Γ V

e=1 ,
y − 1, y ∈ Γ V

mn res q
BV , BV , BV , LV,Fm


;

e=0 ⊂ VarV to be the maximal subset (under inclusion)


further, we take Γ V

among all those subsets that satisfy the above.


(2) the induced morphism Ze(ð †
(kτ ) rµ sh ),Γ
→ ZΓ is birational;
(3) for any variable y ∈ VarV , Ze(ð(kτ ) rµ sh ),Γ ∩ V ⊂ (y = 0) if and only if

Ze(ð(kτ ) rµ sh ),Γ ∩ V ⊂ (y = 0). Consequently, Ze(ð †


(kτ ) rµ sh ),Γ
∩ V ⊂ Zeψ(kτ )µ(h+1) ∩ V
if and only if Ze(ð rµ s ),Γ ∩V ⊂ Zeψ
(kτ ) h
∩V where Zeψ
(kτ )µ(h+1)
is the proper
(kτ )µ(h+1)

transform in Re(ð(kτ ) rµ sh ) of the ð-center Zψ(kτ )µ(h+1) .

Proof. We prove by induction on (kτ )µh ∈ {(11)10} ⊔IndexΨ (cf. (7.2)). (The proof
is analogous to that of Lemma 8.7. We provide details for completeness.)
The initial case is (11)10. In this case, we have

Re(ð11 r1 s0 ) := Re℘ , Ve(ð11 r1 s0 ) := Ve℘ , Ze(ð11 r1 s0 ),Γ := Ze℘,Γ, Ze(ð



11 r1 s0 ),Γ
:= Ze℘,Γ

.
GRASSMANNIANS AND SINGULARITIES 119

Then, in this case, Lemma 8.9 is Lemma 8.7 for Re℘ .


We now suppose that Lemma 8.9 holds for (kτ )µ(h−1) for some (kτ )µh ∈ IndexΨ .
We then consider the case of (kτ )µh ∈ IndexΨ .
We let

(8.39) ρ(ð(kτ ) rµ sh ) : Ve(ð(kτ ) rµ sh ) −→ Ve(ð(kτ ) rµ sh−1 )

be the morphism induced from π(ð(kτ ) rµ sh ) : Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 ) .


Suppose that Ze(ð(kτ ) rµ sh−1 ),Γ , or equivalently Ze(ð

(kτ ) rµ sh−1 ),Γ
, by Lemma 8.9 (3) in
(ð(kτ ) rµ s(h−1) ), is not contained in Zψ′ (kτ )h where Zψ′ (kτ )µh is the proper transform in
Re(ð(kτ ) rµ sh−1 ) of the ð-center Zψ(kτ )µh (of Re(ð(kτ ) rµ−1 ) ).
We then let Ze(ð(kτ ) rµ sh ),Γ (resp. Ze(ð

(kτ ) rµ sh ),Γ
) be the proper transform of Ze(ð(kτ ) rµ sh−1 ),Γ
(resp. Ze† ) in Ve(ð rµ s ) . As Ze†
(ð(kτ ) rµ sh−1 ),Γ (kτ ) h
is closed in Ze(ð rµ s ),Γ and
(ð(kτ ) rµ sh ),Γ (kτ ) h

contains a Zariski open subset of Ze(ð(kτ ) rµ sh ),Γ , it is an irreducible component of


Ze(ð rµ s ),Γ .
(kτ ) h

Further, consider any standard chart V of Re(ð(kτ ) rµ sh ) , lying over a unique standard
chart V′ of Re(ð rµ s ) , such that Ze(ð rµ s ),Γ ∩ V 6= ∅. We set
(kτ ) h−1 (kτ ) h

e=0 = {yV | yV is the proper transform of some yV′ ∈ Γ


Γ e=0′ };
V V

e=1 = {yV | yV is the proper transform of some yV′ ∈ Γ


Γ e=1′ }.
V V

We now prove Lemma 8.9 (1), (2) and (3) in (ð(kτ ) rµ sh ).


Lemma 8.9 (1) in (ð(kτ ) rµ sh ) follows from Lemma 8.9 (1) in (ð(kτ ) rµ sh−1 ) because
Ze(ð rµ s ),Γ is the proper transform of Ze(ð rµ s ),Γ .
(kτ ) h (kτ ) h−1

By construction, we have that the composition Ze(ð



(kτ ) rµ sh ),Γ
→ Ze(ð

(kτ ) rµ sh−1 ),Γ
→ ZΓ
is birational. This proves Lemma 8.9 (2) in (ð(kτ ) rµ sh ).
It remains to prove Lemma 8.9 (3) in (ð(kτ ) rµ sh ). But, it follows from the identical
lines of the corresponding proof of Lemma 8.7 (3). We avoid the repetition.

We now suppose that Ze(ð(kτ ) rµ sh−1 ),Γ , or equivalently Ze(ð



(kτ ) rµ sh−1 ),Γ
, by Lemma 8.9
(3) in (ð(kτ ) rµ s(h−1) ), is contained in the proper transform Zψ′ (kτ )µh of Zψ(kτ )µh .
Consider any standard chart V of Re(ð(kτ ) rµ sh ) , lying over a unique standard chart
V′ of Re(ð rµ s ) , such that Ze(ð rµ s ),Γ ∩ V′ 6= ∅.
(kτ ) h−1 (kτ ) h−1
120 YI HU

′ ′
We let ψ(kτ )µh be the proper transform in the chart V of the ð-set ψ(kτ )µh . Then,

we can express
′ ′ ′
ψ(kτ )µh = {y0 , y1 } ⊂ VarV′ .

We let ζV ∈ VarV be such that E(ð(kτ ) rµ sh ) ∩ V = (ζV = 0) where E(ð(kτ ) rµ sh ) is


the exceptional divisor of the blowup Re(ð(kτ ) rµ sh ) → Re(ð(kτ ) rµ sh−1 ) . Without loss of
generality, we assume that y0′ corresponds to the exceptional ζV on the chart V. We
then let y1 ∈ VarV be the proper transform of y1′ on the chart V.
Here, we observe that
′ ′ ′ e=0
ψ(kτ )µh = {y0 , y1 } ⊂ ΓV′

because Ze(ð(kτ ) rµ sh−1 ),Γ is contained in the proper transform Zψ′ (kτ )µh .
We set,
=0
e=0′ \ψ ′
ΓV = {ζV , yV | yV is the proper transform of some yV′ ∈ Γ
(8.40) V (kτ )µh },

=1
(8.41) e=1′ }.
ΓV = { yV | yV is the proper transform of some yV′ ∈ Γ V

Consider the scheme-theoretic pre-image ρ−1 e


(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ).

Note that scheme-theoretically, we have

ρ−1 e −1 e e
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ∩ V = π(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ∩ V(ð(kτ ) rµ sh ) ∩ V.

Applying Lemma 8.9 (1) in (ð(kτ ) rµ sh−1 ) to Ze(ð(kτ ) rµ sh−1 ),Γ and ρ−1 e
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ),

and applying Proposition 7.10 to Ve(ð r s ) ∩ V, we obtain that the pre-image


(kτ ) µ h

ρ−1 e ∩ V as a closed subscheme of V, is defined by


(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ )

=0 mn q =1
(8.42) yV ∈ ΓV ; yV − 1, yV ∈ ΓV ; BV ; BV ; LV,Fm


.
=0
(Note that we have ζV ∈ ΓV .)
Thus, by setting
=1
e =0 and yV = 1 for all yV ∈ Γ
yV = 0 for all yV ∈ Γ V V

mn q
in BV , BV , LV,Fm


of the above, we obtain
mn q
(8.43) B̃V , B̃V , L̃V,Fm


.
GRASSMANNIANS AND SINGULARITIES 121


Again, for any F̄ ∈ FV,Γ , if a term of LV,F contains y1 ∈ VarV , then it contains
ζV y1 , hence L̃V,F does not contain y1 . We keep those equations of (8.43) such that
they contain the variable y1 and obtain
mn q
(8.44) B̂V , B̂V ,

viewed as a system of equations in y1 . Then, Proposition 7.10 (1) and (2), (8.44)
is a linear system of equations in y1 . (We point out that y1 here can correspond to
either y0 or y1 as in Proposition 7.10.) Furthermore, one sees that

(ρ−1 e e ′
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ∩ V)/(Z(ð(kτ ) rµ sh−1 ),Γ ∩ V )

is defined by the linear system (8.44).


There are the following two cases for (8.44):
(⋆a) the ranks of the linear system (8.44) equal one at general points of Ze(ð

(kτ ) rµ sh−1 ),Γ
.
(⋆b) the ranks of the linear system (8.44) equal zero at general points of Ze(ð

(kτ ) rµ sh−1 ),Γ
,
hence at all points of Ze† .
(ð(kτ ) rµ sh−1 ),Γ

Proof of Lemma 8.9 in (ð(kτ ) rµ sh ) under the condition (⋆a).

By the condition (⋆a), there exists a Zariski open subset Ze(ð


†◦
(kτ ) rµ sh−1 ),Γ
of Ze(ð

(kτ ) rµ sh−1 ),Γ

such that the rank of the linear system (8.44) equals one at any point of Z e †◦
.
(ð(kτ ) rµ sh−1 ),Γ

By solving y1 from the linear system (8.44) over Ze(ð


†◦
(kτ ) rµ sh−1 ),Γ
, we obtain that the
induced morphism

ρ−1 e†◦ e◦
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) −→ Z(ð(kτ ) rµ sh−1 ),Γ

is an isomorphism.
Suppose y1 is identically zero along ρ−1 e†◦
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ). We then set,

(8.45) e=0 = {y1 } ∪ Γ=0


Γ V V

=0
where ΓV is as in (8.40). In this case, we let

(8.46) Ze(ð(kτ ) sh ),Γ = ρ−1 e


(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ∩ Dy1

scheme-theoretically, where Dy1 is the closure of (y1 = 0) in Re(ð(kτ ) rµ sh ) . We remark


here that Dy1 does not depend on the choice of the chart V.
122 YI HU

Suppose y1 is not identically zero along ρ−1 e†◦


(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ). We then set,

=0
(8.47) e=0
Γ V = ΓV

=0
where ΓV is as in (8.40). In this case, we let

(8.48) Ze(ð(kτ ) sh ),Γ = ρ−1 e


(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ).

We always set (under the condition (⋆a))


=1
(8.49) e=1
Γ V = ΓV

=1
where ΓV is as in (8.41).
In each case, we have

ρ−1 e†◦ e
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ⊂ Z(ð(kτ ) rµ sh ),Γ ,

and we let Ze(ð



(kτ ) rµ sh ),Γ
be the closure of ρ−1 e†◦ e
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) in Z(ð(kτ ) rµ sh ),Γ . It is
an irreducible component of Ze(ð rµ s ),Γ because Ze†
(kτ ) h
is closed in Ze(ð rµ s ),Γ
(ð(kτ ) rµ sh ),Γ (kτ ) h

and contains the Zariski open subset −1 e †◦


ρ(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) of Ze(ð(kτ ) rµ sh ),Γ . Then,
we have that the composition

Ze(ð

(kτ ) rµ sh ),Γ
−→ Ze(ð

(kτ ) rµ sh−1 ),Γ
−→ ZΓ

is birational. This proves Lemma 8.9 (2) in (ð(kτ ) rµ sh ).


In each case of (8.45) and (8.47), by the paragraph of (8.42), we have that
Ze(ð(kτ ) rµ sh ),Γ ∩ V as a closed subscheme of V is defined by the equations as stated in
the Lemma. This proves Lemma 8.9 (1) in (ð(kτ ) rµ sh ).
It remains to prove Lemma 8.9 (3) in (ð(kτ ) rµ sh ). But, it follows from the identical
lines of the corresponding proof of Lemma 8.7 (3). We avoid the repetition.

Proof of Lemma 8.9 in (ð(kτ ) rµ sh ) under the condition (⋆b).

Under the condition (⋆b), we have that

ρ−1 e† e†
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) −→ Z(ð(kτ ) rµ sh−1 ),Γ

can be canonically identified with the trivial P[ξ0 ,ξ1 ] -bundle:

ρ−1 e† e†
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) = Z(ð(kτ ) rµ sh−1 ),Γ × P[ξ0 ,ξ1 ] .
GRASSMANNIANS AND SINGULARITIES 123

In this case, we define

Ze(ð(kτ ) rµ sh ),Γ = ρ−1 e


(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ∩ ((ξ0 , ξ1 ) = (1, 1)),

Ze(ð

(kτ ) rµ sh ),Γ
:= ρ−1 e†
(ð(kτ ) rµ sh ) (Z(ð(kτ ) rµ sh−1 ),Γ ) ∩ ((ξ0 , ξ1 ) = (1, 1)),

both scheme-theoretically.
The induced morphism Ze(ð †
(kτ ) rµ sh ),Γ
→ Ze(ð

(kτ ) rµ sh−1 ),Γ
is an isomorphism. There-
fore, Ze(ð

(kτ ) rµ sh ),Γ
→ Ze(ð

(kτ ) rµ sh−1 ),Γ
→ ZΓ is birational. Again, one sees that Ze(ð †
(kτ ) rµ sh ),Γ
e
is an irreducible component of Z(ð rµ s ),Γ . This proves Lemma 8.9 (2) in (ð(kτ ) rµ sh ).
(kτ ) h

Further, under the condition (⋆b), we set


e =0 = Γ=0 ,
Γ e=1 = {y1 } ∪ Γ=1 .
Γ
V V V V

Then, by the paragraph of (8.42), we have that Ze(ð(kτ ) rµ sh ),Γ ∩ V, as a closed


subscheme of V, is defined by the equations as stated in the Lemma. This proves
Lemma 8.9 (1) in (ð(kτ ) rµ sh ).
It remains to prove Lemma 8.9 (3) in (ð(kτ ) rµ sh ). But, again, it follows from the
identical lines of the corresponding proof of Lemma 8.7 (3). We avoid the repetition.
This completes the proof of Lemma 8.9. 

Definition 8.10. We call Ze(ð(kτ ) rµ sh ),Γ the ð-transform of ZΓ in Ve(ð(kτ ) rµ sh ) for any
(kτ )µh ∈ IndexΨ .

We need the final case of Lemma 8.9. We set

Zeð,Γ := Ze(ð(kτ ) rµ sh ),Γ , Ze̺,Γ



:= Ze(ð

(kτ ) rµ sh ),Γ

where k = Υ, τ = tFΥ , µ = κΥtFΥ , h = ς(ΥtFΥ )κΥtF . That is,


Υ

Ze̺,Γ = Ze(ð(ΥtF ) rκΥt sς(Υt ),Γ , Ze̺,Γ



= Ze(ð

r sς(Υt ),Γ .
Υ FΥ FΥ )κΥtF (Υt FΥ ) κΥtF FΥ )κΥtF
Υ Υ Υ

Corollary 8.11. Fix any standard chart V of Reð such that Zeð,Γ ∩ V 6= ∅. Then,
Zeð,Γ ∩ V, as a closed subscheme of V, is defined by the following relations
e=0 ;
y, ∀ y ∈ Γ e=1 ;
y − 1, ∀ y ∈ Γ
V V

mn q
(8.50) BV , BV , LV,Fm


.

Furthermore, the induced morphism Zeð,Γ



→ ZΓ is birational.
124 YI HU

9. Main Statements on the Final Scheme Veð

Let p be an arbitrarily fixed prime number. Let F be either Q or a finite field with
p elements. In this entire section, every scheme is defined over Z, consequently, is
defined over F, and is considered as a scheme over the perfect field F.
Let Γ be any subset VarUm . Assume that ZΓ is integral. Let Zeð,Γ be the ð-
transform of ZΓ in Veð . We now investigate local properties of the ð-transform Zeð,Γ,
using standard charts. (As just mentioned, ZΓ and Zeð,Γ are F-schemes.) We remind
the reader that Zeð,∅ = Veð when Γ = ∅.
Fix any standard chart V of Reð . By Corollary 8.11, the scheme Zeð,Γ ∩ V, if
nonempty, as a closed subscheme of the chart V of Reð , is defined by

(9.1) e =0 ;
y, y ∈ Γ e=1 ;
y − 1, y ∈ Γ q
BV ;
V V

(9.2) BV,(kτ ) , ∀ s = (kτ ) ∈ SFk \sFk , Fk ∈ Fm ,



(9.3) LV,Fk : ∀ Fk ∈ FV,Γ .

We remind the reader that any main binomial is indexed by some (kτ ) ∈ IndexBmn ;
it corresponds to the k-th Plücker equation F̄k ∈ Fm and a non-leading term of F̄k
indexed by some s ∈ SFk \sFk . Via this correspondence, we sometime write s = (kτ ).
Consequently, we may also write Bs = B(kτ ) .
Fix any k ∈ [Υ]. We let Gk be the set of all equations in (9.2) and (9.3), called
F
the block k of the defining equations of Zeð,Γ. We let G = Υ Gk . k=1

Throughout the remainder of this section, we let z be a fixed closed point of Veð ∩V.
Observe that all the binomial equations B(kτ ) terminate on the chart Reð , according
to Corollary 7.11. In particular, both terms of BV,(kτ ) do not vanish at z.
The standard chart V must be lying over a unique standard chart V[0] of Reϑ[0] =
RF . By Definition 4.16, the chart V[0] is uniquely indexed by a set

(9.4) ΛoFm = {(vsF,o , vsF,o ) | F̄ ∈ Fm }.

And, given ΛoFm , we have the set Λ⋆Fm = ΛFm \ΛoFm .


We let π := πV,V[0] : V −→ V[0] be the (induced) projection. Consider any
k ∈ [Υ]. Recall that we have the set BFmn
k
= {B(kτ ) | 1 ≤ τ ≤ tFk }. In addition,
GRASSMANNIANS AND SINGULARITIES 125

the notion of termination of BV[0] ,(kτ ) at the point πV,V[0] (z) have been introduced
in Definition 5.6 for all (kτ ) ∈ IndexBmn .

Definition 9.1. Consider the point z of V. Consider (kτ ) ∈ IndexBmn . We say


BV,(kτ ) is original at z if BV[0] ,(kτ ) terminates at the point π(z); we let BForik be the set
of all original BV,(kτ ) at z. We say BV,(kτ ) is ℏ-intrinsic (or simply, intrinsic) at z
if BV[0] ,(kτ ) is not original; we let BFinck be the set of all ℏ-intrinsic BV,(kτ ) at z.

Thus, we have
BFmn
k
= BForik ⊔ BFinck .

We point out that this decomposition depends on the fixed point z ∈ Veð ∩ V.
When both terms of BV[0] ,(kτ ) do not vanish at π(z), the form of BV[0] ,(kτ ) re-
mains unchanged throughout the ℏ-blowups, except the changes of the names of the
variables, whence the name of original. On the other hand, if BV,(kτ ) is ℏ-intrinsic
(e.g., when Fk is Γ-irrelevant) and π(z) ∈ V , then both terms of BV[0] ,(kτ ) vanishes
at π(z). Thus, some ℏ-blowup with ℏ ∈ a has to affect a neighborhood of πð,ℏ (z)
where πð,ℏ : Reð −→ Reℏ is the induced blowup morphism, and, the form of BV[0] ,(kτ )
must have positively changed when reaching at the final destination on V, around
the point z, whence the name of ℏ-intrinsic.
By Proposition 4.20, on the chart V[0] , we have that Veϑ[0] ∩ V[0] , as a closed
subscheme of V[0] , is defined by

res q
BV [0]
; BV [0]
;

(9.5) xV[0] ,(us ,vs ) xV[0] ,uF − xV[0] ,(m,uF ) xV[0] ,us xV[0] ,vs , ∀ s ∈ SF \sF ,
X
(9.6) LV[0] ,F : sgn(s)xV[0] ,(us ,vs ) , ∀ F̄ ∈ Fm .
s∈SF

Here, recall that uF := usF and sF ∈ SF is the index for the leading term of F̄ . Also
recall from the proof of Proposition 4.17 that on the chart V[0] , xV[0] ,u = xu , and
xV[0] ,(u,v) is the in-homogenization of x(u,v) .

Lemma 9.2. Consider the fixed point z ∈ Veð ∩ V. Fix any F̄ ∈ Fm . Assume that
BFori 6= ∅, equivalently, BFinc 6= BFmn . Then, xV[0] ,uF (π(z)) 6= 0, hence, xV,uF exists as
126 YI HU

a coordinate variable in VarV and xV,uF (z) 6= 0. Equivalently, if xV[0] ,uF (π(z)) = 0,
then BFori = ∅ and BFinc = BFmn .

Proof. This follows from definition. 

Fix F̄ ∈ Fm . To continue, we let

SFori = {s ∈ SF \sF | Bs ∈ BFori }, SFinc = {s ∈ SF \sF | Bs ∈ BFinc }.

There are two extreme cases for the decomposition BFmn = BFori ⊔ BFinc : BFmn = BFori
and BFmn = BFinc , where the latter must be the case when xV[0] ,uF (π(z)) = 0.
To analyze all possibilities, we first divide it into two classess:

sF,o = sF and sF,o 6= sF .

(See (9.4) for the explanation of sF,o.)

Suppose sF,o = sF and BFori 6= ∅. Then, we can write SFori = {s1 , · · · , sℓ } for some
1 ≤ ℓ ≤ tF and SF \sF = SFori ⊔ SFinc . In this case, xV,(m,uF ) = xV,(us , vs ) = 1.
F,o F,o

Hence, the binomial equations of BFori on the chart V are

(9.7) BV,si : xV,(us ,vs ) xuF − xV,us xV,vs , 1 ≤ i ≤ ℓ.


i i i i

Suppose sF,o 6= sF , and BFori 6= ∅. Then, in this case, as xV,(us , vs ) = 1 and


F,o F,o

xV,uF (z) 6= 0 (by Lemma 9.2), we must have sF,o ∈ SFori . Hence, we can write
SFori = {sF,o, s1 , · · · , sℓ } for some 1 ≤ ℓ ≤ tF . Therefore, the binomial equations of
BFori on the chart V are

(9.8) BV,so : xV,uF − xV,(m,uF ) xV,us xV,vs ,


F,o F,o

(9.9) BV,si : xV,(us ,vs ) xV,uF − xV,(m,uF ) xV,us xV,vs = 0, 1 ≤ i ≤ ℓ.


i i i i

Indeed, we must have ℓ ≥ 1 in the above. To see this, consider π(z) ∈ V[0] ∩ V[0] ,
then, LV[0] ,F is satisfied by π(z). Thus, one sees that there must be an index s1 6= sF,o
such that xV[0] ,(us1 ,vs1 ) (π(z)) 6= 0. Hence, BV,s1 is original because xV[0] ,uF (π(z)) 6= 0
as well.
Next, we consider SFinc . We suppose SFinc 6= ∅. We write

SFinc = {t1 < · · · < tq }


GRASSMANNIANS AND SINGULARITIES 127

for some 1 ≤ q ≤ tF .
By Definition 7.9, we let yV,ti be the (unique) terminating central variable for
Bti ∈ SFinc for all 1 ≤ i ≤ q. We write

inc
(9.10) yV,F = {yV,t1 , · · · , yV,tq }.

We then set
inc
∗ inc
∂(BV,F ) ∂(BV,t1 , · · · , BV,tq )
(9.11) J (BV,F ) := := .
∂(yinc
V,F )
∂(yV,t1 , · · · , yV,tq )

Lemma 9.3. Suppose z ∈ Veð ∩ V. We have the following.

(1) The matrix J ∗ (BV,F


inc
) is invertible lower-triangular at z;
(2) Consider any Ḡ ∈ Fm such that Ḡ < F̄ . Then, the variable yV,ti does not
appear in the main binomial equation BV,s , either original or ℏ-intrinsic at
z, for all s ∈ SG \sG and all 1 ≤ i ≤ q.
(3) Further, for any H̄ ∈ Fm , if yV,ti appears in LV,H , then it appears in a term
that vanishes at z, for all 1 ≤ i ≤ q.

Proof. (1). Note that BV,ti , 1 ≤ i ≤ q, terminates at z. As the term of BV,ti that
∂BV,ti
contains yV,ti is linear in yV,ti according to Corollary 7.12, we have ∂yV,ti
(z) 6= 0. By
Proposition 7.10 (3), when j < i, at the point z, the ℏ-intrinsic binomial BV,tj must
terminate earlier than the ℏ-intrinsic binomial BV,ti does. Hence, one sees that the
variable yV,ti , as the terminating central variable of BV,ti at z, can not appear in
∂BV,tj
BV,tj since BV,tj already terminates. Hence, ∂yV,ti
(z) = 0 when j < i. This implies
that the matrix J ∗ (BV,F
inc
) is lower-triangular with nonzero entries along the diagonal
at z, hence also invertible at z.
(2). For any Ḡ ∈ Fm such that Ḡ < F̄ , we have that BV,s with s ∈ SG \sG ,
either original or ℏ-intrinsic at z, must terminate earlier than BV,ti does at z, for
all 1 ≤ i ≤ q. Hence, by the identical arguments as in (1) (for the case j < i), we
conclude that the variable yV,ti does not appear in BV,s .
(3). Now, if yV,ti appears in LV,H for some H̄ ∈ Fm , so does ζV yV,ti where
ζV ∈ VarV is the terminating exceptional parameter (see Definition 7.9 for the
explanation of ζV ). Since ζV (z) = 0, the statement follows. 
128 YI HU

Definition 9.4. Fix and consider any variable y ∈ VarV that appears in some
equations of the block Gk for some k ∈ [Υ] such that y(z) 6= 0. (See (9.2), (9.3),
and the subsequent paragraph for the explanation of Gk ). We say that y is pleasant
if either y does not appear in any equation of the block Gj for all 1 ≤ j < k, or else,
when it does, it appears in a term that vanishes at z.

By this definition, the terminating central variables yV,ti in Lemma 9.3 are pleas-
ant due to (2) and (3) of that lemma.

Definition 9.5. A scheme X is smooth if it is a disjoint union of connected smooth


schemes of possibly various dimensions.

Theorem 9.6. Let Γ be any subset VarUm . Assume that ZΓ is integral. Let Zeð,Γ be
the ð-transform of ZΓ in Veð . Then, Zeð,Γ is smooth over Spec F. Consequently, Ze† ð,Γ

is smooth over Spec F.


In particular, when Γ = ∅, we obtain that Veð is smooth over Spec F.

Proof. Fix any closed point z ∈ Zeð,Γ and let V be a standard chart of Reð such that
z ∈ V. Let Γe=0 and Γ
e=1 be as in Corollary 8.11. We let
V V

eV = Γ
Γ e=0 ⊔ Γ
e=1 .
V V

e=0 and y = 1 for all y ∈ Γ


By setting y = 0 for all y ∈ Γ e=1 , we obtain an affine
V V

subspace VΓ of V. That is,

e =0 ; y = 1, y ∈ Γ
VΓ = {y = 0, y ∈ Γ e =1 } ⊂ V.
V V

As an affine space, VΓ comes equipped with the set of coordinate variables

eV }.
{y | y ∈ VarV \Γ

For any polynomial f ∈ k[y]y∈VarV , we let f |ΓeV be obtained from f by setting all
e=0 to be 0 and setting all variables in Γ
variables in Γ e=0 to be 1. This way, f |e
V V ΓV
becomes a polynomial over VΓ .
GRASSMANNIANS AND SINGULARITIES 129

Then, Zeð,Γ ∩ V can be identified with the closed subscheme of VΓ defined by

q
(9.12) BV |ΓeV , ∀ B q ∈ Bq ,

(9.13) BV,(kτ ) |ΓeV , ∀ (kτ ) ∈ IndexBmn ,



(9.14) LV,Fk |ΓeV , ∀ F̄k ∈ FV,Γ

For any subset P of polynomials over V, we let P |ΓeV = {f |ΓeV | f ∈ P }. This way,
mn q
we have BV |ΓeV , BV |ΓeV , etc..
In what follows, we focus on the polynomial equations in (9.13) and (9.14), treated
as polynomials in k[y]y∈VarV \ΓeV . We will analyze the Jacobian of these polynomials.
Fix any F̄ = F̄k ∈ Fm with k ∈ [Υ]. We divide the analysis into three cases.

ori mn inc
Case α. First, we suppose BV,F = ∅. Hence, in this case, BV,F = BV,F . (Note that
this must be the case when F is Γ-irrelevant.)

Assume first that F̄ ∈
/ FV,Γ (thus, F must be Γ-irrelevant. cf. (8.14)). In this
case, by Corollary 8.11, LV,F |ΓeV is not one of the defining equations. Solely for the
purpose of uniform writing to be conveniently used later, we set “LV,F |ΓeV ≡ 0” when
⋆ inc inc
F̄ ∈
/ FV,Γ . In such a case, we only need to consider BV,F . We let yV,F be the set of
inc
the terminating central variables for the equations BV,F , determined and listed as
in (9.10) and (9.11) (cf. Lemma 9.3). Then, we set
inc
∂(BV,F |ΓeV )
J ∗ (LV,F |ΓeV , BV,F
mn
|ΓeV ) = J ∗ (LV,F |ΓeV , BV,F
inc
|ΓeV )(z) := inc
.
∂(yV,F )

We remark here again that “LV,F |ΓeV ≡ 0” is not a defining equation.



Next we assume that F̄ ∈ FV, m (this is automatic when F is Γ-relevant). As LF
terminates on RF , there exists s ∈ SF such that xV,(us ,vs ) exists in VarV and the
∂LV,F |Γ
e
term sgn(s)xV,(us ,vs ) of LV,F does not vanish at z. Hence, V
(z) = sgn(s) 6= 0.
∂xV,(u
v s, s)
inc inc
Now, we consider BV,F . We let yV,F be the set of the terminating central variables
inc
for the equations BV,F , determined and listed as in (9.10) and (9.11). Then, we set

inc
∂(LV,F |Γe , BV,F |ΓeV )
∗ mn ∗ inc
J (LV,F |ΓeV , BV,F |ΓeV ) =J (LV,F |ΓeV , BV,F |ΓeV )(z) := V
inc
.
∂(xV,(us ,vs ) , yV,F )
130 YI HU

Then, one sees that at the point z, J ∗ (LV,F |ΓeV , BV,F ) takes form

 ∂LV,F |Γ

e
∂xV,(u
V
(z) 0
J ∗ (LV,F |ΓeV , BV,F
mn
|ΓeV )(z) =  v
s, s) ,
∗ inc
∗ J (BV,F )(z)

where the zero entries in the upper-right corner are due to Lemma 9.3 (3).
Then, in either case of F̄ ∈ ⋆
/ FV,Γ ⋆
and F̄ ∈ FV,Γ , by Lemma 9.3 (1), J ∗ (BV,F
inc
|ΓeV )
is an invertible lower-triangular matrix at the point z, implying that in either case,
J ∗ (LV,F |ΓeV , BV,F |ΓeV ) is of full rank at z.

Further, when F̄ ∈ FV,Γ , observe that the variable xV,(us ,vs ) uniquely appears in
the equations in the block Gk with F = Fk . Hence, it is pleasant (see Definition
9.4). In addition, by Lemma 9.3 (2) and (3), all the terminating central variables
used to compute J ∗ (BV,F
inc
|ΓeV ) are pleasant, too. Thus, we conclude that in this case
⋆ ⋆
(i.e., Case α), in either case of F̄ ∈
/ FV,Γ and F̄ ∈ FV,Γ , all the variables that are
used to compute the (partial) Jacobian J ∗ (LV,F |ΓeV , BV,F
mn
|ΓeV ) are pleasant.

Next, we assume BFori 6= ∅, that is, BFinc 6= BFmn . (In such a case, F must be Γ-
relevant.) As earlier, we let V[0] be the unique chart of RF such that V lies over
V[0] and π = πV,V[0] : V −→ V[0] be the projection. Then, xV[0] ,uF (π(z)) 6= 0, hence,
xV,uF (z) 6= 0, by Lemma 9.2. As discussed in the paragraphs subsequent to Lemma
9.2, we have the following two situations to consider: sF,o = sF and sF,o 6= sF .

Case β. Suppose BFori 6= ∅ and sF,o = sF . Then, we write SFori = {s1 , · · · , sℓ } for some
1 ≤ ℓ ≤ tF . And, we have that the binomial equations of BFori take the following
(original) forms:

(9.15) BV,si : xV,(us ,vs ) xV,uF − xV,us xV,vs , 1 ≤ i ≤ ℓ.


i i i i

Suppose that BFinc 6= ∅. As before, we write SFinc = {t1 < · · · < tq } for some
1 ≤ q ≤ tF . Then, on the chart V[0] , we have

BV[0] ,ti : xV[0] ,(ut ,vt ) xV(10),u − xV[0] ,uti xV[0] ,vti , 1 ≤ i ≤ q.
i i F
GRASSMANNIANS AND SINGULARITIES 131

As BV,ti is ℏ-intrinsic at z and xV[0] ,uF (π(z)) 6= 0, one finds xV[0] ,(ut ,vt ) (π(z)) = 0.
i i

Then, LV,F admits the following form

ℓ q
X X
(9.16) LV,F = sgn(sF ) + sgn(si )xV,(us ,vs ) + sgn(ti )x̃V,(ut ,vt ) ,
i i i i
i=1 i=1

where x̃V,(ut ,vt ) = π ∗ xV[0] ,(ut ,vt ) is the pullback of xV[0] ,(ut ,vt ) . In particular, we
i i i i i i

have

(9.17) x̃V,(ut ,vt ) (z) = 0, ∀ 1 ≤ i ≤ q.


i i

As the chart V is fixed and is clear from the context, in what follows, for simplicity
of writing, we will selectively drop some subindex “ V ”. For instance, we may write
xuF for xV,uF , x(us ,vs ) for xV,(us ,vs ) , etc. A confusion is unlikely.
1 1 1 1

We introduce the following (partial) Jacobian matrix

∂(BV,s1 |ΓeV · · · BV,sℓ |ΓeV , LV,F |ΓeV )


J ∗ (BV,F
ori
|ΓeV , LV,F |ΓeV ) = .
∂(xuF , x(us ,vs ) · · · x(us ,vs ) )
1 1 ℓ ℓ

Then, one calculates and finds


 
x(us ,vs ) xuF 0 ··· 0
 1 1 
 x(us2 ,vs2 ) 0
xuF ··· 0 
 
 .. 
J ∗ (BV,F
ori
, LV,F |ΓeV ) = 
 . .

 
 x(u ,v ) 0 0 ··· xuF 
 sℓ sℓ 
0 sgn(s1 ) sgn(s2 ) · · · sgn(sℓ )

Multiplying the first column by −xuF (6= 0 at z), we obtained


 
−x(us1 ,vs1 ) xuF xuF 0 ··· 0
 
 −x ··· 
 (us ,vs ) xuF 0 xuF 0 
 2 2

 ... .
 
 
 −x(u ,v ) xu 0 0 ··· xuF 
 sℓ sℓ F 
0 sgn(s1 ) sgn(s2 ) · · · sgn(sℓ )
132 YI HU

Multiplying the (i+1)-th column by x(us ,vs ) and added to the first for all 1 ≤ i ≤ h,
i i

we obtain
 
0 xuF 0 ··· 0
 
 0 0 x · · · 0 
 uF 
 .. .. .. .. 
 . . . · · · . .
 
 
 0 xu 0 · · · xu 
 P F F 

i=1 sgn(s )x
i (us ,vs ) sgn(s 1 ) sgn(s 2 ) · · · sgn(s ℓ )
i i

But, at the point z, we have



X X
sgn(si )x(us ,vs ) (z) = −sgn(sF ) − sgn(t)x̃(ut ,vt ) (z) = −sgn(sF ) 6= 0,
i i
i=1 inc
t∈SF

because of (9.16) and (9.17). Thus, we conclude that J ∗ (BV,F


ori
|ΓeV , LV,F |ΓeV ) is a
square matrix of full rank at z.
inc
We now consider BV,F . If BFinc = ∅, there is nothing to consider, and we move on.
Suppose BFinc 6= ∅. We let yV,F
inc
be the set of the terminating central variables for
inc
the equations BV,F , determined and listed as in (9.10) and (9.11). We then set
inc
∂(BV,s1 |ΓeV · · · BV,sℓ |ΓeV , LV,F |ΓeV , BV,F |ΓeV )
J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV ) = inc
.
∂(xuF , x(us1 ,vs1 ) · · · x(us ,vs ) , yV,F )
ℓ ℓ

Then, by Lemma 9.3 (2) and (3), one sees that J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV ) at the
point z takes form
 
J ∗ (BV,F
ori
|ΓeV , LV,F |ΓeV )(z) 0
J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV )(z) =  inc |
∂(BF e )
Γ
.
∗ inc )
∂(yV,F
V
(z)
inc |
∂(BF e )
Γ
Then, applying Lemma 9.3 (1) to inc )
∂(yV,F
V
and the previous discussion on the matrix
J ∗ (BV,F
ori
|ΓeV , LV,F |ΓeV ), we conclude that the (partial) Jacobian J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV )
for the block of equations in Gk , where F = Fk , achieves its full rank at z.
Further, observe here that the variables x(us ,vs ) · · · x(us ,vs ) uniquely appear in
1 1 ℓ ℓ

the block Gk . Also, xV,uF , as the leading variable of F , does not appear in any
equation of Gj with j < k (see Proposition 3.9). In addition, by Lemma 9.3 (2) and
(3), all the terminating central variables used to compute J ∗ (BV,F
inc
|ΓeV ) are pleasant.
GRASSMANNIANS AND SINGULARITIES 133

Thus, we conclude that in this case (i.e., Case β), all the variables that are used to
compute the (partial) Jacobian J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV ) are pleasant.

Case γ. Now, we suppose BFori 6= ∅ and sF,o 6= sF . Then, as earlier, we can write
SFori = {sF,o, s1 , · · · , sℓ } for some 1 ≤ ℓ ≤ tF . And, we have the binomial equations
of BFori as

(9.18) BV,so : xV,uF − xV,(m,uF ) xV,us xV,vs ,


F,o F,o

(9.19) BV,si : xV,(us ,vs ) xV,uF − xV,(m,uF ) xV,usi xV,vsi = 0, 1 ≤ i ≤ ℓ.


i i

Again, we selectively drop some subindex “ V ” below. Then, in this case, one
calculates and finds

∂(BV,sF,o |ΓeV , BV,s1 |ΓeV · · · BV,sℓ |ΓeV , LV,F |ΓeV )


J ∗ (BV,F
ori
|ΓeV , LV,F |ΓeV ) :=
∂(xuF , x(m,uF ) , x(us1 ,vs1 ) · · · x(us ,vs ) )
ℓ ℓ

is equal to

 
1 −xus xvs 0 ··· 0
 F,o F,o 
 x −xus1 xvs1 xuF ··· 0 
 (us1 ,vs1 ) 
 .. 
 . .
 
 
 x(u ,v ) −xus xvs 0 ··· xuF 
 sℓ sℓ ℓ ℓ 
0 sgn(sF ) sgn(s1 ) · · · sgn(sℓ )

Multiplying the first column by xuF (6= 0 at z), we obtained

 
xuF −xus xvs 0 ··· 0
 F,o F,o 
 x x −xus1 xvs1 xuF ··· 0 
 uF (us1 ,vs1 ) 
 .. 
 . .
 
 
 xu x(u ,v ) −xus xvs 0 ··· xuF 
 F sℓ sℓ ℓ ℓ 
0 sgn(sF ) sgn(s1 ) · · · sgn(sℓ )
134 YI HU

Multiplying the second column by x(m,uF ) and added to the first, we obtain
 
0 −xus xvs 0 ··· 0
 F,o F,o 
 0 −xus1 xvs1 xuF ··· 0 
 
 . 
 .. .
 
 
 0 −xus xvs 0 ··· xuF 
 ℓ ℓ 
sgn(sF )x(m,uF ) sgn(sF ) sgn(s1 ) · · · sgn(sℓ )
As sgn(sF )x(m,uF ) , xus xvs and xuF are all not vanishing at z by (9.18), one see
F,o F,o

that the above matrix, hence J ∗ (BV,F


ori
|ΓeV , LV,F |ΓeV ), is of full rank at z.
inc
We now consider BV,F . If BFinc = ∅, there is nothing to consider. Suppose BFinc 6= ∅.
inc
As in Case β, we let yV,F be the set of the terminating central variables for the
inc
equations BV,F , determined and listed as in (9.10) and (9.11). We then set
inc
∂(BV,sF,o |ΓeV , BV,s1 |ΓeV · · · BV,sℓ |ΓeV , LV,F |ΓeV , BV,F )
J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV ) = inc
.
∂(xuF , x(m,uF ) , x(us1 ,vs1 ) · · · x(us ,vs ) , yV,F )
ℓ ℓ

Then, by Lemma 9.3 (2) and (3), one sees that J ∗ (BV,F
ori inc
, LV,F |ΓeV , BV,F ) takes the
following form at the point z
 
∗ ori
J (BV,F |ΓeV , LV,F |ΓeV )(z) 0
J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV )(z) =  inc |
∂(BF e )
Γ
.
∗ V
inc )
∂(yV,F
(z)
inc |
∂(BF e )
Γ
Then, applying Lemma 9.3 (1) to inc )
∂(yV,F
V
and the previous discussion on the matrix
J ∗ (BV,F
ori
|ΓeV , LV,F |ΓeV ), we conclude that the (partial) Jacobian J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV )
for the block of equations in Gk , where F = Fk , achieves its full rank at z.
Further, observe here that the variables x(m,uF ) , x(us ,vs ) · · · x(us ,vs ) uniquely
1 1 ℓ ℓ

appear in the block Gk . Also, xuF , as the leading variable of F , does not appear
in any equation of Gj with j < k. In addition, by Lemma 9.3 (2) and (3), all the
terminating central variables used to compute J ∗ (BV,F
inc
) are pleasant as well. Thus,
we conclude that all the variables that are used to compute the (partial) Jacobian
J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV ) are pleasant, for all F , in this case (i.e., Case γ).

Now, we return to the general case of the decomposition BF = BFori ⊔ BFinc , regard-
less weather BFori = ∅ or not, i.e., we consider all “Cases α, β, and γ” together. In
all the cases, we conclude that the (partial) Jacobian J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV ),
GRASSMANNIANS AND SINGULARITIES 135

where we allow either BFori = ∅ or BFinc = ∅, and also LV,F |ΓeV ≡ 0 is not one of the
irr
defining equation when F̄ ∈ Fm ,Γ , is a square matrix of full rank at the point z and

all the variables that are used to compute it are pleasant, for all F .
Thus, at the point z, by combining the (partial) Jacobian J ∗ (BV,F
ori inc
|ΓeV , LV,F |ΓeV , BV,F |ΓeV )
together for all F̄ ∈ Fm , we obtain a maximal minor, simply denoted by J ∗ , of the
complete Jacobian of all the equations of

{BV,s |ΓeV , LV,F |ΓeV | F̄ ∈ Fm , s ∈ SF \sF }

as follows (here again, we drop some subindex “ V ”)


(9.20)
 
J ∗ (BFori1 |Γe V , LF1 |Γe V , BFinc | e V )(z)
1 Γ
0 ··· 0
 ∗ J ∗ (BFori2 |Γe V , LF2 |Γe V , BFinc | e V )(z) ··· 0 
 2 Γ 
 .. .. .. .. .
 . . . . 
∗ ∗ ∗ J ∗ (BForiΥ |Γe V , LFΥ |Γe V , BFinc | e V )(z)
Υ Γ

All the zero blocks in the upper-right corner are because for every Fk with 1 ≤ k ≤
Υ, the variables that are used to compute J ∗ (BV,F
ori
k
inc
, LV,Fk |ΓeV , BV,F k
) are pleasant.
Hence, by our previous discussions, (9.20) is a square matrix of full rank at z.

Now, we consider the case when Γ = ∅. Thus, we have Z∅ = Um ∩ Grd,E and


Zeð,∅ = Veð . In this case, we let J := J(Bq , BV
mn
, LV,Fm ) be the full Jacobian of
q mn
all the defining equations of BV , BV , and LV,Fm := {LV,F | F̄ ∈ Fm }; we let
∗ ∗ mn
J := J (BV , LV,Fm ) be the matrix of (9.20) in the case of Γ = ∅. Take any
z ∈ Veð , let Tz (Veð ) be the Zariski tangent space of Veð at z. Then, we have

dim Tz (Veð ) = dim Reð − rank J(z) ≤ dim Reð − rank J ∗ (z)

= dim Um + |Bmn | − (|Bmn | + Υ) = dim Um − Υ = dim Veð ,


where dim Reð = dim Um + |Bmn | by (4.32) and rank J ∗ (z) = |Bmn | + Υ by (9.20).
Hence, dim Tz (Veð ) = dim Veð , thus, Veð is smooth at z. Therefore, Veð is smooth.
Consequently, one sees that on any standard chart V of the final scheme Reð , all
q mn
the relations of BV must lie in the ideal generated by relations of BV and LV,Fm ,
thus, can be discarded from the chart V.
Now, we return to a general subset Γ of VarUm as stated in the theorem. By
the previous paragraph, over any standard chart V of Reð with Zeð,Γ ∩ V 6= ∅, we
136 YI HU

q
can discard BV |ΓeV from the defining equations of Zeð,Γ ∩ V and focus only on the
equations of BVmn
|ΓeV and LV,Fm
⋆ |e . In other words, Zeð,Γ ∩ V, if nonempty, as
,Γ ΓV
mn
a closed subcheme of VΓ , is defined by the equations in BV |ΓeV and LV,Fm
⋆ |e .
,Γ ΓV
mn
Then, by (9.20), the rank of the full Jacobian of BV |ΓeV and LV,Fm ⋆ |e equals to
,Γ ΓV

the number of the above defining equations at any closed point z of Zeð,Γ ∩V. Hence,
Zeð,Γ ∩ V is smooth, thus, so is Zeð,Γ .
This proves the theorem. 

Let X be an integral scheme. We say X admits a resolution if there exists a


e and a proper birational morphism from X
smooth scheme X e onto X.

Theorem 9.7. Let Γ be any subset VarUm . Assume that ZΓ is integral. Then, the
morphism Ze† → ZΓ can be decomposed as
ð,Γ

Ze̺,Γ

→ · · · → Zeℏ,Γ

→ Zeℏ†′ ,Γ → · · · → ZF

[j] ,Γ

→ ZF [j−1] ,Γ
→ · · · → ZΓ

such that every morphism Zeℏ,Γ



→ Zeℏ†′ ,Γ in the sequence is Ze(ð

(kτ ) rµ sh ),Γ
→ Ze(ð

(kτ ) rµ sh−1 ),Γ

for some (kτ )µh ∈ IndexΨ , or Ze †


→ Ze †
(ð(kτ ) rµ sh ),Γ for some (kτ )µh ∈
(ð(kτ ) rµ sh−1 ),Γ
IndexΦ , or Zeϑ† [k] ,Γ → Zeϑ† [k−1] ,Γ
for some k ∈ [Υ]. Further, every morphism in the
sequence is surjective, projective, and birational. In particular, Zeð,Γ

→ ZΓ is a
resolution if ZΓ is singular.

Proof. The smoothness of Zeð,Γ



follows from Theorem 9.6; the decomposition of
Zeð,Γ → ZΓ follows from Lemmas 8.3, 8.5, 8.7, and Lemma 8.9.



10. Local Resolution of Singularity

We follow Lafforgue’s presentation of [12] on Mnëv’s universality theorem.


As before, suppose we have a set of vector spaces, E1 , · · · , En such that Eα is of
dimension 1 (or, a free module of rank 1 over Z). We let
M
EI = Eα , ∀ I ⊂ [n],
α∈I

E := E[n] = E1 ⊕ . . . ⊕ En .
(Lafforgues [12] considers the more general case by allowing Eα to be of any finite
dimension.)
GRASSMANNIANS AND SINGULARITIES 137

For any fixed integer 1 ≤ d < n, the Grassmannian

Grd,E = {F ֒→ E | dim F = d}

decomposes into a disjoint union of locally closed strata

Grd,E = {F ֒→ E | dim(F ∩ EI ) = dI , ∀ I ⊂ [n]}


d

indexed by the family d = (dI )I⊂[n] of nonnegative integers dI ∈ N verifying


• d∅ = 0, d[n] = d,
• dI + dJ ≤ dI∪J + dI∩J , for all I, J ⊂ [n].
The family d is called a matroid of rank d on the set [n]. The stratum Grd,E is
d
called a thin Schubert cell.
The Grassmannian Grd,E comes equipped with the (lattice) polytope

∆d,n = {(x1 , · · · , xn ) ∈ Rn | 0 ≤ xα ≤ 1, ∀ α; x1 + · · · + xn = d}.

For any i = (i1 , · · · , id ) ∈ Id,n , we let xi = (x1 , · · · , xn ) be defined by


(
xi = 1, if i ∈ i,
(10.1)
xi = 0, otherwise.

It is known that ∆d,n ∩ Nn = {xi | i ∈ Id,n } and it consists of precisely the vertices
of the polytope ∆d,n .
Then, the matroid d = (dI )I⊂[n] above defines the following subpolytope of ∆d,n
X
∆d,n = {(x1 , · · · , xn ) ∈ ∆d,n | xα ≥ dI , ∀ I ⊂ [n]}.
d
α∈I

This is called the matroid subpolytope of ∆d,n corresponding to d.


Recall that we have a canonical decomposition
M
∧d E = Ei1 ⊗ · · · ⊗ Eid
i∈Id,n

and it gives rise to the Plücker embedding of the Grassmannian

Grd,E ֒→ P(∧d E) = {(pi )i∈Id,n ∈ Gm \(∧d E\{0})}.


138 YI HU

Proposition 10.1. (Proposition, p4, [12]) Let d be any matroid of rank d on the
set [n] as considered above. Then, in the Grassmannian

Grd,E ֒→ P(∧d E) = {(pi )i∈Id,n ∈ Gm \(∧d E\{0})},

the thin Schubert cell Grd,E , as a locally closed subscheme, is defined by


d

pi = 0, / ∆d,n ,
∀ xi ∈
d

pi 6= 0, ∀ xi ∈ ∆d,n .
d

Let d = (dI )I⊂[n] be a matroid of rank d on the set [n] as above. Assume that
d[n]\{α} = d − 1 for all 1 ≤ α ≤ n. Then, the configuration space C d,n defined by the
d
matroid d is the classifying scheme of families of n points

P1 , · · · , Pn

on the projective space Pd−1 such that for any nonempty subset I ⊂ [n], the projec-
tive subspace PI of Pd−1 generated by the points Pα , α ∈ I, is of dimension

dim PI = d − 1 − dI .

Theorem 10.2. (Mnëv, Theorem I. 14, [12]) Let X be an affine scheme of finite
type over Spec Z. Then, there exists a matroid d of rank 3 on the set [n] such that
PGL3 acts freely on the configuration space C 3,n . Further, there exists a positive
d
integer r and an open subset U ⊂ X × Ar projecting surjectively onto X such that
U is isomorphic to the quotient space C 3,n := C 3,n /PGL3 .
d d

Theorem 10.3. (Gelfand, MacPherson, Theorem I. 11, [12]) Let d be any matroid
of rank d on the set [n] as considered above. Then, the action of PGLd−1 on C d,n is
d
d,n
free if and only if dimR ∆ = n − 1. Similarly, The action of Gm /Gm on Grd,n is
n
d d
d,n d,n
free if and only if dimR ∆ = n − 1. Further, when dimR ∆ = n − 1, the quotient
d d
d,n
C /PGLd−1 can be canonically identified with the quotient Grd,E /(Gnm /Gm ).
d d

By the above correspondence, we have the following equivalent version of Theorem


10.2.
GRASSMANNIANS AND SINGULARITIES 139

Theorem 10.4. (Mnëv, Theorem I. 14, [12]) Let X be an affine scheme of finite
type over Spec Z. Then, there exists a matroid d of rank 3 on the set [n] such that
(Gnm /Gm ) acts freely on the thin Schubert cell Gr3,E . Further, there exists a positive
d
r
integer r and an open subset U ⊂ X × A projecting surjectively onto X such that
U is isomorphic to the quotient space Gr3,E := Gr3,E /(Gnm /Gm ).
d d

Theorem 10.5. (Local Resulution) Let X be a scheme of finite presentation over


a field k. Assume further that X is integral and singular. Take any singular closed
point q ∈ X. Then, there exists an open subset V of X containing the point q such
that V admits a resolution.

Proof. The question is local. Hence, we can assume that X is affine.


First, we assume that X is defined over Spec Z.
We apply Theorem 10.4 to X and follow the notations in Theorem 10.4.
We identify U ⊂ X × Ar with the quotient space Gr3,E = Gr3,E /(Gnm /Gm ).
d d
Consider the quotient map

π : Gr3,E −→ Gr3,E = Gr3,E /(Gnm /Gm ).


d d d
We have the diagram
π
Gr3,E Gr3,E = Gr3,E /(Gnm /Gm ) ∼

/ =U / X × Ar
d d d


X.

We can apply Proposition 10.1 to the thin Schubert cell Gr3,E .


d
3,n 3,n
Since ∆ 6= ∅, there exists m ∈ I3,n such that xm ∈ ∆ . We define
d d
(10.2) / ∆3,n }.
Γ := Γd = {i ∈ I3,n | xi ∈
d

Then, we have that


Gr3,E 
 
/ ZΓ  / Um
d

and Gr3,E is an open subset of the Γ-scheme ZΓ ⊂ Um . As X is integral (by


d
assumption), one sees that ZΓ is integral.
140 YI HU

We then let

̟Γ : Zeð,Γ

−→ ZΓ

f 3,E = ̟ −1(Gr3,E ), scheme-


be as in Theorem 9.7. This is a resolution. We set Gr d Γ d
theoretically. Then
f 3,E 3,E
f 3,E : Grd −→ Grd
̟|Gr
d
is a resolution.
Take any general closed point a ∈ Ar , such that

(q, a) ∈ (X × a) ∩ U ⊂ X × Ar .

The quotient map

π : Gr3,E −→ Gr3,E (∼ U)
d d =
is a principal (Gnm /Gm )-bundle, and is étale locally trivial. As any étale locally trivial
principal (Gnm /Gm )-bundle is Zariski locally trivial (that is, (Gnm /Gm ) is special in
the sense of Serre), there exists a suitably small open subset V of X containing q
such that the following hold:

=
• W a := (V × a) ∩ U −→ V ;
• W a , considered as a locally closed subset of Gr3,E (∼ U), admit a lift Wa ⊂ Gr3,E
d = d
such that we have

= ∼
=
Wa −→ W a = (V × a) ∩ U −→ V.
f
Let W a be the closure of Wa in ZΓ (⊃ Gr3,E ). For generic choice of a, we let W a
d
3,E
be the proper transform of W a in Zeð,Γ
† f ). It is immediate that the induced
(⊃ Grd
f
morphism ρ : W −→ W is surjective, projective and birational with a ∈ Ar being
a a
fa = ρ−1 (Wa ). Then, the induced morphism
generic. Let W

f
fa : Wa −→ Wa
ρ|W

is surjective, projective and birational.


It remains to show that Wfa is smooth.
fa .
We take any closed point z ∈ W
GRASSMANNIANS AND SINGULARITIES 141

We can depict the above situation as the following diagram.

fa  
z∈W f 3,E ⊂ Ze†  
Gr Veð
d
/ /
ð,Γ

ρ|W
f ̟Γ
a
  
Gr3,E ⊂ ZΓ 
 
Wa  / / Um
d

= π
 
Gr3,E ∼
 
Wa  / = U / X × Ar
d

=
 

q∈V  / X

Now we let V ⊂ Reð be a standard chart of Reð such that z ∈ V. We let I be


the ideal of F[x]x∈VarV corresponding to Zeð,Γ

∩ V and Ia be the ideal of F[x]x∈VarV
corresponding to W f ∩ V. Further, we let m be the maximal ideal of F[x]
a x∈VarV

corresponding to z ∈ V, n = m/I be the maximal ideal of O e† (Ze ∩ V) corre-



Zð,Γ ∩V ð,Γ

sponding to z ∈ Zeð,Γ

∩ V, and na = m/Ia be the maximal ideal of OW f
f ∩V (W a ∩ V)
a
f ∩ V.
corresponding to z ∈ W a

Then, we have the following digram of exact sequences of vector spaces over
F(z) ∼
= F (see, equation (2.1), page 128, [14])

0 / I/(I ∩ m2 ) / m/m2 / n/n2 / 0,


=
  
2 2
0 / Ia /(Ia ∩ m ) / m/m / na /n2a / 0.

Hence, the left downarrow is injective.


Let Ar = Spec F[t1 , · · · , tr ]. For every i ∈ [r], we first pullback ti via π to obtain an
element of the coordinate ring ZΓ and then lift it to an element t′i of the coordinate
ring Um . We then let t̃i be the proper transform of t′i over the standard chart V.
This way, we obtain t̃1 , · · · , t̃r ∈ F[x]x∈VarV . Then, (t̃1 −a1 , · · · , t̃r −ar ) belong to the
ideal Ia where a = (a1 , · · · , ar ) ∈ Ar . One sees that the images of t̃1 − a1 , · · · , t̃r − ar
in Ia /(Ia ∩ m2 ) are linear independent over F(z) ∼ = F. Since W fa is birational to
= Wa ∼
Wa with Wa ∼ = V, we have that the vector subspace spanned by the images
142 YI HU

of t̃1 − a1 , · · · , t̃r − ar intersects trivially with I/(I ∩ m2 ), both regarded as vector


subspaces of Ia /(Ia ∩ m2 ).
Then, the above implies that we have

dimF(z) na /n2a = dimF(z) m/m2 − dimF(z) Ia /(Ia ∩ m2 )

≤ dimF(z) m/m2 − dimF(z) I/(I ∩ m2 ) − r

= dimF(z) n/n2 − r = dim Zeð,Γ


† fa .
− r = dim W

Here, dimF(z) n/n2 = dim Zeð,Γ because Zeð,Γ



is smooth at z.
fa is smooth for general closed point a ∈ Ar . Thus,
Hence, W

f ∼ ∼
fa : Wa −→ Wa = W a = V
ρ|W

is a resolution. Let Ve = W
fa . Then, Ve −→ V provides a resolution of V as in the
statement of the theorem, when X is defined over Spec Z..
Now, consider the general case when X is of finite presentation over a field k.
The field k is an extension of its unique prime (minimal) subfield F′ . This unique
prime subfield F′ is isomorphic to Q when k has characteristic zero or isomorphic to
Fp when k has characteristic p > 0. That is, F′ ∼
= F. Then, there exists an integral
scheme Y of finite presentation defined over Z, hence defined over F′ , such that

X/k = Y ×Spec F′ Spec k.

Take a singular point x of X, lying over a point y of Y . Replacing X and Y by


suitable open subsets containing x and y, respectively, if necessary, we may assume
that Y admits a resolution, as Y is defined over F′ .
Take a resolution Ye → Y over F′ . Then, we have the cartesian diagram

e := Ye ×Spec F′ Spec k
X / Ye

 
X = Y ×Spec F′ Spec k / Y.

The left vertical arrow is proper and birational.


By base change, we have that X e = Ye ×Spec F′ Spec k is smooth over k. Thus,
e → X is a resolution over k.
X 
GRASSMANNIANS AND SINGULARITIES 143

When the base field k has characteristic zero, the above theorem is well known
from Hironaka’s resolution [4]. When the base field k has positive characteristic, Ab-
hyankar [1] proved resolution of singularities for algebraic threefolds in characteristic
greater than 6.
In general, consider any fixed (singular) integral scheme X. Let K(X) be the
function field of X. Because the Riemann-Zariski space of K(X)/k is quasi-compact,
it remains to glue finitely many local resolutions of X, as obtained in this article,
to obtain a global resolution of X.
In this article, we approach resolution of singularity by performing blowups of (a
chart of) Gr3,E . It is convincible that certain parallel blowups exist for (P2 )n that
can also lead to achieve local resolution of singularity ([9]). (Indeed, when the author
began to work on resolution of singularity, he tried both approaches and switched
between the two for quite a while before settling down on the current approach via
Grassmannians.)

11. Geometric Resolution

Prior to de Jong’s geometric approach [3], resolutions of varieties in general di-


mensions are essentially done by finding good algorithms. In such an approach, one
isolates a set of bounded invariants and prove that after certain finite steps, such in-
variants improve strictly. As the invariants are bounded, the algorithm terminates.
These approaches are nicely presented in Kollar’s book [10].
According to [17], many moduli spaces or deformation spaces exhibit arbitrary
singularities. In other words, all singularities exist geometrically. Since singularities
exist for geometric reasons, one would wonder whether there should be geometric
ways to resolve them, avoiding pure algorithms on polynomials. Being philosophi-
cally optimistic, the author believes that every singular moduli admits a resolution,
in a specific relative sense, such that the resolution itself is also a moduli.
In other words, it would be desirable if the following problem can be answered in
some positive ways.

f
Problem 11.1. For any singular moduli space M, find another moduli space M
that only modifies the boundary objects of M such that every irreducible component
144 YI HU

f endowed with the reduced stack structure, is smooth, and all such irreducible
of M,
components meet transversally.

f should be obtained from the corresponding object of M by


Here, an object of M
adding certain extra data. The extra data should reduce the automorphisms of the
original object, and ideally, should remove all removable obstructions.
See Conjectures 5.4 and 5.5 of [6] for somewhat more precise formulations.

References
[1] Shreeram S. Abhyankar, Resolution of singularities of embedded algebraic surfaces, Springer
Monographs in Mathematics, Acad. Press (1966).
[2] Dan Abramovich, Michael Temkin, Jaroslaw Wlodarczyk, Functorial Embedded Resolution via
Weighted Blowings up, arXiv:1906.07106v3.
[3] Aise Johan de Jong, Smoothness, Semi-stability and Alterations, Publ. Math. IHES 83, (1996)
51-93.
[4] Heisuke Hironaka, Resolution of Singularities of an Algebraic Variety over a Field of Charac-
teristic zero, I, II, Ann. of Math., 2, 79 (1), (1964) 109-203, 205-326.
[5] Heisuke Hironaka, Resolution of Singularities in Positive Characteristics,
http://www.math.harvard.edu/∼hironaka/pRes.pdf
[6] Yi Hu, Relative Resolution and Its Applications, Proceedings of the Sixth International Con-
gress of Chinese Mathematicians. Vol. I, 467-488, Adv. Lect. Math. (ALM), 36, Int. Press,
Somerville, MA, 2017.
[7] Yi Hu, Hilbert Quotients and Hilbert Families of Grassmannians, Manuscript, Fall 2015.
[8] Yi Hu, Desingularizations of the Hilbert Quotients and Hilbert Families of Grassmannians, In
preparation.
[9] Yi Hu, Configuration Spaces and Singularities, Incomplete draft, Jan. 2015.
[10] János Kollár, Lectures on Resolutions of Singularities, Ann. of Math. Studies, No. 166, 2007.
[11] Laurent Lafforgue, Pavages des simplexes, schémas de graphes recollés et compactification des
PGLn+1
r /PGLr . (French. English summary) [Pavings of simplices, schemes of graphs that are
“pasted together” and compactification of PGLn+1
r /PGLr ] Invent. Math. 136 (1999), 233–271.
[12] −, Chirurgie des grassmanniennes. (French) [Surgery on Grassmannians], CRM Monogr. Ser.
19, Amer. Math. Soc., Providence, 2003.
[13] Seok Hyeong Lee and Ravi Vakil, Mnëv-Sturmfels Universality for Schemes, Clay Mathematics
Proceedings, Volume 18, 2013.
[14] Qing Liu, Algebraic Geometry and Arithmetic Curves, Oxford University Press, 2002.
GRASSMANNIANS AND SINGULARITIES 145

[15] Michael McQuillan, Gianluca Marzo, Very Fast, Very Functorial, and Very Easy Resolution
of Singularities, arXiv:1906.06745v1
[16] Nikolai Mnëv, The Universality Theorems on the Classification Problem of Configuration Va-
rieties and Convex Polytope Varieties, In Topology and Geometry, Rohlin Seminar, LNM 1346,
527-543, Springer, Berlin, 1988.
[17] Ravi Vakil, Murphy’s Law in Algebraic Geometry: Badly-Behaved Deformation Spaces, Inven.
Math. 164 569-590 (2006)
[18] Oscar Zariski, Local uniformization theorem on algebraic varieties, Ann. of Math. 41 (1940),
852 - 896.

Department of Mathematics, University of Arizona, USA.

You might also like