Download as pdf or txt
Download as pdf or txt
You are on page 1of 77

Note on Relation between Enhanced Ind-Sheaves and

Enhanced Subanalytic Sheaves∗


Yohei ITO†
arXiv:2109.13991v1 [math.AG] 28 Sep 2021

Abstract
In this paper, we will explain a relation between [DK16, Thm. 9.5.3] and [Kas16,
Thm. 6.2]. For this purpose, we will also explain a relation between enhanced ind-
sheaves and enhanced subanalytic sheaves.

1 Introduction
In [KS01], M. Kashiwara and P. Schapira introduced the notion of ind-sheaves and sub-
analytic sheaves to treat “sheaves” of functions with tempered growth conditions. Ind-
sheaves are defined as ind-objects of the category of sheaves of vector spaces with compact
support. Subanalytic sheaves are defined as sheaves on subanalytic sites. Moreover, the
authors proved that there exists a fully faithful functor from the category of subanalytic
sheaves to the category of ind-sheaves, and its essentially image is equal to the category
of ind-objects of R-constructible sheaves with compact support.
After a groundbreaking development in the theory of irregular meromorphic connec-
tions by K. S. Kedlaya [Ked10, Ked11] and T. Mochizuki [Moc09, Moc11], A. D’Agnolo
and M. Kashiwara introduced the notion of enhanced ind-sheaves extending the notion
of ind-sheaves and established the Riemann–Hilbert correspondence for analytic irregu-
lar holonomic D-modules in [DK16] as below (see [Ito21] for the algebraic case). Let X
be a complex manifold. Then there exists a fully faithful functor SolEX which is called
enhanced solution functor (see [DK16, Def. 9.1.1] for the definition) from the full triangu-
lated subcategory Dbhol (DX ) of the derived category of DX -modules consisting of objects
with holonomic cohomologies to the triangulated category EbR−c (ICX ) of R-constructible
enhanced ind-sheaves:

SolEX : Dbhol (DX )op ֒→ EbR−c (ICX ). (1.1)

Moreover, T. Mochizuki characterized its essentially image by the curve test [Moc16,
Thm. 12.1]. In [Ito20], the author defined C-constructability for enhanced ind-sheaves
and proved that they are nothing but objects of its essentially image. Namely, we obtain

2020 Mathematics Subject Classification: 18F10, 32C38, 35Q15, 32S60

Department of Mathematics, Faculty of Science Division II, Tokyo University of Science, 1-3, Kagu-
razaka, Shinjuku-ku, Tokyo, 162-8601, Japan. E-mail: yitoh@rs.tus.ac.jp

1
an equivalence of categories between the triangulated category Dbhol (DX ) and the one
EbC−c (ICX ) of C-constructible enhanced ind-sheaves:

SolEX : Dbhol (DX )op −→ EbC−c (ICX ).

At the 16th Takagi lecture, M. Kashiwara explained a similar result of (1.1) by using
“enhanced subanalytic sheaves” instead of enhanced ind-sheaves as below. We denote by
Db (Csub
X×R∞ ) the derived category of subanalytic sheaves on a bordered space X × R∞ .
Then there exists a fully faithful functor SolTX (see [Kas16, §5.4] for the definition) from
Dbhol (DX ) to Db (Csub
X×R∞ ):

SolTX : Dbhol (DX )op ֒→ Db (Csub


X×R∞ ). (1.2)

In this paper, we will explain a relation between (1.1) and (1.2). For this purpose,
we will also explain a relation between enhanced ind-sheaves and enhanced subanalytic
sheaves. Although it may be known by experts, it is not in the literature to our knowledge.
The main results of this paper are Theorems 3.18, 3.23, 3.40 and 3.41. One can summarize
these results in the following commutative diagram:

Db (Csub
X×R∞ )
❣❣❣3 O
T,sub ❣❣❣❣❣❣❣
SolX ❣❣❣❣❣❣ ❣❣
❣❣ ❣❣❣❣❣ RE,sub
X
❣❣ ❣❣
❣❣❣❣❣ ?
% ❣
 ❣❣❣❣❣
Dbhol (DX ) t
 ❖ SolE,sub
/ Eb (Csub )
R−c X ⊂ E (Csub
b
X ) g❖
❖❖❖ X O O  ❖t ❖❖❖❖❖ E
❖❖❖ ❖❖❖ ❖❖JX
❖❖❖ E ≀ JE E E
❖❖❖ ❖❖❖
❖❖❖ IX X
IX ≀ JX ❖❖❖ ❖❖❖
SolE ❖❖❖ ❖❖❖
IE
X
❖❖❖ ❖
❖'   X ❖❖❖
'
EbR−c (ICX ) ⊂ EbIR−c (ICX ) ⊂ Eb (ICX ).

See §2.7 for the definition of Eb (ICX ), §3.1 for the definition of Db (Csub
X×R∞ ), §3.3 for the
b sub E,sub b sub E
definitions of E (CX ), RX , §3.4 for the definitions of EIR−c (CX ), IX , JXE , EbR−c (ICX ),
E T,sub
Definition 3.22 for EbR−c (Csub
X ), Definition 3.34 for SolX , Definition 3.36 for SolX , Defi-
T,sub
nition 3.38 for SolX .

Acknowledgement
I would like to thank Dr. Tauchi of Kyushu University for many discussions and giving
many comments.

2
2 Preliminary Notions and Results
In this section, we briefly recall some basic notions and results which will be used in this
paper.

2.1 Subanalytic Subsets


The theory of semi-analytic and subanalytic sets originates in the work of S. Lojasiewicz
[Loj59, Loj64a, Loj64b] and has been elaborated by A. M. Gabrielov [Gab68], H. Hironaka
[Hiro73a, Hiro73b] and R. M. Hardt [Har75, Har77] for subanalytic sets. See also [BM88].
In this subsection, we just recall the definition of semi-analytic subsets and subanalytic
subsets and some properties.
Let M be a real analytic manifold and denote by AM the sheaf of rings of real analytic
functions. For an open subset U of M, we denote by S (AM (U)) the family of subsets of
M of the form p q
[ \
Xij ,
i=1 j=1

where each Xij is either {x ∈ U |fij (x) = 0} or {x ∈ U |fij (x) > 0}, fij ∈ AM (U).

Definition 2.1. A subset A of M is called semi-analytic if for any x ∈ M there exist an


open neighborhood U of x such that A ∩ U ∈ S (AM (U)).

Note that for any open subset U of M, S (AM (U)) is stable under finite intersection,
finite union and complement. Hence, a finite union and a finite intersection of semi-
analytic subsets are semi-analytic and the complement of a semi-analytic subset is also
semi-analytic. Furthermore, the closure and the interior of a semi-analytic subset are
semi-analytic, see [BM88, Cor. 2.8] for the details. Although, the inverse image of a semi-
analytic subset by a morphism of real analytic manifolds is semi-analytic, the operation of
direct image by a proper morphism of real analytic manifolds does not in general preserve
semi-analyticity. However, the class of subanalytic subsets is closed with respect to those
operations as below. Subanalytic subsets are “locally proper projections of semi-analytic
sets”.

Definition 2.2. A subset S of M is called subanalytic if for any x ∈ M there exist an


open neighborhood U of x, a real analytic manifold N and a relatively compact semi-
analytic subset A of M × N such that S ∩ U = pr1 (A), where pr1 : M × N → M is the
first projection.

From the basic properties of semi-analytic subsets, a finite union and a finite inter-
section of subanalytic subsets are subanalytic, and the closure of a subanalytic subset is
subanalytic. Furthermore, the complement (and thus the interior) of a subanalytic subset
is also subanalytic, see [BM88, Thm. 3.10] for details. Note also that the inverse image of
a subanalytic subset by a morphism of real analytic manifolds is subanalytic. Moreover,
the direct image of a subanalytic subset by a proper morphism of real analytic manifolds
is subanalytic, see [Hiro73a, Prop. 3.8] for details.

3
2.2 Sheaves on Sites
The theory of sheaves on topological spaces was created by J. Leray [Ler50] and this notion
was extended to sheaves on sites by A. Grothendieck [SGA4]. A site is a category endowed
with a “Grothendieck topology” on it. The Grothendieck topology was introduced by
A. Grothendieck in order to have a cohomology theory on algebraic varieties. In this
subsection, we shall briefly recall the definition of sheaves on sites and some properties
based on [KS01, §2]. See [KS06, §§16, 17, 18] for more general settings.
Let U be a universe. A set is called U-small if it is isomorphic to a set belonging to U.
A small category1 C is called U-category if for any objects X, Y of C the set HomC (X, Y )
is U-small. If moreover the family Ob(C) of objects (a set in bigger universe) of C is
U-small, then C is called U-small.
Definition 2.3. Let C be a U-small category admitting finite products and fiber products.
(1) For an object U of C, we denote by CU the category of arrows V → U. Namely, the
category CU is given by

Ob(CU ) := {(V, iV ) | V ∈ Ob(C), iV ∈ HomC (V, U)},



HomCU (V, iV ), (W, iW ) := {ϕ ∈ HomC (V, W ) | iW ◦ ϕ = iV }.

For simplicity, we sometimes write V → U instead of (V, iV ).

(2) For an object (V, iV ) ∈ Ob(CU ) and a subset S ⊂ Ob(CU ), let us set

V ×U S := {(V ×U W, prV ) | W ∈ S} ⊂ Ob(CV ),

where prV : V ×U W → V is the projection.

(3) For S1 , S2 ⊂ Ob(CU ), S1 is said to be a refinement of S2 if for any (V1 , iV1 ) ∈ S1


there exist (V2 , iV2 ) ∈ S2 and a morphism ϕ ∈ HomC (V1 , V2 ) such that iV2 ◦ ϕ = iV1 .
In such a situation we write S1  S2 .
Definition 2.4. Let C be a U-small category admitting finite products and fiber products.
A Grothendieck topology on C is the data associating to any U ∈ Ob(C) a family Cov(U)
of subsets of Ob(CU ) satisfying the axioms
(GT1) {idU } ∈ Cov(U),

(GT2) if S1 ∈ Cov(U) is a refinement of S2 ⊂ Ob(CU ), then S2 ∈ Cov(U),

(GT3) if S ∈ Cov(CU ), then V ×U S ∈ Cov(V ) for any (V, iV ) ∈ Ob(CU ),

(GT4) if S1 , S2 ⊂ Ob(CU ), S1 ∈ Cov(U) and V ×U S2 ∈ Cov(V ) for any (V, iV ) ∈ S1 , then


S2 ∈ Cov(U).
A subset S ∈ Cov(U) is called a covering of U.
A site X is a U-small category CX 2 admits finite products and fiber products and
endowed with a Grothendieck topology on CX .
1
In [KS06], a category means a small category.
2
Remark that the category CX is different from the category CX in Definition 2.3 (1).

4
Example 2.5. Let X be a topological space and denote by OpX the category of open
subsets of X. Namely, the category OpX is given by
(
{pt} (U ⊂ V ),
Ob(OpX ) := {U ∈ P(X) | U is open }, HomOpX (U, V ) :=
∅ (otherwise),

where P(X) is the power set of X. Note that for any U ∈ OpX we have (OpX )U = OpU .
We can endow OpX with the followingSGrothendieck topology: a subset S ⊂ Ob((OpX )U )
is a covering of U ∈ Ob(OpX ) if U = V ∈S V . We will keep the same symbol X to denote
its site.
Let k be a commutative unital ring and denote by Mod(k) the category of k-modules.
Definition 2.6. A presheaf of k-modules on a site X is nothing but a contravariant
functor from CX to Mod(k). We denote by s|V the image of s ∈ F (U) by the restriction
morphism F (U) → F (V ) associated a morphism V → U in CX .
Let us denote by Psh(kX ) the category of presheaves of k-modules on a site X.
Note that the category Psh(kX ) is abelian because the category Mod(k) is abelian.
Definition 2.7. Let X be a site. A presheaf F of k-modules on X is called separated if
for any U ∈ Ob(CX ) and any covering S ∈ Cov(U) of U, the natural morphism
!
Y Y
F (U) → ker F (V ) ⇒ F (V ′ ×U V ′′ )
V ∈S V ′ ,V ′′ ∈S

is a monomorphism. Here, the two arrows are associated to F (V ′ ) → F (V ′ ×U V ′′ ), and


F (V ′′ ) → F (V ′ ×U V ′′ ) and the kernel of the double arrow is the kernel of the difference
of the two arrows.
If moreover the natural morphism is an isomorphism, a presheaf F is a sheaf of k-
modules on X. Let us denote by Mod(kX ) the category of shaves of k-modules on X.
Note that the category Mod(kX ) is a full additive subcategory of Psh(kX ). Further-
more, it is abelian as below. To explain this, let us recall the sheaf associated with a
presheaf. Let X be a site. For a presheaf F of k-modules on X and U ∈ Ob(CX ), we set

F + (U) := lim F (S),


−→
S∈Cov(U )
Q Q 
′ ′′
where F (S) := ker V ∈S F (V ) ⇒ V ′ ,V ′′ ∈S F (V ×U V ) . Remark that the relation
 is a pre-order on Cov(U) and hence Cov(U) inherits a structure of a small category.
Moreover, Cov(U)op is filtrant. See [KS01, p. 19] for the details. Then we have the presheaf
op
F + : CX → Mod(k), U 7→ F + (U)

and a functor
(·)+ : Psh(kX ) → Psh(kX ), F 7→ F + .
Theorem 2.8 ([KS01, Thm. 2.1.7], see also [KS06, Thm. 17.4.7 (i), (iii)] for (3), (4)). Let
X be a site.

5
(1) The functor (·)+ : Psh(kX ) → Psh(kX ) is left exact.
(2) For any F ∈ Psh(kX ), F + is a separated presheaf.
(3) For any separated presheaf F , F + is a sheaf.
(4) The functor (·)++ : Psh(kX ) → Mod(kX ) is a left adjoint to the embedding functor
Mod(kX ) → Psh(kX ). Namely, for any F ∈ Psh(kX ) and any G ∈ Mod(kX ), we
have
HomPsh(kX ) (F , G) ≃ HomMod(kX ) (F ++ , G).
For a presheaf F ∈ Psh(kX ), the sheaf F ++ is called the sheaf associated with F . Let
us call the functor (·)++ the sheafification functor. Hence, we have:
Theorem 2.9 ([KS01, Thm. 2.1.10], see also [KS06, Thms. 17.4.9, 17.4.7 (iv), 18.1.6 (v)]).
Let X be a site.
(1) The category Mod(kX ) admits projective limits. More precisely, for any projective
system {Fi }i∈I of sheaves, its projective limit in Psh(kX ) is a sheaf and is a projec-
tive limit in Mod(kX ).
(2) The category Mod(kX ) admits inductive limits. More precisely, for any inductive
system {Fi }i∈I of sheaves, its inductive limit in Mod(kX ) is the sheaf associated
with its inductive limit in Psh(kX ).
(3) The category Mod(kX ) is abelian.
(4) The embedding functor Mod(kX ) → Psh(kX ) is fully faithful and left exact, and the
functor (·)++ : Psh(kX ) → Mod(kX ) is exact.
(5) Filtrant inductive limits in Mod(kX ) are exact.
(6) The U-category Mod(kX ) admits enough injective.
For M ∈ Mod(k), we denote by MX the sheaf associated with the presheaf U 7→ M
and call MX the constant sheaf with stalk M.
There are many operations for sheaves on sites as similar to classical sheaves.
Definition 2.10. Let X and Y be sites. A morphism f : X → Y of sites from X to
Y is a functor tf : CY → CX such that it commutes with fiber products and if for any
V ∈ Ob(CY ) and any S ∈ Cov(V ) then tf (S) ∈ Cov(tf (V )).
For a morphism f : X → Y of sites associated with a functor tf : CY → CX and
F ∈ Psh(kX ), G ∈ Psh(kY ), we set
(f∗ F )(V ) := F (tf (V )) for any V ∈ CY ,
−1
(fpre )G(U) := lim G(V ) for any U ∈ CX .
−→
t
U → f (V )

Then we have functors:


f∗ : Psh(kX ) → Psh(kY ),
−1
fpre : Psh(kY ) → Psh(kX ).
Note that if F ∈ Mod(kX ) then the presheaf f∗ F is a sheaf, see [KS06, Prop. 17.5.1].

6
Definition 2.11. Let f : X → Y be a morphism of sites. The direct image functor f∗
and inverse image functor f −1 are defined by

f∗ : Mod(kX ) → Mod(kY ), F 7→ f∗ F ,
++
f −1 : Mod(kY ) → Mod(kX ), F 7→ fpre
−1
F .

They have the following properties:


Proposition 2.12 ([KS01, Thm. 2.2.1], see also [KS06, Thm. 17.5.2]). Let f : X → Y be
a morphism of sites.
(1) The inverse image functor f −1 : Mod(kY ) → Mod(kX ) is left adjoint to the direct
image functor f∗ : Mod(kX ) → Mod(kY ). Namely, for any F ∈ Mod(kX ) and any
G ∈ Mod(kY ) we have

HomMod(kX ) (f −1 F , G) ≃ HomMod(kY ) (F , f∗ G).

(2) The direct image functor f∗ is left exact and commutes with small projective limits.

(3) The inverse image functor f −1 is exact and commutes with small inductive limits.
Note that for a site X and U ∈ Ob(CX ), we can endow (CX )U with the following
Grothendieck topology: for V ∈ Ob((CX )U ), a subset of Ob((CX )V ) is a covering of V if
it is a covering in CX . We denote by U its site and by iU : U → X the natural morphism
of sites from U to X associated with a functor t iU : CX → (CX )U , V 7→ (U × V, pr1 ),
where pr1 : U × V → U is the first projection. On the other hand, we have the morphism
jU : X → U of sites from X to U associated with a functor t jU : (CX )U → CX , V → V .
Hence we get functors:
−1
jU i−1
U
/ /
Mod(kU ) o Mod(kX ) o Mod(kU ).
jU ∗ iU ∗

Definition 2.13. Let X be a site and U ∈ Ob(CX ).


(1) The exact functor i−1
U : Mod(kX ) → Mod(kU ) is called the restriction functor to U.
For simplicity, we some times write F |U instead of i−1
U F for F ∈ Mod(kX ).

(2) The exact functor jU−1 : Mod(kU ) → Mod(kX ) is called the extension functor from
U. Let us set iU ! := jU−1 .

(3) We set ΓU F := iU ∗ i−1 −1


U F and FU := iU ! iU F for F ∈ Mod(kX ).

Clearly, the functor ΓU is left exact and the functor (·)U is exact, and there exist a
canonical morphism id → ΓU of functors and an isomorphism Γ(X; · ) ◦ ΓU ≃ Γ(U; · ).
Note that jU ∗ = i−1 −1
U and hence the functor iU ! is a left adjoint to the functor iU . Namely,
for any F ∈ Mod(kU ) and any G ∈ Mod(kX ) we have:

HomMod(kX ) (iU ! F , G) ≃ HomMod(kU ) (F , i−1


U G).

See [KS01, Prop. 2.3.4] for the details. Moreover we have:

7
Proposition 2.14 ([KS01, Prop. 2.3.6]). Let X be a site and U ∈ Ob(CX ). We assume
that for any V ∈ Ob(CX ), HomCX (V, U) has at most one element. Then we have
∼ ∼
(1) i−1 −1
U ◦ iU ∗ −→ id, id −→ iU ◦ iU !

(2) Functors iU ∗ and iU ! are fully faithful.

(3) There exist a canonical morphism (·)U → id of functors. Moreover, for any F ∈
Mod(kX ) the canonical morphism FU → F is a monomorphism .
If sites X and Y have terminal objects which are denoted by same symbols, and a
morphism f : X → Y of sites associated the functor tf : CY → CX satisfies tf (Y ) = X, for
any V ∈ CY we have the following commutative diagrams:
f f
XO / YO X / Y
iU iV jU jV
 
U f |U
/ V, U f |U
/ V,

where we set U := tf (V ) and a morphism f |U : U → Y of sites is induced by f . In this


case, we have
iU ! ◦ (f |U )−1 ≃ f −1 ◦ iV ! , (f |U )∗ ◦ i−1 −1
U ≃ iV ◦ f∗ .

By using the first one and the fact that f −1 kY ≃ kX , we have f −1 (kY V ) ≃ kXU .
For F , G ∈ Psh(kX ), we set

HomkX (F , F ) (U) := HomMod(kU ) (F |U , G|U ) for any U ∈ CX ,
 
pre
F ⊗ G (U) := F (U) ⊗ G(U) for any U ∈ CX .
kX k

Then we have bifunctors:

HomkX (·, ·) : Psh(kX )op × Psh(kX ) → Psh(kX ), (F , G) 7→ HomkX (F , G),


HomMod(kX ) (·, ·) : Psh(kX )op × Psh(kX ) → Mod(k), (F , G) 7→ HomMod(kX ) (F , G),
pre pre
(·) ⊗ (·) : Psh(kX ) × Psh(kX ) → Psh(kX ), (F , G) 7→ F ⊗ G.
kX kX

Note that if F , G ∈ Mod(kX ) then the presheaf HomkX (F , G) is a sheaf, see [KS06,
Prop. 17.7.1 (i)].
Definition 2.15. The internal hom functor and the tensor product functor are defined
by

HomkX (·, ·) : Mod(kX )op × Mod(kX ) → Mod(kX ), (F , G) 7→ HomkX (F , G),


HomMod(kX ) (·, ·) : Mod(kX )op × Mod(kX ) → Mod(k), (F , G) 7→ HomMod(kX ) (F , G),
 ++
pre
(·) ⊗ (·) : Mod(kX ) × Mod(kX ) → Mod(kX ), (F , G) 7→ F ⊗ G .
kX kX

We sometimes write Hom, ⊗ instead of HomkX , ⊗ , respectively.


kX

8
Remark that the internal hom functor is left exact in each variable and the tensor
product functor is right exact in each variable. They have the following properties.
Proposition 2.16 ([KS01, Props. 2.4.2, 2.4.3], see also [KS06, Prop. 17.7.3, Thms. 18.2.3,
Lem. 18.3.1]). Let f : X → Y be a morphism of sites and F , F1, F2 , H ∈
Mod(kX ), G, G1 , G2 ∈ Mod(kY ). Then we have
(1) for any U ∈ Ob(CX ), we have i−1 −1 −1
U Hom(F , G) ≃ Hom(iU F , iU G),

(2) HomkX (kX , F ) ≃ F ,


(3) kX ⊗ F ≃ F ,
(4) HomkX (F1 ⊗ F2 , H) ≃ HomkX (F1 , Hom(F2 , H)),

(5) HomkY (G, f∗ F ) −→ f∗ HomkX (f −1 G, F ),

(6) f −1 (G1 ⊗ G2 ) −→ f −1 G1 ⊗ f −1 G2 .
Let us recall the notion of (sheaves of) R-modules.
Definition 2.17. Let X be a site and k a commutative unital ring. A sheaf of k-algebras
is a sheaf R of k-modules on a site X such that for any U ∈ Ob(CX ), R(U) is a k-
algebra and for any morphism V → U in CX , the restriction morphism R(U) → R(V ) is
a morphism of k-algebras.
For sheaves R, S of k-algebras, a morphism of sheaves of k-algebras from R to S is a
morphism f : R → S of sheaves such that for any U ∈ CX the morphism f (U) : R(U) →
S(U) is a morphism of k-algebras.
Example 2.18. The constant sheaf kX on a site X is a sheaf of k-algebras.
A sheaf of k-algebras is called a kX -algebra. A sheaf of Z-algebras is simply called
a sheaf of rings. Moreover, for a kX -algebra R, we denote Rop the opposite kX -algebra
which is defined by Rop (U) := R(U)op for any U ∈ CX . Here, R(U)op is the opposite ring
of R(U).
Definition 2.19. Let X be a site and R a kX -algebra.
A presheaf of R-modules is a presheaf F ∈ Psh(kX ) such that for any U ∈ CX , F (U)
is a left R(U)-module and for any morphism V → U in CX , the restriction morphism
F (U) → F (V ) commutes with the action R, that is, (rs)|V = r|V s|V for any s ∈ F (U)
and r ∈ R(U).
For presheaves M, N of R-modules, a morphism of presheaves of R-modules from M
to N is a morphism ϕ : M → N of presheaves such that for any U ∈ CX the morphism
ϕ(U) : M(U) → N (U) is a morphism of R(U)-modules.
Let us denote by Psh(R) the category of presheaves of R-modules.
Clearly, the category Psh(R) is abelian.
Definition 2.20. Let X be a site and R be a kX -algebra.
A sheaf of R-modules is a presheaf of R-modules which is a sheaf of k-modules. For
simplicity, a sheaf of R-modules is called a R-module. A Rop -module is called a right
R-module.
A morphism of R-modules is a morphism of presheaves of R-modules. Let us denote
by Mod(R) the category of R-modules.

9
Note that the category Mod(R) has same properties of Theorem 2.9. For example,
it is abelian, see also [KS06, Thm. 18.1.6 (1)]. Moreover, the category Mod(R) is a
Grothendieck category, see [KS06, Def. 8.3.24, Thm. 18.1.6 (v)] for details. In particular,
it admits enough injectives, see also [KS06, Thm. 9.6.2].
Note also that the forgetful functor

f or : Mod(R) → Mod(kX )

is faithful and conservation but not fully faithful in general.


The sheafification functor (·)++ : Psh(kX ) → Mod(kX ) induces a functor

(·)++ : Psh(R) → Mod(R),

which will be denoted by the same symbol. Note that it is also left adjoint to the embed-
ding functor Mod(R) → Psh(R). See [KS06, Lem. 18.1.4] for details. Moreover, for any
U ∈ CX , we have functors:

(·)U : Mod(R) → Mod(R),


ΓU : Mod(R) → Mod(R),
Γ(U; · ) : Mod(R) → Mod(R(U))

and the functor (·)U is exact and the functors ΓU , Γ(U; · ) are left exact, see [KS06, §18.1]
for details.
For M, N ∈ Psh(R), we set

HomR (M, N ) (U) := HomPsh(R|U ) (M|U , N |U ) for any U ∈ CX ,
 
pre
M ⊗ N (U) := M(U) ⊗ N (U) for any U ∈ CX .
R R(U )

Then we have bifunctors:

HomPsh(R) (·, ·) : Psh(R)op × Psh(R) → Psh(kX ), (M, N ) 7→ HomR (M, N ),


HomPsh(R) (·, ·) : Psh(R)op × Psh(R) → Mod(k), (M, N ) 7→ HomPsh(R) (M, N ),
pre pre
(·) ⊗ (·) : Psh(Rop ) × Psh(R) → Psh(kX ), (M, N ) 7→ M ⊗ N .
R R

Note that if M, N ∈ Mod(R) then the presheaf HomR (M, N ) is a sheaf, see [KS06,
Lem. 18.2.1 (i)].

Definition 2.21. The internal hom functor and the tensor product functor are defined
by

HomR (·, ·) : Mod(R)op × Mod(R) → Mod(kX ), (M, N ) 7→ HomR (M, N ),


HomMod(R) (·, ·) : Mod(R)op × Mod(R) → Mod(kX ), (M, N ) 7→ HomMod(R) (M, N ),
 ++
pre
op
(·) ⊗ (·) : Mod(R ) × Mod(R) → Mod(kX ), (M, N ) 7→ M ⊗ N .
R R

10
Note that the internal hom functor is left exact in each variable and the tensor product
functor is right exact in each variable. Note also that if R is a sheaf of commutative rings
then HomR (M, N ), M ⊗ N ∈ Mod(R) for M, N ∈ Mod(R).
R
Moreover we have:
Proposition 2.22 ([KS06, Rem. 18.2.6]). Let Ri (i = 1, 2, 3, 4) be kX -algebras. Then we
have left exact functors in each variable
HomR1 (·, ·) : Mod(R1 ⊗ Rop 2 ) × Mod(R1 ⊗ R3 ) → Mod(R2 ⊗ R3 ),
op op op

HomMod(R1 ) (·, ·) : Mod(R1 ⊗ Rop


2 ) × Mod(R1 ⊗ R3 ) → Mod((R2 ⊗ R3 )(X))
op op op

and a right exact functor in each variable


(·) ⊗ (·) : Mod(R1 ⊗ Rop
2 ) × Mod(R2 ⊗ R3 ) → Mod(R1 ⊗ R3 ).
op op
R2

Moreover, for any i Mj ∈ Mod(Ri ⊗ Rop


j ) (i, j = 1, 2, 3, 4), there exist natural isomor-
phisms in Mod(R1 ⊗ R4 )op

   
1 M2 ⊗ 2 M3 ⊗ 3 M4 ≃ 1 M2 ⊗ 2 M3 ⊗ 3 M4 ,
R2 R3 R2 R3
 
HomR2 (2 M1 , HomR3 (3 M2 , 3 M4 )) ≃ HomR3 3 M2 ⊗ 2 M1 , 3 M4 .
R2

Let f : X → Y be a morphism of sites. Then f∗ R is a kY -algebra for a kX -algebra R,


see [KS06, Lem. 18.3.1 (i)]. However, f −1 S is not necessarily a ring for a kY -algebra S.
If the morphism f : X → Y of sites associated to the functor tf : CY → CX is left exact,
that is, the functor tf is left exact, then the sheaf f −1 S is a kX -algebra. Moreover, in this
case, the direct image and the inverse image functors f∗ , f −1 in the sense of sheaves of
k-modules induce the functors:
f∗ : Mod(f −1 S) → Mod(S),
f −1 : Mod(S) → Mod(f −1 S).
Note also that the direct image functor f∗ is left exact and the inverse image functor f −1
is exact, see [KS06, Lem. 18.3.1 (ii)] for the details of these results. Moreover, we have:
Proposition 2.23 ([KS06, Lems. 18.3.1 (ii) (c), 18.3.2]). Let f : X → Y be a left exact
morphism of sites and S be a kY -algebra. Then we have
(1) for any M ∈ Mod(S op ) and any N ∈ Mod(S),
 
−1
f M ⊗ N ≃ f −1 M ⊗ f −1 N ,
S f −1 S

(2) for any M ∈ Mod(S) and any N ∈ Mod(f −1 S),


f∗ Homf −1 S (f −1 M, N ) ≃ HomS (M, f∗ N ),
HomMod(f −1 S) (f −1 M, N ) ≃ HomMod(S) (M, f∗ N ).
In particular, the functor f −1 is left adjoint to the functor f∗ .

11
At the end of this subsection, let us briefly recall the derived category D∗ (Mod(R))
(∗ = ub, +, −, b) of sheaves on a site. Let R be a kX -algebra which is flat as a kX -
module. We shall often write for short D(R) (resp. D∗ (R)) instead of Dub (Mod(R))
(resp. D∗ (Mod(R))). For a left exact morphism f : Z → X of sites and U ∈ CX , we have
(derived) functors

Rf∗ : D(f −1 R) −→ D(R),


f −1 : DR) −→ D(f −1R),
L
(·) ⊗ (·) : D(Rop ) × D(R) −→ D(kX ),
R
RHomR (·, ·) : D(R)op × D(R) −→ D(kX ),
RHomR (·, ·) : D(R)op × D(R) −→ D(k),
(·)U : D(R) −→ D(R),
RΓU : D(R) −→ D(R),
RΓ(U; · ) : D(R) −→ D(R(U)).

Note that these functors have same properties of the case of classical sheaves. We shall
skip the explanation of it. See [KS06, §§18.4, 18.5, 18.6] for the details.

2.3 Subanalytic Sheaves


The notion of subanalytic sheaves are introduced by M. Kashiwara and P. Schapira in
[KS01] to treat functions with growth conditions in the formalism of sheaves. They
are defined as sheaves on a subanalytic site and have examples such as the subanalytic
sheaf of tempered distributions, the one of tempered C ∞ -functions, the one of Whitney
C ∞ -functions. In this subsection, let us briefly recall the notion of subanalytic sheaves.
Reference are made to [KS01, §6], [Pre08], see also Subsection 2.2.
Throughout this subsection, we assume that k is a field. Let M be a real analytic
manifold and denote by Opsub M the category of subanalytic open subsets of M. We can
endow OpM with the following Grothendieck topology: a subset S ⊂ Ob((Opsub
sub
M )U ) is
a covering of U ∈ Opsub M if for any compact subset K of M there exists a finite subset
sub
S0 ⊂ S of S such that U ∩ K = (∪V ∈S0 V ) ∩ K. We denote by M such a site and call
it the subanalytic site.
A subanalytic sheaf (resp. presheaf) of k-modules on M is a sheaf (resp. presheaf)
of k-modules on the subanalytic site M sub . We shall write Mod(ksub sub
M ) (resp. Psh(kM ))
instead of Mod(kM sub ) (resp. Psh(kM sub )). For simplicity, we shall call subanalytic sheaves
(resp. presheaves) of k-modules subanalytic sheaves (resp. presheaves).
Note that the forgetful functor induces an equivalence of categories (see e.g. [Pre08,
Rem. 1.1.2])

Mod(kM sub ) −→ Mod(kM sub, c ),
where the site M sub, c is the category Opsub,
M
c
of relatively compact subanalytic open subsets
with the following Grothendieck topology: a subset S ⊂ Ob((Opsub M )U ) is a covering of
U ∈ Opsub,
M
c
if for any compact subset K of M there exists a finite subset S0 ⊂ S of S
such that U ∩ K = (∪V ∈S0 V ) ∩ K. Hence we have:

12
Proposition 2.24 ([Pre08, Prop. 1.1.4]). Let M be a real analytic manifold and F a
presheaf on M sub, c . If F satisfies the following conditions then it is a sheaf on M sub, c ,

that is, it is an object of Mod(ksub
M ) −→ Mod(kM sub, c )
(1) F (∅) = 0,
(2) for any relatively compact subanalytic open subsets U, V , the sequence
0 / F (U ∪ V ) / F (U) ⊕ F (V ) / F (U ∩ V )
s ✤ / (s|U , s|V )
(t, u) ✤ / t|U ∩V − u|U ∩V
is exact.
Clearly, we have the natural morphism ρM : M → M sub of sites. Hence, there exist
functors
ρM ∗ : Mod(kM ) → Mod(ksub
M ),
ρ−1 sub
M : Mod(kM ) → Mod(kM ).

We have already seen in Subsection 2.2, a pair (ρ−1


M , ρM ∗ ) is an adjoint pair, the functor ρM ∗
−1
is left exact and the functor ρM is exact. Note that there exists a canonical isomorphism

ρ−1
M ◦ ρM ∗ −→ id of functors. Hence the functor ρM ∗ is a fully faithful functor, see [Pre08,
Prop. 1.1.8]. Note that the restriction of ρM ∗ to the category of R-constructible sheaves is
R−c
exact, see [Pre08, Prop. 1.1.9]. Let us denote by ρM ∗ the exact functor. Moreover there
exists a left adjoint functor ρM ! of the functor ρ−1
M which is defined by the following: for
any F ∈ Mod(kM ), ρM ! F is given by the sheaf associated to the presheaf U 7→ Γ U ; F |U ,
where U is the closure of U in M. See [Pre08, Prop. 1.1.13 (i)] for the details. Note that
the functor ρM ! is exact and fully faithful, and there exist the canonical isomorphism

id −→ ρ−1 M ◦ ρM ! of functors and an isomorphism

Homsub (ρM ! F , G) ≃ Hom(F , ρ−1


M G)

in Mod(kM ) for any F ∈ Mod(kM ) and any G ∈ Mod(ksub M ), see [Pre08, Props. 1.1.14,
1.1.5]. Here, the functor Homsub is defined as below.
Let f : M → N be a morphism of real analytic manifolds. Then it induces the mor-
phism f : M sub → N sub of subanalytic sites from M sub to N sub which is denoted by the
same symbol. We have already seen in Subsection 2.2, there exist many functors
(·)++ : Psh(ksub sub
M ) → Mod(kM ),
f∗ : Mod(ksub sub
M ) → Mod(kN ),
f −1 : Mod(ksub sub
N ) → Mod(kM ),
(·) ⊗ (·) : Mod(ksub sub sub
M ) × Mod(kM ) → Mod(kM ),
Ihomsub (·, ·) : Mod(ksub
M )
op
× Mod(ksub sub
M ) → Mod(kM ),
HomMod(ksub
M )
(·, ·) : Mod(ksub
M )
op
× Mod(ksub
M ) → Mod(k),

(·)U : Mod(ksub sub


M ) → Mod(kM ),
ΓU : Mod(ksub sub
M ) → Mod(kM ),
Γ(U; · ) : Mod(ksub
M ) → Mod(k).

13
In this paper, we shall write Ihomsub instead of the internal hom functor on Mod(ksub M ),
that is, HomkM sub (·, ·) in the notion of Subsection 2.2. Let us set Homsub := ρ−1
M ◦Ihom sub
.
Moreover, we have a functor which is called the proper direct image functor:

f!! : Mod(ksub sub


M ) → Mod(kN ), f!! F := lim f F ≃ lim f Γ F ,
−→ ∗ U −→ ∗ K
U K

where U ranges through the family of relatively compact open subanalytic subsets and
K ranges through the family of subanalytic compact subsets. Note that the functor f!!
is left exact and if f : M → N is proper then we have f∗ ≃ f!! . Moreover, these functor
have many properties, see [Pre08] for the details. We shall skip the explanation of it.
We shall write D∗ (ksub ∗
M ) (∗ = ub, +, −, b) instead of D (kM sub ). Then there exist (de-
rived) functors of them, we have already seen in Subsection 2.2 for several ones. Moreover,
the functors below are well defined, see [Pre08, Cor. 2.3.3] for the functors Rf∗ , Rf!! :

RρM ∗ : Db (kM ) ֒→ Db (ksub


M ),
R−c b b sub
ρM ∗ : DR−c (kM ) ֒→ D (kM ),
ρ−1 b sub b
M : D (kM ) −→ D (kM ),
ρM ! : Db (kM ) ֒→ Db (ksub
M ),
(·)++ : Db (Psh(ksub b sub
M )) −→ D (kM ),
Rf∗ : Db (ksub b sub
M ) −→ D (kN ),
f −1 : Db (ksub b sub
N ) −→ D (kM ),
Rf!! : Db (ksub b sub
M ) −→ D (kN ),
(·) ⊗ (·) : Db (ksub b sub b sub
M ) × D (kM ) −→ D (kM ),
RIhomsub (·, ·) : Db (ksub
M )
op
× Db (ksub + sub
M ) −→ D (kM ),
RHomsub (·, ·) : Db (ksub
M )
op
× Db (ksub +
M ) −→ D (kM ),
RHomMod(ksub
M )
(·, ·) : Db (ksub
M )
op
× Db (ksub +
M ) −→ D (k),

(·)U : Db (ksub b sub


M ) −→ D (kM ),
RΓU : Db (ksub b sub
M ) −→ D (kM ),
RΓ(U; · ) : Db (ksub b
M ) −→ D (k),

where U is a subanalytic open subset of M. Note that the functor Rf!! admits a right
adjoint functor, see [Pre08, §2.4] for the details:

f ! : Db (ksub b sub
N ) −→ D (kM ).

Moreover, these functor have many properties as similar to classical sheaves. In this
subsection, we just recall the following properties, see [Pre08] for the details.

Theorem 2.25. Let f : M → N be a morphism of real analytic manifolds and F , F1 , F2 ∈


Db (ksub b sub b b b
M ), G, G1 , G2 ∈ D (kN ), K ∈ D (kM ), L ∈ D (kN ), J ∈ DR−c (kM ).

(1) RρM ∗ Hom(ρ−1


M F , K) ≃ RIhom
sub
(F , RρM ∗ K),
RHomsub (ρM ! K, F ) ≃ RHom(K, ρ−1
M F ),

14
(2) RIhomsub (Rf!! F , G) ≃ Rf∗ RIhomsub (F , f ! G),
Rf∗ RIhomsub (f −1 G, F ) ≃ RIhomsub (G, Rf∗ F ),

(3) RIhomsub (F1 ⊗ F2 , F ) ≃ RIhomsub F1 , RIhomsub (F2 , F ) ,

(4) f −1 (F1 ⊗ F2 ) ≃ f −1 F1 ⊗ f −1 F2 ,
Rf!! (F ⊗ f −1 G) ≃ Rf!! F ⊗ G,

(5) f ! RIhomsub (G1 , G2 ) ≃ RIhomsub (f −1 G1 , f ! G2 ),

(6) for a cartesian diagram


f′
M ′ sub / N ′ sub
g′ g
 
M sub f
/ N sub

we have g −1 f!! F ≃ f!!′ g ′−1 F , g ! f∗ F ≃ f∗′ g ′! F ,

(7) f ! (G ⊗ ρM ! L) ≃ f ! G ⊗ ρM ! f −1 L,

(8) RIhomsub (J , F ) ⊗ ρM ! K ≃ RIhomsub (J , F ⊗ ρM ! K).

At the end of this subsection, let us summarize the commutativity of the various
functors in the table below. Here, “ ◦ ” means that the functors commute, and “ × ” they
do not.

⊗ f −1 Rf∗ f! Rf!!
Rρ∗ × × ◦ ◦ ×

ρ∗R−c ◦ ◦ ◦ ◦ ×

ρ−1 ◦ ◦ ◦ × ◦
ρ! ◦ ◦ × × ×

2.4 Ind-Sheaves
The theory of (classical) sheaves is not well suited to the study of various objects in Analy-
sis which are not defined by local properties, such as holomorphic functions with tempered
growth. The notion of ind-sheaves was introduced by M. Kashiwara and P. Schapira in
[KS01] to treat functions with such as growth conditions in the formalism of sheaves. In
this subsection, we shall briefly recall the notion of ind-sheaves. References are made to
[KS01, KS06].
Let M be a good topological space, that is, a topological space which is locally
compact, Hausdorff, countable at infinity and has finite soft dimension. We denote by

15
Modc (kM ) the category of sheaves of k-vector spaces on M with compact support. An
ind-sheaf of k-vector spaces on M is an ind-object of Modc (kX ), that is, inductive limit

“ lim ”Fi := lim HomModc (kM ) ( · , Fi )


−→ −→
i∈I i∈I

of a small filtrant inductive system {Fi }i∈I in Modc (kX ). We call an ind-sheaf of k-vector
spaces an ind-sheaf for simplicity. Let us denote IkM the category of ind-sheaves of k-
vector spaces on M. Note that it is abelian. Note also that there exists a natural exact
embedding ιM : Mod(kM ) → IkM of categories. We sometimes omit it. It has an exact
left adjoint αM , that has in turn an exact fully faithful left adjoint functor βM :
ιM
/
Mod(kM ) o αM
/ IkM .
βM

The category IkM does not have enough injectives. Nevertheless, we can construct the
derived category Db (IkM ) for ind-sheaves and the Grothendieck six operations among
them. We denote by ⊗ and RIhom the operations of tensor product and internal hom
respectively. If f : M → N be a continuous map, we denote by f −1 , Rf∗ , f ! and Rf!! the
operations of inverse image, direct image, proper inverse image and proper direct image,
respectively. We set also RHom := αM ◦ RIhom. We thus have (derived) functors

(·) ⊗ (·) : Db (IkM ) × Db (IkM ) −→ Db (IkM ),


ιM : Db (kM ) −→ Db (IkM ), RIhom(·, ·) : Db (IkM )op × Db (IkM ) −→ D+ (IkM ),
αM : Db (IkM ) −→ Db (kM ), RHom(·, ·) : Db (IkM )op × Db (IkM ) −→ D+ (kM ),
βM : Db (kM ) −→ Db (IkM ), RHom(·, ·) : Db (IkM )op × Db (IkM ) −→ D+ (k),
Rf∗ : Db (IkM ) −→ Db (IkN ),
Z (·) : Db (IkM ) −→ Db (IkM ), f −1 : Db (IkN ) −→ Db (IkM ),
RΓZ : Db (IkM ) −→ Db (IkM ), Rf!! : Db (IkM ) −→ Db (IkN ),
f ! : Db (IkN ) −→ Db (IkM ),

where Z is a locally closed subset of M. These functor have many properties as similar
to classical sheaves, see [KS01, KS06] for details. Moreover, we can consider the notion
of ind-sheaves with ring action. We shall skip the explanation of them.
At the end of this subsection, let us summarize the commutativity of the various
functors in the table below. Here, “ ◦ ” means that the functors commute, and “ × ” they
do not.

⊗ f −1 Rf∗ f! Rf!!
ι ◦ ◦ ◦ ◦ ×
α ◦ ◦ ◦ × ◦
β ◦ ◦ × × ×

16
2.5 Relation between Ind-sheaves and Subanalytic Sheaves
In this subsection, we shall recall the relation between ind-sheaves and subanalytic sheaves.
Reference are made to [KS01, §§6.3, 7.1] and [Pre13, §A.2].
Let M be a real analytic manifold and k a field. We denote by ModcR−c (kM ) the
abelian category of R-constructible sheaves on M with compact support and by IR−c kM
the category of ind-objects of ModcR−c (kM ). Moreover let us denote by DbIR−c (IkM ) the
full triangulated subcategory of Db (IkM ) consisting of objects whose cohomologies are
contained in IR−c kM . A functor JM : IkM → Mod(ksub M ) is defined by
!
JM “ lim ”Fi := lim ρM ∗ Fi .
−→ −→
i∈I i∈I

Note that for any F ∈ IkM the subanalytic sheaf JM F is given by Γ(U; JM F ) =
HomIkM (ιM kU , F ) for each open subanalytic subset U. Note also that the functor JM
is left exact and admits a left adjoint
IM : Mod(ksub
M ) → IkM

which is fully faithful, exact and commutes with lim. Then we have (derived) functors
−→
IM : D (kM ) → Db (IkM ),
b sub

RJM : Db (IkM ) → Db (ksub


M )

and a pair (IM , RJM ) of functors is an adjoint pair and there exists a canonical isomor-

phism id −→ RJM ◦ IM . Moreover, we have
RJM RIhom(IM (·), ·) ≃ RIhomsub (·, RJM (·)).
Theorem 2.26 ([KS01, Thm. 6.3.5], see also [Pre13, A.2.1]). There exists an equivalence
of abelian categories:
IM
/
Mod(ksub
M ) o
∼ IR−c kM .
JM

Furthermore, there exists an equivalence of triangulated categories:


IM
/
Db (ksub
M ) o
∼ DbIR−c (IkM ).
RJM

At the end of this subsection, let us summarize the commutativity between the various

functors and functors I, RJ. We will denote by λM : DbIR−c (IkM ) −→ Db (ksub
M ) the inverse
b sub ∼ b
functor of IM : D (kM ) −→ DIR−c (IkM ). Let f : M → N be a morphism of real analytic
manifolds. Then we have:

αM ◦ I M ≃ ρ−1
M , RJM ◦ ιM ≃ ρM ∗ ,
IM ◦ ρM ! ≃ βM , ρ−1
M ◦ RJM ≃ αM , IM ◦ ρMR−c
∗ ≃ ιM |ModR−c (kM ) ,
IM ◦ f −1 ≃ f −1 ◦ IN , RJM ◦ βM ≃ ρM ! , λM ◦ f ≃ f −1 ◦ λN ,
−1

Rf!! ◦ IM ≃ IN ◦ Rf!! , Rf∗ ◦ RJM ≃ RJN ◦ Rf∗ , Rf!! ◦ λM ≃ λN ◦ Rf!! ,


IM ◦ f ! ≃ f ! ◦ IN , RJM ◦ f ! ≃ f ! ◦ RJN , λM (· ⊗ ·) ≃ λM (·) ⊗ λM (·).
IM (· ⊗ ·) ≃ IM (·) ⊗ IM (·),

17
2.6 Ind-Sheaves on Bordered Spaces
In this subsection, we shall recall the notion of bordered spaces and the notions of ind-
sheaves and subanalytic sheaves on bordered spaces. Reference are made to [DK16, §3]
for the details of bordered spaces and ind-sheaves on bordered spaces.
Throughout this subsection, we assume that k is a field.

2.6.1 Bordered Spaces


A bordered space is a pair M∞ = (M, M̌ ) of a good topological space M̌ and an open
subset M ⊂ M̌ . A morphism f : (M, M̌ ) → (N, Ň) of bordered spaces is a continuous map
f : M → N such that the first projection M̌ × Ň → M̌ is proper on the closure Γf of the
graph Γf of f in M̌ × Ň . Note that a morphism fˇ: M̌ → Ň such that fˇ(M) ⊂ N induces
a morphism of bordered spaces from (M, M̌ ) to (N, Ň ). The category of good topological
spaces is embedded into that of bordered spaces by the identification M = (M, M).
For a locally closed subset Z of M, we set Z∞ := (Z, Z), where Z is the closure of
Z in M̌ , and denote by iZ∞ : Z∞ → M∞ a morphism of bordered spaces induced by the
natural embedding iZ : Z ֒→ M. Moreover, let us denote by jM∞ : M∞ → M̌ a morphism
of bordered spaces induced by the natural embedding jM : M ֒→ M̌.

2.6.2 Ind-Sheaves on Bordered Spaces


A quotient category

Db (IkM∞ ) := Db (IkM̌ )/Db (IkM̌ \M )

is called the category of ind-sheaves on M∞ = (M, M̌ ), where Db (IkM̌ \M ) is identified with


its essentially image in Db (IkM̌ ) by the fully faithful functor RiM̌\M !! ≃ RiM̌ \M ∗ . Here
iM̌ \M : M̌ \M → M is the closed embedding. An object of Db (IkM∞ ) is called ind-sheaves
on M∞ . The quotient functor

q : Db (IkM̌ ) → Db (IkM∞ )

has a left adjoint l and a right adjoint r, both fully faithful, defined by

l(qF ) := ιM̌ kM ⊗ F,
r(qF ) := RIhom(ιM̌ kM , F ).

Moreover they induce equivalences of categories


∼ ∼
l : Db (IkM∞ ) −→ {F ∈ Db (IkM̌ ) | ιM̌ kM ⊗ F −→ F },
∼ ∼
r : Db (IkM∞ ) −→ {F ∈ Db (IkM̌ ) | F −→ RIhom(ιM̌ kM , F )},

respectively.
For a morphism f : M∞ → N∞ of bordered spaces, we have the Grothendieck opera-
tions ⊗, RIhom, Rf∗ , Rf!! , f −1 , f ! , see [DK16, Definitions 3.3.1 and 3.3.4] for the defini-
tion. Moreover, these functor have many properties as similar to classical sheaves. We
shall skip the explanation of them. See [DK16, §3.3] for the details.

18
−1 !
Note that the functors jM ∞
≃ jM ∞
: Db (IkM̌ ) → Db (IkM∞ ) are isomorphic to the quo-
tient functor and the functor RjM∞ !! : Db (IkM∞ ) → Db (IkM̌ ) (resp. RjM∞ ∗ : Db (IkM∞ ) →
Db (IkM̌ )) is isomorphic to the functor l (resp. r).
Note also that there exists an embedding functor
−1
ιM∞ : Db (kM ) ֒→ Db (IkM∞ ), F 7→ jM −1
ι j F ( ≃ jM
∞ M̌ M !
ι RiM ∗ F )
∞ M̌

which has an exact left adjoint

αM∞ : Db (IkM∞ ) → Db (kM ), F 7→ i−1 −1


M αM̌ RjM∞ !! F ( ≃ iM αM̌ RjM∞ ∗ F )

that has in turn an exact fully faithful left adjoint functor


−1
βM∞ : Db (kM ) ֒→ Db (IkM∞ ), F 7→ jM −1
β j F ( ≃ jM
∞ M̌ M !
β RiM ∗ F ).
∞ M̌

It is clear that the quotient category Db (kM∞ ) := Db (kM̌ )/Db (kM̌ \M ) is equivalent
to the derived category Db (kM ) of the abelian category Mod(kM ). We sometimes write
Db (kM∞ ) for Db (kM ), when considered as a full subcategory of Db (IkM∞ ).

2.7 Enhanced Ind-Sheaves


It was a long-standing problem to generalize the Riemann–Hilbert correspondence for
regular holonomic D-modules to the (not necessarily regular) holonomic D-module case.
One of the difficulties was that we could not find an appropriate substitute of the target
category of the regular case. In [DK16], the authors solved it by using the enhanced
ind-sheaves. In this subsection, we shall recall a more general notion of enhanced ind-
sheaves on topological spaces, that is, the notion of enhanced ind-sheaves on bordered
spaces. Reference are made to [KS16a] and [DK19]. Moreover we also refer to [DK16]
and [KS16b] for the notion of enhanced ind-sheaves on good topological spaces.
Let M∞ = (M, M̌ ) be a bordered space. We set R∞ := (R, R) for R := R⊔{−∞, +∞},
and let t ∈ R be the affine coordinate. We consider the morphisms of bordered spaces
p1 , p2 , µ π
M∞ × R2∞ −−−−−→ M∞ × R∞ −→ M∞

given by the maps p1 (x, t1 , t2 ) := (x, t1 ), p2 (x, t1 , t2 ) := (x, t2 ), µ(x, t1 , t2 ) := (x, t1 + t2 )


and π(x, t) := x. Then the convolution functors for ind-sheaves on M∞ × R∞
+
(·) ⊗ (·) : Db (IkM∞ ×R∞ ) × Db (IkM∞ ×R∞ ) → Db (IkM∞ ×R∞ ),
RIhom+ (·, ·) : Db (IkM∞ ×R∞ )op × Db (IkM∞ ×R∞ ) → Db (IkM∞ ×R∞ )

are defined by
+
F1 ⊗ F2 := Rµ!! (p−1 −1
1 F1 ⊗ p2 F2 ),
RIhom+ (F1 , F2 ) := Rp1∗ RIhom(p−1 !
2 F1 , µ F2 ),

where F1 , F2 ∈ Db (IkM∞ ×R∞ ). Now we define the triangulated category of enhanced


ind-sheaves on a bordered space M∞ by

Eb (IkM∞ ) := Db (IkM∞ ×R∞ )/π −1Db (IkM∞ ).

19
The quotient functor
QM∞ : Db (IkM∞ ×R∞ ) → Eb (IkM∞ )
has fully faithful left and right adjoints

LEM∞ , REM∞ : Eb (IkM∞ ) → Db (IkM∞ ×R∞ )

defined by
+
LEM∞ (QM∞ (F )) := ιM∞ ×R∞ (k{t≥0} ⊕ k{t≤0} ) ⊗ F,
REM∞ (QM∞ (F )) := RIhom+ (ιM∞ ×R∞ (k{t≥0} ⊕ k{t≤0} ), F )

for F ∈ Db (IkM∞ ×R∞ ). We sometimes denote QM∞ (resp. LEM∞ , REM∞ ) by Q (resp.
LE , RE ) for short. Moreover they induce equivalences of categories
 
∼ + ∼
E b b
LM∞ : E (IkM∞ ) −→ F ∈ D (IkM∞ ×R∞ ) | ιM∞ ×R∞ (k{t≥0} ⊕ k{t≤0} ) ⊗ F −→ F ,
n o
E b ∼ b ∼ +
RM∞ : E (IkM∞ ) −→ F ∈ D (IkM∞ ×R∞ ) | F −→ RIhom (ιM∞ ×R∞ (k{t≥0} ⊕ k{t≤0} ), F ) ,

where {t ≥ 0} stands for {(x, t) ∈ M × R | t ≥ 0}and {t ≤ 0} is defined similarly. Then


we have the following standard t-structure on Eb (ICM∞ ) which is induced by the standard
t-structure on Db (ICM∞ ×R∞ )

E≤0 (ICM∞ ) = {K ∈ Eb (ICM∞ ) | LEM∞ K ∈ D≤0 (ICM∞ ×R∞ )},


E≥0 (ICM∞ ) = {K ∈ Eb (ICM∞ ) | LEM∞ K ∈ D≥0 (ICM∞ ×R∞ )}.

We denote by
Hn : Eb (ICM∞ ) → E0 (ICM∞ )
the n-th cohomology functor, where we set E0 (ICM∞ ) := E≤0 (ICM∞ ) ∩ E≥0(ICM∞ ).
The convolution functors for enhanced ind-sheaves on bordered spaces are well defined.
We denote them by the same symbols:
+
(·) ⊗ (·) : Eb (IkM∞ ) × Eb (IkM∞ ) → Eb(IkM∞ ),
RIhom+ (·, ·) : Eb(IkM∞ )op × Eb (IkM∞ ) → Eb(IkM∞ ).

For a morphism f : M∞ → N∞ of bordered spaces, we can define also the operations

Ef∗ , E!! : Eb (IkM∞ ) → Eb (IkN∞ ),


Ef −1 , Ef ! : Eb (IkN∞ ) → Eb (IkM∞ ),

for enhanced ind-sheaves
 on bordered spaces. For example, we define Ef ∗ QM ∞ F :=
b
QN∞ RfR∞ ∗ F for F ∈ D (IkM∞ ×R∞ ), where fR∞ : M∞ × R∞ → N∞ × R∞ is the natural
morphism of bordered spaces associated to f . The other operations are defined similarly.
Note that there exists a morphism Ef!! → Ef∗ of functors Eb (IkM∞ ) → Eb (IkN∞ ) and it
is an isomorphism if f is proper.

20
Moreover we have external hom functors
RIhomE (·, ·) : Eb (IkM∞ )op × Eb (IkM∞ ) → Db (IkM∞ ),
RHomE (·, ·) : Eb (IkM∞ )op × Eb (IkM∞ ) → Db (kM ),
RHomE (·, ·) : Eb (IkM∞ )op × Eb (IkM∞ ) → Db (k),
which are defined by
RIhomE (K1 , K2 ) := Rπ∗ RIhom(LEM∞ K1 , LEM∞ K2 ),
RHomE (K1 , K2 ) := αM∞ RIhomE (K1 , K2 ),

RHomE (K1 , K2 ) := RΓ M; RHomE (K1 , K2 ) ,
for K1 , K2 ∈ Eb (IkM∞ ). For F ∈ Db (IkM∞ ) and K ∈ Eb (IkM∞ ) the objects
π −1 F ⊗ K := QM∞ (π −1 F ⊗ LEM∞ K),

RIhom(π −1 F, K) := QM∞ RIhom(π −1 F, REM∞ K) .
in Eb (IkM∞ ) are well defined. Hence, we have functors
π −1 (·) ⊗ (·) : Db (IkM∞ ) × Eb (IkM∞ ) → Db (IkM∞ ×R∞ ),
RIhom(π −1 (·), ·) : Db (IkM∞ )op × Eb (IkM∞ ) → Db (IkM∞ ×R∞ ).
We set !
kEM∞ := QM∞ q “ lim ” ιM̌ ×R k{t≥a} ∈ Eb (IkM∞ ).
−→
a→+∞

Then we have a natural embedding functor


eM∞ : Db (IkM∞ ) → Eb (IkM∞ ), F 7→ kEM∞ ⊗ π −1 F.
E
Let us define ωM ∞
:= eM∞ (ιM∞ ωM ) ∈ Eb (IkM∞ ) where ωM ∈ Db (kM∞ )(≃ Db (kM )) is the
dualizing complex, see [KS90, Definition 3.1.16] for the details. Then we have the Verdier
duality functor for enhanced ind-sheaves on bordered spaces
DEM∞ : Eb (IkM∞ )op → Eb (IkM∞ ), K 7→ RIhom+ (K, ωM
E

).
Note that there exists an isomorphism DEM∞ (eM∞ ιM∞ F ) ≃ eM∞ (DM F ) in Eb (IkM∞ ) for
any F ∈ Db (kM ). Let i0 : M∞ → M∞ × R∞ be a morphism of bordered spaces induced
by x 7→ (x, 0). We set
IshM∞ := i!0 ◦ REM∞ : Eb (IkM∞ ) → Db (IkM∞ ),
shM∞ := αM∞ ◦ IshM∞ : Eb (IkM∞ ) → Db (kM )
and call them the ind-sheafification functor, the sheafification functor, for enhanced ind-
sheaves on bordered spaces, respectively. Note that a pair (eM∞ , IshM∞ ) is an adjoint
∼ ∼
pair and there exist isomorphisms F −→ IshM∞ eM∞ F for F ∈ Db (IkM ) and F −→
shM∞ eM∞ ιM∞ F for F ∈ Db (kM ). See [DK21, §3] for details.
For a continuous function ϕ : U → R defined on an open subset U ⊂ M, we set the
exponential enhanced ind-sheaf by
+
EϕU |M∞ := kEM∞ ⊗ QM∞ ιM∞ ×R∞ k{t+ϕ≥0} ,
where {t + ϕ ≥ 0} stands for {(x, t) ∈ M × R | x ∈ U, t + ϕ(x) ≥ 0}.

21
3 Main Results
The main theorem of this paper is Theorems 3.18, 3.23, 3.40 and 3.41.

3.1 Subanalytic Sheaves on Real Analytic Bordered spaces


The notion of subanalytic sheaves on bordered spaces was introduced by M. Kashiwara.
Although it has already been explained in [Kas16, §§3.4–3.7], we will explain in detail in
this subsection again3 .
A real analytic bordered space is a bordered space M∞ = (M, M̌ ) such that M̌ is a real
analytic manifold and M is an open subanalytic subset. A morphism f : (M, M̌) → (N, Ň )
of real analytic bordered spaces is a morphism of bordered spaces such that the graph
Γf of f is a subanalytic subset of M̌ × Ň . Note that a morphism fˇ: M̌ → Ň of real
analytic manifolds such that fˇ(M) ⊂ N induces a morphism of bordered spaces of real
analytic bordered spaces from (M, M̌ ) to (N, Ň). The category of real analytic manifolds
is embedded into that of real analytic bordered spaces by the identification M = (M, M).
Let M∞ = (M, M̌ ) be a real analytic bordered space. We denote by Opsub M∞
(resp. Opsub,
M∞
c
) the category of open subsets of M which are subanalytic (resp. subanalytic
and relatively compact) in M̌ . It is clear that they admit finite products and fiber prod-
ucts. Although the following proposition are well known by experts, we will prove in this
section.
Proposition 3.1. Let M∞ = (M, M̌ ) be a real analytic bordered space.
(1) The category Opsub
M∞ can be  endowed with the following Grothendieck
 topology: a
subset S ⊂ Ob (Opsub M∞ U) is a covering of U ∈ Ob Op sub
M∞ if for any compact
[
subset K of M̌ there exists a finite subset S ′ ⊂ S of S such that K ∩U = K ∩ V.
V ∈S ′

sub, c
(2) The category Op M∞ can be  endowed with the following Grothendieck
 topology: a
sub, c sub, c
subset S ⊂ Ob (OpM∞ )U is a covering of U ∈ Ob OpM∞ if for any compact
[
subset K of M̌ there exists a finite subset S ′ ⊂ S of S such that K ∩U = K ∩ V.
V ∈S ′

Proof. Since the proof of (2) is similar, we only prove (1). It is clear that the condition
(GT1) of Definition 2.4 is satisfied.
Let us prove that the condition (GT2) of Definition 2.4 is satisfied. Let S1 ⊂
Ob((Opsub sub
M∞ )U ) be a covering of U ∈ Ob(OpM∞ ) which is a refinement of S2 ⊂
Ob((Opsub
M∞ )U ) and K a compact subset of M̌ . Then there exists a finite subset S1 of

S1 such that K ∩ U = K ∩ (∪V ∈S1′ V ). Since S1 is a refinement of S2 , for any V1 ∈ S1′ ,


there exists V2 ∈ S2 such that V1 ⊂ V2 . We write S2′ for a subset of S2 consisting of such
elements. Since ∪V1 ∈S1′ V1 ⊂ ∪V2 ∈S2′ V2 , we have
[ [
K∩U =K ∩ V1 ⊂ K ∩ V2 (⊂ K ∩ U)
V1 ∈S1′ V2 ∈S2′

3
In [Kas16], the author introduced the notion of subanalytic sheaves on subanalytic bordered spaces.
In this paper, we shall only consider them on real analytic bordered spaces.

22
and hence K ∩ U = K ∩ (∪V2 ∈S2′ V2 ). This implies that S2 is a covering of U and the
condition (GT2) of Definition 2.4 is satisfied.
Let us prove that the condition (GT3) of Definition 2.4 is satisfied. Let S ⊂
Ob((Opsub sub sub
M∞ )U ) be a covering of U ∈ Ob(OpM∞ ), V ∈ Ob((OpM∞ )U ) and K a compact
subset of M̌ . Then there exists a finite subset S ′ of S such that K ∩ U = K ∩ (∪W ∈S ′ W ).
Then the set V ×U S ′ := {V ∩ W | W ∈ S ′ } is a finite subset of V ×U S and we have
!
[ [ [
K∩ f=K∩
W (V ∩ W ) = K ∩ W ∩ V = (K ∩ U) ∩ V = K ∩ V.
f ∈V ×U S ′
W W ∈S ′ W ∈S ′

This implies that V ×U S is a covering of V and the condition (GT3) of Definition 2.4 is
satisfied.
Let us prove that the condition (GT4) of Definition 2.4 is satisfied. Let S1 be a covering
of U ∈ Ob(Opsub sub
M∞ ), S2 a subset of Ob((OpM∞ )U ) such that for any V ∈ S1 , V ×U S2 is a
covering of V , and K a compact subset of M̌. Then there exists a finite subset S1′ of S1
such that K ∩ U = K ∩ (∪V ∈S1′ V ). Moreover, for any V ∈ S1′ (⊂ S1 ) there exists a finite
subset SfV of V ×U S2 such that K ∩ V = K ∩ (∪W f ∈Sf
V
f ). For any V ∈ S ′ let us set
W 1

SV := {W ∈ S2 | V ∩ W ∈ Sf
V },

fV . Then S ′ is a finite subset of S2 and we have


and set S2′ := ∪V ∈S1′ S 2
  
[ [ [ [ [ [ [
K∩U = (K∩V ) = K ∩  f  = K∩
W f = K∩
W (V ∩W ).
V ∈S1′ V ∈S1′ f ∈Sf
W V
V ∈S1′ f ∈Sf
W V
V ∈S1′ W ∈SV

Since ∪V ∈S1′ ∪W ∈SV (V ∩W ) ⊂ ∪V ∈S1′ ∪W ∈SV W = ∪W ∈S2′ W , we have K ∩U ⊂ K ∩∪W ∈S2′ W


and hence [
K∩U =K∩ W.
W ∈S2′

This implies that S2 is a covering of U and the condition (GT4) of Definition 2.4 is
satisfied.
sub sub,c sub, c
Let us denote by M∞ (resp. M∞ ) the site Opsub
M∞ (resp. OpM∞ ) with the above
Grothendieck topology. Note that the forgetful functor induces an equivalence of cate-
gories:

f or : Mod(kM∞ sub ) −→ Mod(k sub, c ).
M∞
sub
Indeed, for F ∈ Mod(kM∞
sub, c ) and for V ∈ Ob(Op
M∞ ) we set

e ) :=
F(V lim F (U ∩ V ).
←−
U ∈Ob(Opsub,
M∞ )
c

Then we have Fe ∈ Mod(kM∞ e


sub ) and the functor F 7→ F is the inverse functor of the

forgetful functor.

23
Definition 3.2. Let M∞ = (M, M̌) be a real analytic bordered space. An object
of Mod(kM∞sub ) ≃ Mod(k sub, c ) is called a subanalytic sheaf on M∞ . We shall write
M∞
sub
Mod(kM∞ ) instead of Mod(kM∞ sub ) ≃ Mod(k sub, c ), for simplicity.
M∞
The category of subanalytic sheaves on M∞ is abelian, and admits projective limits
and inductive limits, by Theorem 2.9 (1), (2) and (3). Moreover, subanalytic sheaves can
be characterized as follows:
Lemma 3.3. Let M∞ = (M, M̌ ) be a real analytic bordered space. We assume that a
sub, c
presheaf F on M∞ satisfies the following two conditions:
(1) F (∅) = 0,
(2) for any subanalytic open subsets U, V which are relatively compact in M̌, a sequence
0 / F (U ∪ V ) / F (U) ⊕ F (V ) / F (U ∩ V )
s ✤ / (s|U , s|V )
(t, u) ✤ / t|U ∩V − u|U ∩V
is exact.
sub, c
Then the presheaf F is a sheaf on M∞ .
 
Proof. First, let us prove that, for any finite family {Ui }ni=1 in Ob Opsub,
M∞
c
, a sequence
 
S
n
α L
n β L
0 / F Ui / F (Ui ) / F (Ui ∩ Uj )
i=1 i=1 1≤i<j≤n

s ✤ / (s|Ui )ni=1

(ti )ni=1 ✤ / ti |Ui ∩Uj − tj |Ui ∩Uj 1≤i<j≤n

is exact. We shall prove it by induction on n. The cases of n = 1, 2 are obvious. Let us


assume that n > 2 and the results of the case of k ≤ n − 1 have been proved. We shall
consider the following commutative diagram:
0 0

  n−1  n−1  
S
n
α′ S β′ S
0 / F Ui / F Ui ⊕ F (Un ) / F Ui ∩ Un
i=1 i=1 i=1
▲▲▲
▲▲▲
▲▲α▲ α′′ γ
▲▲▲
▲▲&  
L
n β ′′′ L
n−1
F (Ui ) / F (Ui ∩ Un )
i=1 PPP i=1
PPP
PPP β
PPP
β ′′ PPP
PPP
 PP
L L' 
F (Ui ∩ Uj ) / F (Ui ∩ Uj ),
1≤i<j≤n−1 1≤i<j≤n

24
where
 
α′ (s) := s|Sn−1 Ui , s|Un , β ′ (t, u) := t|Sn−1 Ui ∩Un − u|Sn−1 Ui ∩Un ,
i=1 i=1 i=1
′′ n−1  ′′ n

α (t, u) := (t|Ui )i=1 ,u , β ((ti )i=1 ) := ti |Ui ∩Uj − tj |Ui∩Uj 1≤i<j≤n−1 ,
n−1 n−1
γ(v) := (v|Ui∩Un )i=1 , β ′′′ ((ti )ni=1 ) := (ti |Ui ∩Un − tn |Ui∩Un )i=1 .

Remark that α = α′′ ◦ α′ and β = β ′′ ⊕ β ′′′ . By the second assumption, a sequence


n−1  n−1  n−1 
S α′ S β′ S
0 / F Ui ∪ Un / F Ui ⊕ F (Un ) / F Ui ∩ Un
i=1 i=1 i=1

n−1 
S L
n−1
is exact. Moreover, by the induction hypothesis, γ : F Ui ∩ Un → F (Ui ∩ Un ) is
i=1 i=1
injective, and a sequence
n−1 
/ F
S L
n−1 L
0 Ui / F (Ui ) / F (Ui ∩ Uj )
i=1 i=1 1≤i<j≤n−1

n−1
s′ ✤ / (s′ |Ui )i=1
n−1 ✤ 
(ti )i=1 / ti |Uij − tj |Uij 1≤i<j≤n−1

is exact, and hence the following sequence is exact:


n−1 
S α′′ L
n−1 β ′′ L
0 / F Ui ⊕ F (Un ) / F (Ui ) ⊕ F (Un ) / F (Ui ∩ Uj ).
i=1 i=1 1≤i<j≤n−1

Since α′ , α′′ are injective, α is also injective. It is clear that β ◦ α = 0 and hence we have
Im α ⊂ Ker β. Let (ti )ni=1 ∈ Ker β. Then we have β ((ti )ni=1 ) =0 andhence β ′′ ((ti )ni=1 ) =
S
n−1
β ′′′ ((ti )ni=1 ) = 0. Since Im α′′ = Ker β ′′ , there exists (t, u) ∈ F Ui ⊕ F (Un ) such that
n−1
 i=1
n−1

α′′ (t, u) = (ti )i=1 , tn . Moreover, since γ(β ′ (t, u)) = β ′′′ (α′′ (t, u)) = β ′′′ (ti )i=1 , tn =
β ′′′ ((ti )ni=1 ) = 0 and γ is injective,
 n we

 have β (t, u) = 0. So that, (t, u) ∈ Ker β = Im α ,
′ ′

S
and hence there exists s ∈ F Ui such that α′(s) = (t, u). Moreover we have α(s) =
i=1
n−1
α′′ (α′ (s)) = α′′ (t, u) = ((ti )i=1 , tn ) = (ti )ni=1 , and hence Im α = Ker β.
sub,c
Let us prove that the presheaf F is a sheaf on M∞ , that is, for any covering S of
U ∈ Ob Opsub,c M∞ a sequence

ϕ Q ψ Q
0 / F (U) / F (V ) / F (V ′ ∩ V ′′ )
V ∈S V ′ ,V ′′ ∈S

s ✤ / (s|V )V ∈S
(tV )V ∈S ✤ / (tV ′ |V ′ ∩V ′′ − tV ′′ |V ′ ∩V ′′ )V ′ ,V ′′ ∈S

25
is exact. Since U is relatively compact in M̌ and the definition of Grothendieck
S topology
on M∞ sub, c
, there exists a finite subset {Ui }ni=1 of S such that U = ni=1 Ui , and hence a
sequence

α L
n β L
0 / F (U) / F (Ui ) / F (Ui ∩ Uj )
i=1 1≤i<j≤n

s ✤ / (s|Ui )ni=1

(ti )ni=1 ✤ / ti |Ui ∩Uj − tj |Ui ∩Uj 1≤i<j≤n

is exact, as already proved. Let s ∈ F (U) and assume that ϕ(s) = 0. Then we have
(sV )V ∈S = 0, and hence (sUi )ni=1 = 0. This implies that α(s) = 0. Since α is injective, we
have s = 0 and hence ϕ is injective. It is clear that  ψ ◦ ϕ = 0 and hence Im ϕ ⊂ Ker ψ.
Let (tV )V ∈S ∈ Ker ψ. Then we have ψ (tV )V ∈S = (tV ′ |V ′ ∩V ′′ − tV ′′ |V ′ ∩V ′′ )V ′ ,V ′′ ∈S = 0.

In particular, we have ti |Ui ∩Uj − tj |Ui ∩Uj 1≤i<j≤n = 0. This implies that β ((tUi )ni=1 ) = 0.
Since (tUi )ni=1 ∈ Ker β = Im α, there exists s ∈ F (U) such that for any i = 1, . . . , n one
has s|Ui = tUi . Let us prove that for any V ∈ S \ {Ui | i = 1, . . . , n} one has s|V = tV .
Since (tV )V ∈S ∈ Ker ψ, in particular, for any i = 1, . . . , n we have

(s|V − tV )|V ∩Ui = (s|V )|V ∩Ui − tV |V ∩Ui = (s|Ui )|V ∩Ui − tV |V ∩Ui = tUi |V ∩Ui − tV |V ∩Ui = 0.
L
n
Moreover, a map F (V ) → F (V ∩Ui ) is injective, as already proved, and hence s|V = tV .
i=1
This implies that ϕ(s) = (tV )V ∈S and we have Im ϕ = Ker ψ.
Therefore, the proof is completed.
From now on, let us describe various functors for subanalytic sheaves. As already seen
in Theorem 2.8, there exists the left exact functor

(·)+ : Psh(kM∞
sub ) → Psh(kM sub )

where for any F ∈ Psh(ksub


M∞ ) and any U ∈ Ob (OpM∞ ) we set

F + (U) := lim F (S)


−→
S∈Cov(U )

and F (S) is the kernel of a map


Y Y
F (V ) → F (V ′ ∩ V ′′ ), (tV )V ∈S 7−→ (tV ′ |V ′ ∩V ′′ − tV ′′ |V ′ ∩V ′′ )V ′ ,V ′′ ∈S .
V ∈S V ′ ,V ′′ ∈S

Moreover, we have the exact functor

(·)++ : Psh(kM∞
sub ) → Mod(k
sub
M∞ )

which is called the sheafification functor and is the left adjoint of natural embedding
functor Mod(ksub
M∞ ) → Psh(kM∞sub ).

26
sub
Note that there exists the natural morphism ρM∞ : M → M∞ of sites. We sometimes
write ρ instead of ρM∞ , for simplicity. Then we have the functors
ρM∞ ∗ : Mod(kM ) → Mod(ksub
M∞ ),
ρ−1 sub
M∞ : Mod(kM∞ ) → Mod(kM )

which are defined by


(ρM∞ ∗ F )(U) := F (U)

for any F ∈ Mod(kM ) and any U ∈ Ob OpsubM∞ , and by
++
ρ−1 −1
M∞ G := ρpre G

for any G ∈ Mod(ksub −1


M∞ ). Here, the presheaf ρpre G is given by

ρ−1
pre G(V ) := lim G(U)
−→
V ⊂U

for any open subset V of M and U ranges through the family Ob Opsub M∞ . Note that a
pair (ρ−1 −1
M∞ , ρM∞ ∗ ) is an adjoint pair, the functor ρM∞ ∗ is left exact and the functor ρM∞
is exact, as already seen in Proposition 2.12. Note also that there exists a canonical

isomorphism ρ−1M∞ ◦ ρM∞ ∗ −→ id of functors and an isomorphism

ρM∞ ∗ Hom(ρ−1
M∞ F , G) ≃ Hom(F , ρM∞ ∗ G)

in Mod(kM ) for any F ∈ Mod(ksub M∞ ) and any G ∈ Mod(kM ). Hence the functor ρM∞ ∗
is a fully faithful functor. A sheaf F ∈ Mod(kM ) is said to be R-constructible on M∞
if jM ! F ∈ ModR−c (kM̌ ) where jM : M ֒→ M̌ is the natural embedding. Let us denote by
ModR−c (kM∞ ) the category of R-constructible sheaves on M∞ . Note that the restriction
of ρM∞ ∗ to ModR−c (kM∞ ) is exact. We shall denote by
R−c
ρM ∞∗
: ModR−c (kM∞ ) → Mod(ksub
M∞ )

the exact functor. By the same argument as [Pre08, Prop. 1.1.14], we have the left adjoint
functor
ρM∞ ! : Mod(kM ) → Mod(ksub
M∞ )

of ρ−1
M∞ which is given by the following: for any F ∈ Mod(kM ), ρ F is the sheaf
 M∞ !
associated to the presheaf U 7→ Γ(U; F |U ), where U ∈ Ob OpsubM∞ and U is the closure
of U in M. Note that the functor ρM ! is exact and fully faithful, and there exist the

canonical isomorphism id −→ ρ−1M∞ ◦ ρM∞ ! of functors and an isomorphism

ρ−1 −1
M∞ Hom(ρM∞ ! F , G) ≃ Hom(F , ρM∞ G)

in Mod(kM ) for any F ∈ Mod(kM ) and any G ∈ Mod(ksub M∞ ).


Let f : M∞ → N∞ be a morphism of real analytic bordered spaces. Then we obtain a
sub sub
morphism of sites from M∞ to N∞ associated with a map Opsub sub
N∞ → OpM∞ , V 7→ f
−1
(V ).
sub sub
We shall denote by the same symbol f : M∞ → N∞ . As already seen in Definition 2.11
and Proposition 2.12, there exist the direct image functor and the inverse image functor
f∗ : Mod(ksub sub
M∞ ) → Mod(kN∞ ),
f −1 : Mod(ksub sub
N∞ ) → Mod(kM∞ )

27
which are defined by

(f∗ F )(V ) := F (f −1 (V )) ≃ HomMod(ksub


M )
(kf −1 (V ) , F )


for any F ∈ Mod(ksub sub
M∞ ) and any V ∈ Ob OpM∞ , and by

f −1 G := (fpre
−1
G)++

for any G ∈ Mod(ksub −1


N∞ ). Here, the presheaf fpre G is given by

−1
(fpre G)(U) := lim G(V )
−→
U ⊂f −1 (V )

 
for any U ∈ Ob Opsub M∞ and V ranges through the family {W ∈ Ob OpN∞
sub
| U ⊂
−1 −1
f (W )}. We have already seen in Proposition 2.12, a pair (f , f∗ ) is an adjoint pair,
the direct image functor f∗ is left exact and the inverse image functor f −1 is exact.
For any U ∈ Ob Opsub M∞ , we obtain a real analytic bordered space (U, M̌). We shall
denote by U∞ = (U, M̌)4 the real analytic bordered space and denote by iU∞ : U∞ →
M∞ a morphism of real analytic bordered spaces induced by the natural embedding
sub sub
U ֒→ M. Hence, we obtain a morphism iU∞ : U∞ → M∞ of real analytic bordered
spaces. Moreover, we have a morphism from M∞ to U∞ induced by the map Opsub
sub sub
U∞ →
sub 5
OpM∞ , V 7→ V , and the inverse image functor of it which is denoted by iU∞ !! is left
adjoint to the functor i−1 sub
U∞ , namely, for any F ∈ Mod(kU∞ ) and any G ∈ Mod(kM∞ ), we
sub

have
HomMod(ksubM )
(iU∞ !! F , G) ≃ HomMod(ksub
U )
(F , i−1
U∞ G).
∞ ∞

We have already seen in Subsection 2.2, there exist functors

(·)U∞ := iU∞ !! i−1 sub sub


U∞ : Mod(kU ∞ ) → Mod(kM∞ ),
ΓU∞ := iU∞ ∗ i−1 sub sub
U∞ : Mod(kM∞ ) → Mod(kM∞ ),
Γ(U; · ) : Mod(ksub
M∞ ) → Mod(k).

As already seen in Subsection 2.2, we have the tensor product functor and the internal
hom functor6

(·) ⊗ (·) : Mod(ksub sub sub


M∞ ) × Mod(kM∞ ) → Mod(kM∞ ),
Ihomsub (·, ·) : Mod(ksub
M∞ )
op
× Mod(ksub sub
M∞ ) → Mod(kM∞ )

which are defined by


 pre ++
F ⊗ G := F ⊗ G ,

Ihomsub (F , G) (U) := HomMod(ksub (i−1 F , i−1
U ) U∞ U∞ G)

4
It is clear that (U, U ) ≃ (U, M̌ ) as bordered spaces.
5
We shall write iU∞ !! instead of iU∞ ! in the notion of Subsection 2.2.
6
In this paper, we shall write Ihomsub instead of the internal hom functor on Mod(ksub
M∞ ), that is,
HomkM sub (·, ·) in the notion of Subsection 2.2.

28
 pre
for any F , G ∈ Mod(kM ) and any U ∈ Ob Opsub
M∞ . Here, the presheaf F ⊗ G is given by
 pre 
F ⊗ G (U) := F (U) ⊗ G(U)

for any U ∈ Ob Opsub
M∞ . Moreover we have external hom functors

Homsub (·, ·) := ρ−1


M∞ Ihom
sub
: Mod(ksub
M∞ )
op
× Mod(ksub
M∞ ) → Mod(kM ),
HomMod(ksub
M )
(·, ·) : Mod(ksub
M∞ )
op
× Mod(ksub
M∞ ) → Mod(k).

For a morphism f : M → N of real analytic manifolds, the proper direct image functor
f!! : Mod(ksub sub
M∞ ) → Mod(kN∞ )

is given by
(f!! F )(V ) := lim HomMod(ksub (kf −1 (V ) , FU∞ )
−→ M∞ )
U
   
sub,c sub,c
for any V ∈ Ob OpN∞ . Here, U ranges through the family Ob OpM∞ such that
f −1 (V ) ∩ U → V is proper, U is the closure of U in M. Note that the functor f!! is left
exact and if f : M → N is proper then we have f∗ ≃ f!! .
These functor have many properties as similar to classical sheaves. We shall skip the
explanation of it. 
We shall write D∗ (ksub
M∞ ) (∗ = ub, +, −, b) instead of D

Mod ksub
M∞ and denoted
b b
by DR−c (kM∞ ) the full triangulated subcategory of D (kM ) consisting of objects whose
cohomologies are contained in ModR−c (kM∞ ). Recall that the abelian category Mod(ksub
M∞ )
admits enough injectives, see Theorem 2.9 (6). For a morphism f : M∞ → N∞ of real
analytic bordered spaces and U ∈ Ob Opsub M∞ , there exist (derived) functors:

RρM∞ ∗ : Db (kM ) ֒→ Db (ksub


M∞ ),
R−c
ρM ∞∗
: DbR−c (kM∞ ) ֒→ Db (ksub
M∞ ),
ρ−1 b sub b
M∞ : D (kM∞ ) −→ D (kM ),
ρM∞ ! : Db (kM ) ֒→ Db (ksub
M∞ ),
(·)++ : Db (Psh(ksub b sub
M∞ )) −→ D (kM∞ ),
Rf∗ : Db (ksub b sub
M∞ ) −→ D (kN∞ ),
f −1 : Db (ksub b sub
N∞ ) −→ D (kM∞ ),
Rf!! : Db (ksub b sub
M∞ ) −→ D (kN∞ ),
(·) ⊗ (·) : Db (ksub b sub b sub
M∞ ) × D (kM∞ ) −→ D (kM∞ ),
RIhomsub (·, ·) : Db (ksub
M∞ )
op
× Db (ksub + sub
M∞ ) −→ D (kM∞ ),
RHomsub (·, ·) : Db (ksub
M∞ )
op
× Db (ksub +
M∞ ) −→ D (kM∞ ),
RHomMod(ksub
M )
(·, ·) : Db (ksub
M∞ )
op
× Db (ksub +
M∞ ) −→ D (k),

(·)U : Db (ksub b sub


M∞ ) −→ D (kM∞ ),
RΓU : Db (ksub b sub
M∞ ) −→ D (kM∞ ),
RΓ(U; · ) : Db (ksub b
M∞ ) −→ D (k).

29
Note that the functor Rf!! admits a right adjoint functor, see e.g. [Kas16, §3.6]:

f ! : Db (ksub b sub
N∞ ) −→ D (kM∞ ).

Note also that if f : M → N is topological submersive then f ! (·) ≃ ρM∞ ∗ ωM/N ⊗ f −1 (·),
where ωM/N ∈ Db (kM ) is the relative dualizing complex which is defined in [KS90,
Def. 3.1.16 (i)].
Moreover, these functor have many properties as similar to classical (subanalytic)
sheaves.
Proposition 3.4. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces and
F , F1, F2 ∈ Db (ksub b sub b b b
M∞ ), G, G1 , G2 ∈ D (kN∞ ), K ∈ D (kM ), L ∈ D (kN ), J ∈ DR−c (kM∞ ).

(1) RρM∞ ∗ Hom(ρ−1


M∞ F , K) ≃ RIhom
sub
(F , RρM∞ ∗ K),
RHomsub (ρM∞ ! K, F ) ≃ RHom(K, ρ−1
M∞ F ),

(2) RIhomsub (Rf!! F , G) ≃ Rf∗ RIhomsub (F , f ! G),


Rf∗ RIhomsub (f −1 G, F ) ≃ RIhomsub (G, Rf∗ F ),

RIhomsub (F1 ⊗ F2 , F ) ≃ RIhomsub F1 , RIhomsub (F2 , F ) ,

(3) f −1 (F1 ⊗ F2 ) ≃ f −1 F1 ⊗ f −1 F2 ,
Rf!! (F ⊗ f −1 G) ≃ Rf!! F ⊗ G,
f ! RIhomsub (G1 , G2 ) ≃ RIhomsub (f −1 G1 , f ! G2 ),
(4) for a cartesian diagram
′ f′ ′
M∞ / N∞
g′ g
 
M∞ f
/ N∞

we have g −1 Rf!! F ≃ Rf!!′ g ′−1F , g ! Rf∗ F ≃ Rf∗′ g ′! F ,


(5) f ! (G ⊗ ρM∞ ! L) ≃ f ! G ⊗ ρM∞ ! f −1 L,
RIhomsub (J , F ) ⊗ ρM∞ ! K ≃ RIhomsub (J , F ⊗ ρM∞ ! K),
(6) there exist the commutativity of the various functors

⊗ f −1 Rf∗ f! Rf!!
Rρ∗ × × ◦ ◦ ×

ρ∗R−c ◦ ◦ ◦ ◦ ×

ρ−1 ◦ ◦ ◦ × ◦
ρ! ◦ ◦ × × ×

where “ ◦ ” means that the functors commute, and “ × ” they do not.

30
Since the proof of this proposition is similar to the case of classical sheaves, we shall
skip the proof.
At the end of this subsection, we shall describe a relation between ind-sheaves on
M∞ and subanalytic sheaves on M∞ . Recall that jM∞ : M∞ → M̌ is a morphism of real
analytic bordered spaces associated to the natural embedding jM : M ֒→ M̌ . Then for
any F ∈ Mod(ksub sub
M∞ ) and any G ∈ Mod(kM̌ ) we have

−1 ∼
jM ∞
RjM∞ ∗ F −→ F ,
−1 ∼
jM ∞
RjM∞ !! F ←− F ,
−1
jM∞ !! jM ∞
G ≃ ρM̌ ∗ kM ⊗ G,
−1
jM∞ ∗ jM ∞
G ≃ Ihomsub (ρM̌ ∗ kM , G),
−1
and hence functors jM ∞
, RjM∞ !! , RjM∞ ∗ induce equivalences of categories:
∼ ∼
e
j : Db (ksub b sub
M∞ ) −→ {F ∈ D (kM̌ ) | ρM∞ ∗ kM ⊗ F −→ F }

≃ {F ∈ Db (ksub
M̌ ) | F −→ Ihom
sub
(ρM∞ ∗ kM , F )}.

Recall that the category of ind-sheaves on M∞ is defined by

Db (IkM∞ ) := Db (IkM̌ )/Db (IkM̌ \M )

and there exist (derived) functors

IM̌ : Db (ksub b
M̌ ) → D (IkM̌ ),
RJM̌ : Db (IkM̌ ) → Db (ksub
M̌ ).

Let us consider the following functors

IM∞ : Db (ksub b
M∞ ) → D (IkM∞ ), F 7→ qIM̌ RjM∞ !! F ,
RJM∞ : Db (IkM∞ ) → Db (ksub −1
M∞ ), F 7→ jM∞ RJM̌ RjM∞ ∗ F,

where q : Db (IkM̌ ) → Db (IkM∞ ) is the quotient functor.


We denote by DbIR−c (IkM∞ ) the full triangulated subcategory of Db (IkM∞ ) consisting
of objects such that RjM∞ !! F ∈ DbIR−c (IkM̌ ). Then the following lemma follows from
[KS01, Lem. 7.1.3].

Lemma 3.5. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces associated


with a morphism fˇ: M̌ → Ň of real analytic manifolds. The functors below are well
defined:

(1) ιM∞ : DbR−c (kM ) → DbIR−c (IkM∞ ),

(2) βM∞ : Db (kM ) → DbIR−c (IkM∞ ),

(3) (·) ⊗ (·) : DbIR−c (IkM∞ ) × DbIR−c (IkM∞ ) → DbIR−c (IkM∞ ),

(4) RIhom(ιM∞ (·), ·) : DbR−c (kM∞ )op × DbIR−c (IkM∞ ) → DbIR−c (IkM∞ ),

31
(5) f −1 : DbIR−c (IkN∞ ) → DbIR−c (IkM∞ ),
(6) Rf!! : DbIR−c (IkM∞ ) → DbIR−c (IkN∞ ),
(7) f ! : DbIR−c (IkN∞ ) → DbIR−c (IkM∞ ).
Proof. (1) Let F ∈ DbR−c (kM ). Then we have jM ! F ∈ DbR−c (kM̌ ). Note that there exists
an isomorphism RjM∞ !! ιM∞ F ≃ ιM̌ jM ! F in Db (IkM̌ ). Since the functor ιM̌ : DbR−c (kM̌ ) →
DbIR−c (IkM̌ ) is well defined, we have RjM∞ !! ιM∞ F ∈ DbIR−c (IkM̌ ). This implies that
ιM∞ F ∈ DbIR−c (IkM∞ ).
(2) Let F ∈ Db (kM ). Then we have jM ! F ∈ Db (kM̌ ). Note that there exists an
isomorphism RjM∞ !! βM∞ F ≃ βM̌ jM ! F in Db (IkM̌ ). By [KS01, Lem. 7.1.3 (v)], we
have βM̌ jM ! F ∈ DbIR−c (kM̌ ), and hence RjM∞ !! βM∞ F ∈ DbIR−c (kM̌ ). This implies that
βM∞ F ∈ DbIR−c (IkM∞ ).
(3) Let F1 , F2 ∈ DbIR−c (IkM∞ ). Then we have RjM∞ !! F1 , RjM∞ !! F2 ∈ DbIR−c (IkM̌ ). Note
that there exists an isomorphism RjM∞ !! (F1 ⊗ F2 ) ≃ RjM∞ !! F1 ⊗ RjM∞ !! F2 . By [KS01,
Lem. 7.1.3 (iii)], we have RjM∞ !! F1 ⊗RjM∞ !! F2 ∈ DbIR−c (kM̌ ), and hence RjM∞ !! (F1 ⊗F2 ) ∈
DbIR−c (kM̌ ). This implies that F1 ⊗ F2 ∈ DbIR−c (IkM∞ ).
(4) Let F ∈ DbR−c (kM∞ ) and G ∈ DbIR−c (IkM∞ ). Then we have jM ! F ∈ DbR−c (kM̌ ) and
RjM∞ !! G ∈ DbIR−c (IkM̌ ). Note that there exist isomorphisms in Db (IkM̌ )
−1 −1
RjM∞ !! RIhom(ιM∞ F , G) ≃ RjM∞ !! jM ∞
RjM∞ ∗ RIhom(ιM∞ F , jM ∞
RjM∞ !! G)
−1
≃ RjM∞ !! jM∞ RIhom(RjM∞ !! ιM∞ F , RjM∞ !! G)
≃ ιM̌ kM ⊗ RIhom(ιM̌ jM ! F , RjM∞ !! G),
−1 −1
where the first isomorphism follows from jM ∞
RjM∞ ∗ ≃ id and id ≃ jM ∞
RjM∞ !! , in the
−1
second isomorphism we used the adjointness of (RjM∞ !! , jM∞ ) and the last isomorphism
−1
follows from RjM∞ !! jM ∞
(·) ≃ ιM̌ kM ⊗ (·). By [KS01, Lem. 7.1.3 (iii), (iv)], we have
ιM̌ kM ⊗ RIhom(ιM̌ jM ! F , RjM∞ !! G) ∈ DbIR−c (kM̌ ), and hence RjM∞ !! RIhom(ιM∞ F , G) ∈
DbIR−c (kM̌ ). This implies that RIhom(ιM∞ F , G) ∈ DbIR−c (IkM∞ ).
(5) Let G ∈ DbIR−c (IkN∞ ). Then we have RjN∞ !! G ∈ DbIR−c (IkŇ ). Note that there exists
an isomorphism RjM∞ !! f −1 G ≃ fˇ−1 RjN∞ !! G in Db (IkŇ ). Since the functor fˇ: Db (IkŇ ) →
Db (IkM̌ ) is well defined, we have RjM∞ !! f −1 G ∈ DbIR−c (IkM̌ ). This implies that f −1 G ∈
DbIR−c (IkM∞ ).
(6) Let F ∈ DbIR−c (IkM∞ ). Then we have RjM∞ !! F ∈ DbIR−c (IkM̌ ). Note that there exists
an isomorphism RjN∞ !! Rf!! F ≃ fˇ!! RjM∞ !! F in Db (IkM̌ ). By [KS01, Lem. 7.1.3 (i), (iv)],
we have RjN∞ !! Rf!! F ∈ DbIR−c (IkŇ ). This implies that Rf!! F ∈ DbIR−c (IkN∞ ).
(7) Let G ∈ DbIR−c (IkN∞ ). Then we have RjN∞ !! G ∈ DbIR−c (IkŇ ). Note that there exist
isomorphism in Db (IkM̌ )
RjM∞ !! f ! G ≃ RjM∞ !! j −1 fˇ! RjN∞ !! G,
M∞
≃ ιM̌ kM ⊗ fˇ! RjN∞ !! G
≃ fˇ−1 ι kN ⊗ fˇ! RjN∞ !! G

≃ fˇ! (βŇ kN ⊗ RjN∞ !! G),

32
−1
where in the second isomorphism we used RjM∞ !! jM ∞
≃ ιM̌ kM ⊗ (·) and the last iso-
morphism follows from [KS01, Thm. 5.6.3] and the fact that ιM̌ ≃ βM̌ . By [KS01,
Lem. 7.1.3 (ii), (iii), (v)], we have fˇ! (βŇ kN ⊗ RjN∞ !! G) ∈ DbIR−c (IkM̌ ), and hence
RjM∞ !! f ! G ∈ DbIR−c (IkM̌ ). This implies that f ! G ∈ DbIR−c (IkM∞ ).
Let us describe a relation between ind-sheaves on M∞ and subanalytic sheaves on M∞ .
Proposition 3.6. Let M∞ = (M, M̌ ) be a real analytic bordered space. Then we have
(1) a pair (IM∞ , RJM∞ ) is an adjoint pair and there exists a canonical isomorphism

id −→ RJM∞ ◦ IM∞ ,

(2) there exists an equivalence of triangulated categories:


I M∞
/
Db (ksub
M∞ ) o
∼ DbIR−c (IkM∞ ).
RJM∞

Proof. (1) Let F ∈ Db (ksub b


M∞ ) and G ∈ D (IkM∞ ). Then we have

HomDb (IkM∞ ) (IM∞ F , G) = HomDb (IkM∞ ) (qIM̌ RjM∞ !! F , G)


≃ HomDb (IkM̌ ) (IM̌ RjM∞ !! F , RjM∞ ∗ G)
≃ HomDb (ksub ) (RjM∞ !! F , RJM̌ RjM∞ ∗ G)

−1
≃ HomDb (ksub
M )
(F , RjM ∞
RJM̌ RjM∞ ∗ G)

≃ HomDb (ksub
M )
(F , RJM∞ G),

−1 −1
where in the first isomorphism we used the fact that q = jM ∞
and a pair (jM ∞
, RjM∞ ∗ )
is an adjoint pair, the second isomorphism follows from a pair (IM̌ , RJM̌ ) is an adjoint
−1
pair and the third isomorphism follows from a pair (RjM∞ !! , jM ∞
) is an adjoint pair. This
implies that a pair (IM∞ , RJM∞ ) is an adjoint pair.
Hence, there exists a natural morphism id → RJM∞ ◦ IM∞ of functors. Moreover, for
any F ∈ Db (IkM∞ ), we have isomorphisms in Db (IkM∞ )
−1
(RJM∞ ◦ IM∞ )(F ) ≃ jM ∞
RJM̌ RjM∞ ∗ qIM̌ RjM∞ !! F
−1 −1
≃ jM∞ RJM̌ RjM∞ ∗ jM I RjM∞ !! F
∞ M̌
−1
≃ jM∞ RJM̌ RIhom(ιM̌ kM , IM̌ RjM∞ !! F )
−1
≃ jM ∞
RJM̌ RIhom(IM̌ ρM̌ kM , IM̌ RjM∞ !! F )
−1
≃ jM ∞
RIhomsub (ρM̌ kM , RJM̌ IM̌ RjM∞ !! F )
−1
≃ jM ∞
RIhomsub (ρM̌ kM , RjM∞ !! F )
−1 −1
≃ jM ∞
RjM∞ ∗ jM ∞
RjM∞ !! F
≃F
−1
where the second isomorphism follows from q = jM ∞
, in the third isomorphism we used
−1
the fact that RjM∞ ∗ jM∞ (·) ≃ RIhom(ιM̌ kM , ·), the fourth isomorphism follows from
IM̌ ◦ρR−c

= ιM̌ |ModR−c (kM̌ ) , the fifth isomorphism follows from the adjointness of (IM̌ , RJM̌ )

and in the sixth we used the fact id −→ RJM̌ ◦ IM̌ .

33
(2) First, let us prove that for any F ∈ Db (ksub b
M∞ ) one has IM∞ (F ) ∈ DIR−c (IkM∞ ). There
exist isomorphisms in Db (IkM̌ )

RjM∞ !! IM∞ (F ) ≃ RjM∞ !! qIM̌ RjM∞ !! F


≃ ιM̌ kM ⊗ IM̌ RjM∞ !! F
≃ IM̌ ρM̌ kM ⊗ IM̌ RjM∞ !! F
≃ IM̌ (ρM̌ kM ⊗ RjM∞ !! F )

where the third isomorphism follows from IM̌ ◦ ρR−c M̌


= ιM̌ |ModR−c (kM̌ ) and in the fourth
isomorphism we used the fact that IM̌ (· ⊗ ·) ≃ IM̌ (·) ⊗ IM̌ (·). Since IM̌ (ρM̌ kM ⊗
RjM∞ !! F ) ∈ DbIR−c (IkM̌ ), we have RjM∞ !! IM∞ (F ) ∈ DbIR−c (IkM̌ ). This implies IM∞ (F ) ∈
DbIR−c (IkM∞ ).
By (1), the functor IM∞ is fully faithful. Let us prove that the functor IM∞ is essentially
surjective. Let G ∈ DbIR−c (IkM∞ ). Then we have RjM∞ !! G ∈ DbIR−c (IkM̌ ). By Theorem
2.26, there exists F ∈ Db (ksub M̌
) such that RjM∞ !! G ≃ IM̌ F and hence we have G ≃
−1
jM∞ IM̌ F . Moreover, there exist isomorphisms in Db (IkM∞ )
−1 −1 −1
jM I RjM∞ !! jM
∞ M̌ ∞
F ≃ jM I (ρM̌ kM ⊗ F )
∞ M̌
−1
≃ jM∞ (IM̌ ρM̌ kM ⊗ IM̌ F )
−1
≃ jM ∞
(ιM̌ kM ⊗ IM̌ F )
−1 −1
≃ jM∞ RjM∞ !! jM I F
∞ M̌
−1
≃ jM∞ IM̌ F

where in the second isomorphism we used the fact that IM̌ (· ⊗ ·) ≃ IM̌ (·) ⊗ IM̌ (·), the
third isomorphism follows from IM̌ ◦ ρR−c

= ιM̌ |ModR−c (kM̌ ) and in the fourth isomorphism
−1
we used the fact that (RjM∞ !! ◦ jM∞ )(·) ≃ ιM̌ kM ⊗ (·). Hence we have
−1 −1 −1 −1
G ≃ jM I F ≃ jM
∞ M̌
I RjM∞ !! jM
∞ M̌ ∞
F ≃ IM∞ (jM ∞
F ).

This implies that the functor IM∞ is essentially surjective.


Therefore, the proof is completed.

We will denote by λM∞ : DbIR−c (IkM∞ ) −→ Db (ksub
M ) the inverse functor of
b sub ∼ b
IM∞ : D (kM∞ ) −→ DIR−c (IkM∞ ).
Proposition 3.7. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces
associated with a morphism fˇ: M̌ → Ň of real analytic manifolds.
(1) For any F ∈ Db (kM∞ ) and any G ∈ Db (IkM∞ ), we have

RJM∞ RIhom(IM∞ F , G) ≃ RIhomsub (F , RJM∞ G).

(2) For any L ∈ Db (kM ), any F , F1, F2 ∈ Db (ksub b sub


M∞ ) and any G ∈ D (kN∞ ), we have

(i) αM∞ IM∞ F ≃ ρ−1


M∞ F ,
(ii) IM∞ ρM∞ ! L ≃ βM∞ L,
(iii) IM∞ f −1 G ≃ f −1 IN∞ G,

34
(iv) Rf!! IM∞ F ≃ IN∞ Rf!! F ,
(v) IM∞ f ! G ≃ f ! IN∞ G,
(vi) IM∞ (F1 ⊗ F2 ) ≃ IM∞ (F1 ) ⊗ IM∞ (F2 ).

(3) For any L ∈ Db (kM ), any F ∈ Db (IkM∞ ) and any G ∈ Db (IkN∞ ), we have

(i) RJM∞ ιM∞ F ≃ ρM∞ ∗ F ,


(ii) ρ−1
M∞ RJM∞ F ≃ αM∞ F ,

(iii) RJM∞ βM∞ L ≃ ρM∞ ! L,


(iv) Rf∗ RJM∞ F ≃ RJN∞ Rf∗ F ,
(v) RJM∞ f ! G ≃ f ! RJN∞ G.

(4) For any F, F1 , F2 ∈ DbIR−c (IkM∞ ), any G ∈ DbIR−c (IkN∞ ) and any L ∈ DbR−c (kM∞ ),
we have
R−c
(i) IM∞ ρM ∞∗
L ≃ ιM∞ L,
(ii) λM∞ f −1 G ≃ f −1 λN∞ G,
(iii) Rf!! λM∞ F ≃ λN∞ Rf!! F ,
(iv) λM∞ (F1 ⊗ F2 ) ≃ λM∞ (F1 ) ⊗ λM∞ (F2 ).

Proof. (1) Let F ∈ Db (kM∞ ) and G ∈ Db (IkM∞ ). Then we have


−1 −1
RJM∞ RIhom(IM∞ F , G) = jM ∞
RJM̌ RjM∞ ∗ RIhom(jM I RjM∞ !! F , G)
∞ M̌
−1
≃ jM ∞
RIhomsub (RjM∞ !! F , RJM̌ RjM∞ ∗ G)
≃ RIhomsub (jM
−1

−1
RjM∞ !! F , jM ∞
RJM̌ RjM∞ ∗ G)
≃ RIhomsub (F , RJM∞ G),
−1
where in the second isomorphism we used the adjointness of (jM ∞
, RjM∞ ∗ ) and (IM̌ , RjM̌ )
−1
and the last isomorphism follows from jM ∞
◦ RjM∞ !! ≃ id.

(2)(i) Let F ∈ Db (ksub


M∞ ). Then we have

αM∞ IM∞ F ≃ i−1 −1


M αM̌ RjM∞ !! jM∞ IM̌ RjM∞ !! F
≃ i−1
M αM̌ (ιM̌ kM ⊗ IM̌ RjM∞ !! F )
−1
≃ iM αM̌ ιM̌ kM ⊗ i−1
M αM̌ IM̌ RjM∞ !! F
−1 −1
≃ iM ρM̌ RjM∞ !! F
≃ ρ−1 −1
M∞ jM∞ RjM∞ !! F
≃ ρ−1
M∞ F ,

−1
where in the second isomorphism we used the fact that RjM∞ !! jM ∞
≃ ιM̌ kM ⊗ (·) and the
−1
fourth isomorphism follows from αM̌ ◦ IM̌ ≃ ρM̌ .

35
(ii) Let L ∈ Db (kM ). Since IM̌ ◦ ρM̌ ! ≃ βM̌ , we have
−1
IM∞ ρM∞ ! L ≃ jM I RjM∞ !! ρM∞ ! L
∞ M̌
−1
≃ jM I ρ j L
∞ M̌ M̌ ! M !
−1
≃ jM∞ βM̌ jM ! L
≃ βM∞ F .

(iii) Let G ∈ Db (ksub ˇ−1 ≃ fˇ−1 ◦ I , we have


N∞ ). Since IM̌ ◦ f Ň

IM∞ f −1 G ≃ jM−1
I RjM∞ !! f −1 G
∞ M̌

≃ j −1 I fˇ−1 RjN∞ !! G
M∞ M̌
−1 ˇ−1
≃ jM ∞
f IŇ RjN∞ !! G
−1 −1
≃ f jN∞ IŇ RjN∞ !! G
≃ f −1 IN∞ G.

(iv) Let F ∈ Db (ksub ˇ ˇ


M∞ ). Since IŇ ◦ f!! ≃ f!! ◦ IM̌ , we have
−1
Rf!! IM∞ F ≃ Rf!! jM I RjM∞ !! F
∞ M̌

≃ j −1 Rfˇ!! I RjM∞ !! F
N∞ M̌

≃ −1 ˇ
jN∞ IŇ Rf!! RjM∞ !! F
−1
≃ jN I RjN∞ !! Rf!! F
∞ Ň

≃ IN∞ Rf!! F .

(v) Let G ∈ Db (ksub ! b


N∞ ). By Lemma 3.5 (7) we have f IN∞ G ∈ DIR−c (IkM∞ ) and hence by
Proposition 3.6 (2) we have IM∞ RJM∞ f ! IN∞ G ≃ f ! IN∞ G. Then we have
f ! IN∞ G ≃ IM∞ RJM∞ f ! IN∞ G ≃ IM∞ f ! RJN∞ IN∞ G ≃ IM∞ f ! G,
where the second isomorphism follows from (3)(v) and in the last isomorphism we used
RJM∞ ◦ IM∞ ≃ id.
(vi) Let F1 , F2 ∈ Db (ksub
M∞ ). Then we have
−1
RjM∞ !! (F1 ⊗ F2 ) ≃ RjM∞ !! (jM ∞
RjM∞ !! F1 ⊗ F2 ) ≃ RjM∞ !! F1 ⊗ RjM∞ !! F2 ,
and hence we have
IM̌ RjM∞ !! (F1 ⊗ F2 ) ≃ IM̌ RjM∞ !! F1 ⊗ IM̌ RjM∞ !! F2 .
So that, we have
−1
IM∞ (F1 ⊗ F2 ) ≃ jM I RjM∞ !! (F1 ⊗ F2 )
∞ M̌
−1 −1
≃ jM∞ IM̌ RjM∞ !! F1 ⊗ jM I RjM∞ !! F2
∞ M̌

≃ IM ∞ F 1 ⊗ IM ∞ F 2 .

36
(3)(i) Let F ∈ Db (IkM∞ ). Since RJM̌ ◦ ιM̌ ≃ ρM̌ ∗ , we have
−1
RJM∞ ιM∞ F ≃ jM ∞
RJM̌ RjM∞ ∗ ιM∞ F
−1
≃ jM∞ RJM̌ ιM̌ RjM∞ ∗ F
−1
≃ jM ρ RjM∞ ∗ F
∞ M̌ ∗
−1
≃ ρM∞ ∗ jM ∞
RjM∞ ∗ F
≃ ρM∞ ∗ F.

(ii) Let F ∈ Db (IkM∞ ). Since ρ−1



◦ RJM̌ ≃ αM̌ , we have

ρ−1 −1 −1
M∞ RJM∞ F ≃ ρM∞ jM∞ RJM̌ RjM∞ ∗ F
−1 −1
≃ jM ρ RJM̌ RjM∞ ∗ F
∞ M̌
−1
≃ jM α RjM∞ ∗ F
∞ M̌
−1
≃ αM∞ jM ∞
RjM∞ ∗ F
≃ αM∞ F.

(iii) Let L ∈ Db (kM ). By using (2)(ii) and the fact that RJM∞ ◦ IM∞ ≃ id, we have

RJM∞ βM∞ L ≃ RJM∞ IM∞ ρM∞ ! L ≃ ρM∞ ! F.

(iv) Let F ∈ Db (IkM∞ ) and G ∈ Db (ksub


N∞ ). Then we have

HomDb (IkN∞ ) (G, Rf∗ RJM∞ F ) ≃ HomDb (IkM∞ ) (IM∞ f −1 G, F )


≃ HomDb (IkM∞ ) (f −1 IN∞ G, F )
≃ HomDb (IkN∞ ) (G, RJN∞ Rf∗ F ),

where the first and last isomorphisms follow from adjointness of (f −1 , Rf∗ ) and (I, RJ)
and in the second isomorphism we used (2)(iii). Hence there exists an isomorphism
Rf∗ RJM∞ F ≃ RJN∞ Rf∗ F in Db (IkN∞ ).
(v) Let G ∈ Db (IkN∞ ). Since RJ ◦ fˇ! ≃ fˇ! ◦ RJ , we have
M̌ Ň

RJM∞ f ! G ≃ jM−1

RJM̌ RjM∞ ∗ f ! G
≃ j −1 RJ fˇ! RjN∞ ∗ G
M∞ M̌
−1 ˇ!
≃ jM∞ f RJŇ RjN∞ ∗ G
≃ f ! jN
−1

RJŇ RjN∞ ∗ G
!
≃ f RJN∞ G.

37
(4)(i) Let L ∈ DbR−c (kM∞ ). Then we have ιM∞ L ∈ DbIR−c (IkM∞ ) by Lemma 3.5 (1).
Hence, by using (3)(i) and Proposition 3.6 (2), we have

ιM∞ L ≃ IM∞ RJM∞ ιM∞ L ≃ IM∞ ρM∞ L.

(ii) Let G ∈ DbIR−c (IkN∞ ). Then we have G ≃ IN∞ λN∞ G by Proposition 3.6 (2). Moreover,
by using (2)(iii) we have

λM∞ f −1 G ≃ λM∞ f −1 IN∞ λN∞ G ≃ λM∞ IM∞ f −1 λN∞ G ≃ f −1 λN∞ G.

(iii) Let F ∈ DbIR−c (IkM∞ ). Then we have F ≃ IM∞ λM∞ F by Proposition 3.6 (2). Hence,
by using (2)(iv) we have

λN∞ f!! F ≃ λN∞ f!! IM∞ λM∞ F ≃ λN∞ IN∞ f!! λM∞ F ≃ f!! λM∞ F.

(iv) Let F1 , F2 ∈ DbIR−c (IkM∞ ). Then we have Fi ≃ IM∞ λM∞ Fi (i = 1, 2) by Proposition


3.6 (2). Hence, by using (2)(vi) we have

λM∞ (F1 ⊗F2 ) ≃ λM∞ (IM∞ λM∞ F1 ⊗ IM∞ λM∞ F2 ) ≃ λM∞ IM∞ (λM∞ F1 ⊗ λM∞ F2 ) ≃ λM∞ F1 ⊗λM∞ F2 .

3.2 Convolutions for Subanalytic Sheaves on Real Analytic Bor-


dered Spaces
In this subsection, let us define convolution functors for subanalytic sheaved on real
analytic bordered spaces.
Let M∞ = (M, M̌ ) be a real analytic bordered space. We set R∞ := (R, R) for
R := R ⊔ {−∞, +∞}, and let t ∈ R be the affine coordinate. We consider the morphisms
of real analytic bordered spaces
p1 , p2 , µ π
M∞ × R2∞ −−−−−→ M∞ × R∞ −→ M∞

given by the maps p1 (x, t1 , t2 ) := (x, t1 ), p2 (x, t1 , t2 ) := (x, t2 ), µ(x, t1 , t2 ) := (x, t1 + t2 )


and π(x, t) := x.
Then, the convolution functors for subanalytic sheaves on M∞ × R∞
+
(·) ⊗ (·) : Db (ksub b sub b sub
M∞ ×R∞ ) × D (kM∞ ×R∞ ) → D (kM∞ ×R∞ ),
RIhom+,sub (·, ·) : Db (ksub
M∞ ×R∞ )
op
× Db (ksub b sub
M∞ ×R∞ ) → D (kM∞ ×R∞ )

are defined by
+
F1 ⊗ F2 := Rµ!! (p−1 −1
1 F1 ⊗ p2 F2 ),
RIhom+,sub (F1 , F2 ) := Rp1∗ RIhomsub (p−1 !
2 F1 , µ F2 ),

38
for F1 , F2 ∈ Db (ksub b sub
M∞ ×R∞ ). Note that for any F , F1 , F2 , F3 ∈ D (kM∞ ×R∞ ) there exist
isomorphisms
+ +
F1 ⊗ F2 ≃ F2 ⊗ F1 ,
   
+ + + +
F1 ⊗ F2 ⊗ F3 ≃ F1 ⊗ F2 ⊗ F3 ,
+
k{t=0} ⊗ F ≃ F ≃ RIhom+,sub (k{t=0} , F ),

where {t = 0} stands for {(x, t) ∈ M × R | t = 0}. Hence, the category Db (ksub


M∞ ×R∞ ) has
+
a structure of commutative tensor category with ⊗ as tensor product functor and k{t=0}
as unit object.
The convolution functors have several properties as similar to the tensor product
functor and the internal hom functor. For a morphism of real analytic bordered spaces
f : M∞ → N∞ , let us denoted by fR∞ : M∞ × R∞ → M∞ × R∞ the morphism f × idR∞
of real analytic bordered spaces.

Proposition 3.8. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces,


F , F1, F2 , ∈ Db (ksub b sub
M∞ ×R∞ ) and G, G1 , G2 , ∈ D (kN∞ ×R∞ ). There exist isomorphism
 
+,sub
+ 
RIhom F1 ⊗ F2 , F ≃ RIhom+,sub F1 , RIhom+,sub (F2 , F ) ,
 
+
+,sub

HomDb (ksub
M∞ ×R∞ ) F 1 ⊗ F 2 , F ≃ Hom Db (ksub
M∞ ×R∞ ) F 1 , RIhom (F 2 , F ) ,

RfR∞ ∗ RIhom+,sub fR−1 G, F ≃ RIhom+,sub (G, RfR∞ ∗ F ) ,
 ∞

+ +
fR∞ F1 ⊗ F2 ≃ fR−1
−1

F1 ⊗ fR−1 ∞
F2 ,

RIhom+,sub (RfR∞ !! F , G) ≃ RfR∞ ∗ RIhom+,sub F , fR! ∞ G ,
 
+ +
−1
RfR∞ !! F ⊗ fR∞ G ≃ RfR∞ !! F ⊗ G,

fR! ∞ RIhom+,sub (G1 , G2 ) ≃ RIhom+,sub fR−1 G , fR! ∞ G2 .
∞ 1

Proof. First, let us prove the second isomorphism. By using the adjointness, we have
 
+ 
HomDb (ksub
M ×R )
F1 ⊗ F2 , F3 ≃ HomDb (ksub
M ×R )
Rµ!! (p−1 −1
1 F1 ⊗ p2 F2 ), F3
∞ ∞ ∞ ∞

≃ HomDb (ksub
M ) F1 , Rp1∗ RIhomsub (p−1 !
2 F2 , µ F3 )
∞ ×R∞

≃ RIhomsub F1 , RIhom+,sub (F2 , F3 ) .

Let us prove the first isomorphism. By using the second isomorphism, for any F0 ∈

39
Db (kM∞ ×R∞ ), there exist isomorphisms
  
+
+,sub
HomDb (ksub
M∞ ×R∞ )
F0 , RIhom F1 ⊗ F2 , F3
   
+ +
≃ HomDb (ksub
M∞ ×R∞ )
F0 ⊗ F1 ⊗ F2 , F3
  
+ +
≃ HomDb (ksub
M∞ ×R∞ )
F0 ⊗ F1 ⊗ F2 , F3
 
+
+,sub
≃ HomDb (ksub
M ×R )
F0 ⊗ F1 , RIhom (F2 , F3 )
∞ ∞

≃ HomDb (ksub
M ) F0 , RIhom+,sub F1 , RIhom+,sub (F2, F3 ) .
∞ ×R∞

 
+ 
Hence, we have RIhom+,sub F1 ⊗ F2 , F3 ≃ RIhom+,sub F1 , RIhom+,sub (F2 , F3) .

Let us denote by feR∞ : M∞ × R2∞ → N∞ × R2∞ the morphism f × idR2∞ of real analytic
bordered spaces. Then there exist commutative diagrams:
⋆ π
Db (ksub
M∞ ×R2∞ )
/ Db (ksub
M∞ ×R∞ )
/ Db (ksub
M∞ )

feR∞ f R∞ f
  
b
D (ksub
N∞ ×R2∞ ) ⋆
/ D b
(ksub
N∞ ×R∞ ) π
/ D (ksub
b
N∞ ),

where ⋆ = p1 , p2 , µ, respectively. Hence, we have

RfR∞ ∗ RIhom+,sub (fR−1



G, F ) ≃ RfR∞ ∗ Rp1∗ RIhomsub (p−1 −1 !
1 fR∞ G, µ F )

≃ Rp1∗ RfeR∞ ∗ RIhomsub (fe−1 p−1 G, µ! F ) R∞ 1

≃ Rp1∗ RIhom (p1 G, RfeR∞ ∗ µ! F )


sub
−1

≃ Rp1∗ RIhomsub (p−1 !


1 G, µ RfR∞ ∗ F )
≃ RIhom+,sub (G, RfR∞ ∗ F ),

where in the fourth isomorphism we used Proposition 3.4 (4). By using Proposition 3.4
(3), (4) we have
   
+ +
−1 −1 −1 −1
fR∞ F1 ⊗ F2 ≃ fR∞ Rµ!! p1 F1 ⊗ p2 F2
 
+
e−1 −1 −1
≃ Rµ!! fR∞ p1 F1 ⊗ p2 F2
 
+
e−1 −1 e−1 −1
≃ Rµ!! fR∞ p1 F1 ⊗ fR∞ p2 F2
 
+
−1 −1 −1 −1
≃ Rµ!! p1 fR∞ F1 ⊗ p2 fR∞ F2
+
≃ fR−1

F1 ⊗ fR−1

F2 ,

40
and
   
+ +
−1
RfR∞ !! F ⊗ fR∞ G ≃ RfR∞ !! Rµ!! p−11 F ⊗ p−1 −1
2 fR∞ G
 
+
e
≃ Rµ!! RfR∞ !! p−1 ⊗ feR−1 p−1 G
1 F ∞ 2
 
+
≃ Rµ!! RfeR∞ !! p−1
1 F
−1
⊗ p2 G
 
+
≃ Rµ!! p−11 RfR∞ !! F
−1
⊗ p2 G
+
≃ RfR∞ !! F ⊗ G.

Moreover, we have

fR! ∞ RIhom+,sub (G1 , G2 ) ≃ fR! ∞ Rp1∗ RIhomsub (p−1 −1


1 G1 , p2 G2 )

≃ Rp1∗ fe! RIhomsub (p−1 G1 , p−1 G2 )


R∞ 1 2

≃ Rp1∗ RIhom (feR−1


sub
p−1 G1 , feR! ∞ p−1
∞ 1 2 G2 )
sub −1 −1 −1 !
≃ Rp1∗ RIhom (p1 fR∞ G1 , p2 fR∞ G2 )

≃ RIhom+,sub fR−1 G , fR! ∞ G2 .
∞ 1

By using Proposition 3.4 (2), (4) there exits isomorphisms



RIhom+,sub (RfR∞ !! F , G) ≃ Rp1∗ RIhomsub p−1 !
2 RfR∞ !! F , µ G
 
≃ Rp1∗ RIhom sub e −1
RfR∞ !! p2 F , µ G!

 
≃ Rp1∗ RfeR∞ ∗ RIhomsub p−1 2 F , R e! µ! G
fR∞
sub

≃ RfR∞ ∗ Rp1∗ RIhom p2 F , µ RfR! ∞ G
−1 !

≃ RfR∞ ∗ RIhom+,sub F , fR! ∞ G .

Proposition 3.9. Let F , G, H ∈ Db (ksub b sub


M∞ ×R∞ ) and F0 ∈ D (kM∞ ). Then there exist
isomorphisms
 
+ +
π F0 ⊗ F ⊗ G ≃ π −1 F0 ⊗ F ⊗ G,
−1

 
RIhomsub π −1 F0 , RIhom+,sub (F , G) ≃ RIhom+,sub π −1 F0 ⊗ F , G

≃ RIhom+,sub F , RIhomsub (π −1 F0 , G ,
 
sub
+ 
Rπ∗ RIhom F ⊗ G, H ≃ Rπ∗ RIhomsub F , RIhom+,sub (G, H) .

Proof. First let us remark that π ◦ p1 = π ◦ p2 = π ◦ µ.

41
By using Proposition 3.4 (3), we have
 
+ 
π F0 ⊗ F ⊗ G ≃ π −1 F0 ⊗ Rµ!! p−1
−1
1 F ⊗ p −1
2 G

≃ Rµ!! µ−1 π −1 F0 ⊗ p−1 1 F ⊗ p2 G
−1

≃ Rµ!! p−1
1 π −1
F 0 ⊗ p −1
1 F ⊗ p −1
2 G

≃ Rµ!! p−1
1 (π −1
F 0 ⊗ F ) ⊗ p −1
2 G
+
≃ π −1 F0 ⊗ F ⊗ G.

By using Proposition 3.4 (2),


 
RIhomsub π −1 F0 , RIhom+,sub (F1 , F2) ≃ RIhomsub π −1 F0 , Rp1∗ RIhomsub (p−1 !
2 F1 , µ F2 )

≃ Rp1∗ RIhomsub p−1 −1
1 π F0 , RIhom
sub −1
(p2 F1 , µ! F2 )

≃ Rp1∗ RIhomsub p−1 −1 −1 !
1 π F 0 ⊗ p2 F 1 , µ F 2 ) .

Moreover, by using Proposition 3.4 (3) we have


 
Rp1∗ RIhomsub p−1 −1 −1 !
1 π F0 ⊗ p2 F1 , µ F2 ) ≃ Rp1∗ RIhom
sub
p−1 −1 −1
2 π F 0 ⊗ p2 F 1 , µ F 2 )
!

≃ Rp1∗ RIhomsub p−1 2 (π −1
F 0 ⊗ F 1 ), µ !
F 2 )
+,sub −1

≃ RIhom π F0 ⊗ F1 , F2

and by using Proposition 3.4 (2), (3) we have



Rp1∗ RIhomsub p−1 −1 −1 !
1 π F 0 ⊗ p2 F 1 , µ F 2 )

≃ Rp1∗ RIhomsub µ−1 π −1 F0 ⊗ p−1 !
2 F1 , µ F2 )

≃ Rp1∗ RIhomsub p−1 2 F1 , RIhom
sub
(µ−1π −1 F0 , µ! F2 )

≃ Rp1∗ RIhomsub p−1 2 F 1 , µ !
RIhom sub
(π −1
F 0 , F 2 )
+,sub

≃ RIhom F1 , RIhomsub (π −1 F0 , F2) .

Let us prove the last assertion. For any G0 ∈ Db (ksub M∞ ), there exist isomorphisms
  
+
sub
HomDb (ksub
M∞ )
G0 , Rπ∗ RIhom F ⊗ G, H
  
+
−1 sub
≃ HomDb (ksub
M∞ ×R∞ )
π G0 , RIhom F ⊗ G, H
   
+
−1
≃ HomDb (ksub
M∞ ×R∞ )
π G0 ⊗ F ⊗ G , H
 
−1
+
≃ HomDb (ksub
M∞ ×R∞ )
π G0 ⊗ F ⊗ G, H
−1 +,sub

≃ HomDb (ksub
M∞ ×R∞ ) π G0 ⊗ F , RIhom (G, H)

≃ HomDb (ksub
M )
G0 , Rπ∗ RIhomsub F , RIhom+,sub (G, H) .

42
where in the third (resp. fourth) isomorphism we used the first assertion (resp. Proposition
3.8). Hence, we have an isomorphism
 
sub
+ 
Rπ∗ RIhom F ⊗ G, H ≃ Rπ∗ RIhomsub F , RIhom+,sub (G, H) .

Lemma 3.10. For any F ∈ Db (ksub b sub


M∞ ×R∞ ) and any G ∈ D (kN∞ ), We have

+
π −1 G ⊗ F ≃ π −1 (G ⊗ Rπ!! F ),
RIhom+,sub (π −1 G, F ) ≃ π ! RIhom+,sub (G, Rπ∗ F ),
RIhom+,sub (F , π ! G) ≃ π ! RIhom+,sub (Rπ!! F , G).

Proof. Note that there exist the following Cartesian diagrams (i=1, 2)
µ p1
M∞ × R2∞ / M∞ × R∞ o M∞ × R2∞
pi ✷ π ✷ p2
  
M∞ × R∞ π
/ M∞ o π M∞ × R∞ .

Then by using Proposition 3.4 (3), (4), we have


+ 
π −1 G ⊗ F ≃ Rµ!! p−1 −1 −1
1 π G ⊗ p2 F

≃ Rµ!! µ−1 π −1 G ⊗ p−1
2 F
≃ π −1 G ⊗ Rµ!! p−1
2 F
−1 −1
≃ π G ⊗ π Rπ!! F
≃ π −1 (F ⊗ Rπ!! F ).

Moreover, by using Proposition 3.4 (2), (3), (4), we have

RIhom+,sub (π −1 G, F ) ≃ Rp1∗ RIhomsub (p−1 −1 !


2 π G, µ F )
≃ Rp1∗ RIhomsub (µ−1 π −1 G, µ! F )
≃ Rp1∗ µ! RIhomsub (π −1 G, F )
≃ π ! Rπ∗ RIhomsub (π −1 G, F )
≃ π ! RIhom+,sub (G, Rπ∗ F )

and

RIhom+,sub (F , π ! G) ≃ Rp1∗ RIhomsub (p−1 ! !


2 F , µ π G)
≃ Rp1∗ RIhomsub (p−1 ! !
2 F , p2 π G)
≃ Rp1∗ p!2 RIhomsub (F , π !G)
≃ π ! Rπ∗ RIhomsub (F , π ! G)
≃ π ! RIhom+,sub (Rπ!! F , G).

43
Since Rπ!! k{t≤0} ≃ Rπ!! k{t≥0} ≃ 0 and π −1 kM ≃ kM ×R , we have
Corollary 3.11. For any F ∈ Db (ksub b sub
M∞ ×R∞ ) and any G ∈ D (kM∞ ), we have

+ 
k{t≥0} ⊕ k{t≤0} ⊗ π −1 G ≃ RIhom+,sub k{t≥0} ⊕ k{t≤0} , π −1 G ≃ 0,
+
kM ×R ⊗ F ≃ π −1 Rπ!! F ( ≃ π ! Rπ!! F [−1]),
RIhom+,sub (kM ×R , F ) ≃ π ! Rπ∗ F ( ≃ π −1 Rπ∗ F [1]).

At the end of this subsection, we shall prove the following proposition.


Proposition 3.12. Let M∞ = (M, M̌ ) be a real analytic bordered space.
(1) For any F1 , F2 ∈ Db (ksub b
M∞ ×R∞ ), there exits an isomorphism in D (IkM∞ ×R∞ ) :
 
+ +
IM∞ ×R∞ F1 ⊗ IM∞ ×R∞ F2 ≃ IM∞ ×R∞ F1 ⊗ F2 .

(2) For any F ∈ Db (ksub b


M∞ ×R∞ ) and any G ∈ D (IkM∞ ×R∞ ), there exist an isomorphism
in Db (ksub
M∞ ×R∞ ) :

RJM∞ ×R∞ RIhom+ (IM∞ ×R∞ F , G) ≃ RIhom+,sub (F , RJM∞ ×R∞ G) .

(3) The functor


+
(·) ⊗ (·) : EbIR−c (IkM∞ ×R∞ ) × EbIR−c (IkM∞ ×R∞ ) → DbIR−c (IkM∞ ×R∞ )

is well defined.
(4) For any F1 , F2 ∈ DbIR−c (IkM∞ ×R∞ ), there exits an isomorphism in Db (ksub
M∞ ×R∞ )
 
+ +
λM∞ ×R∞ F1 ⊗ λM∞ ×R∞ F2 ≃ λM∞ ×R∞ F1 ⊗ F2 .

3
Proof. (1) Let us denote by S the closure of {(t1 , t2 , t3 ) ∈ R3 | t1 + t2 + t3 = 0} in R ,
and consider the morphisms pe1 , pe2 , µ e : S → R given by pe1 (t1 , t2 , t3 ) = t1 , pe2 (t1 , t2 , t3 ) =
e(t1 , t2 , t3 ) = t1 + t2 = −t3 . We shall denote by the same symbols the corresponding
t2 , µ
morphisms M̌ × S → M̌ × R. Then there exists a commutative diagram

k
M∞ × R2∞ / M̌ × S

u u
e
 
M∞ × R∞ jM∞ ×R∞
/ M̌ × R,

where u = p1 , p2 , µ and k is the morphism associated by the embedding R2 ֒→ S, (t1 , t2 ) 7→


(t1 , t2 , −t1 − t2 ). Note that for any F1 , F2 ∈ Db (IkM∞ ×R∞ ) there exists an isomorphism
+
−1
F1 ⊗ F2 ≃ jM ∞ ×R∞
Re p−1
µ!! (e e−1
1 RjM∞ ×R∞ !! F1 ⊗ p 2 RjM∞ ×R∞ !! F2 ).

44
This assertion can be proved as similarly to [DK16, Lem. 4.3.9]. Then, we have isomor-
phism
+
−1 −1 −1

IM∞ ×R∞ F1 ⊗IM∞ ×R∞ F2 ≃ jM ∞ ×R∞
Re
µ !! p
e 1 RjM ∞ ×R ∞ !! I M ∞ ×R ∞ F 1 ⊗ p
e 2 RjM ∞ ×R ∞ !! I M ∞ ×R ∞ F 2 .

Moreover, we have isomorphisms


−1
pe−1
1 RjM∞ ×R∞ !! IM∞ ×R∞ F1 ≃ pe−1
1 RjM∞ ×R∞ !! jM∞ ×R∞ IM̌×R RjM∞ ×R∞ !! F1
≃ pe−1
1 (ιM̌ ×R kM ×R ⊗ IM̌×R RjM∞ ×R∞ !! F1 )
≃ pe−1
1 (IM̌ ×R ρM̌ ×R∗ kM ×R ⊗ IM̌×R RjM∞ ×R∞ !! F1 )

≃ pe−1
1 IM̌ ×R ρM̌ ×R∗ kM ×R ⊗ RjM∞ ×R∞ !! F1

≃ IM̌ ×S pe−1
1 ρM̌ ×R∗ kM ×R ⊗ RjM∞ ×R∞ !! F1

≃ IM̌ ×S pe−1
1 RjM∞ ×R∞ !! jM−1
×R Rj M ∞ ×R ∞ !! F 1
 ∞

≃ IM̌ ×S pe−11 Rj F
M∞ ×R∞ !! 1 ,

−1
where in the second (resp. sixth) isomorphism we used the fact that RjM∞ ×R∞ !! jM ∞ ×R∞

−1
ιM̌ ×R kM ×R ⊗ (·) (resp. RjM∞ ×R∞ !! jM∞ ×R∞ ≃ ρM̌ ×R∗ kM ×R ⊗ (·)). See also the end of
Subsection 2.5. In a similar way, we have an isomorphism

pe−1 e−1
2 RjM∞ ×R∞ !! IM∞ ×R∞ F2 ≃ IM̌×S p 2 RjM∞ ×R∞ !! F2 .

Hence, there exist isomorphisms


+
−1

IM∞ ×R∞ F1 ⊗ IM∞ ×R∞ F2 ≃ jM ∞ ×R∞
µ!!IM̌ ×S pe−1
Re e−1
1 RjM∞ ×R∞ !! F1 ⊗ p 2 RjM∞ ×R∞ !! F2
−1

≃ jM I
∞ ×R∞ M̌ ×R
µ!! pe−1
Re e−1
1 RjM∞ ×R∞ !! F1 ⊗ p 2 RjM∞ ×R∞ !! F2
−1

≃ IM∞ ×R∞ jM ∞ ×R∞
µ!! pe−1
Re e−1
1 RjM∞ ×R∞ !! F1 ⊗ p 2 RjM∞ ×R∞ !! F2 .

Moreover, we have
−1

jM ∞ ×R∞
Reµ!! pe−1
1 RjM∞ ×R∞ !! F1 ⊗ pe−1
2 RjM∞ ×R∞ !! F2

≃ Rµ!! k −1 Rk!! p−1 −1
1 F1 ⊗ Rk!! p2 F2

≃ Rµ!! p−1 −1
1 F 1 ⊗ p2 F 2
 
+ +
and hence we have IM∞ ×R∞ F1 ⊗ IM∞ ×R∞ F2 ≃ IM∞ ×R∞ F1 ⊗ F2 .

(2) For any F0 ∈ Db (ksub


M∞ ×R∞ ), there exist isomorphisms

+

HomDb (ksub
M ) F 0 , RJ M ∞ ×R ∞ RIhom (I M ∞ ×R ∞ F , G)
∞ ×R∞
 
+
≃ HomDb (ksub
M∞ ×R∞ )
IM∞ ×R∞ F0 ⊗ IM∞ ×R∞ F , G
   
+
≃ HomDb (ksub
M∞ ×R∞ )
IM∞ ×R∞ F0 ⊗ F , G
+,sub

≃ HomDb (ksub
M ×R ) F 0 , RIhom (F , RJ M ∞ ×R ∞ G) ,
∞ ∞

45
where in the first and last isomorphisms we used Proposition 3.6 (1) and Proposition
3.8, and the second isomorphism follows from (1). Hence we have an isomorphism
RJM∞ ×R∞ RIhom+ (IM∞ ×R∞ F , G) ≃ RIhom+,sub (F , RJM∞ ×R∞ G) .
(3) Let F1 , F2 ∈ DbIR−c (IkM∞ ×R∞ ). By Proposition 3.6 (2), there exist F1 , F2 ∈ Db (ksub
M∞ )
such that F1 ≃ IM∞ ×R∞ F1 , F2 ≃ IM∞ ×R∞ F2 . Moreover, by using the first assertion, we
have
+ +
F1 ⊗ F2 ≃ IM∞ ×R∞ F1 ⊗ IM∞ ×R∞ F1 ≃ IM∞ ×R∞ (F1 ⊗ F2 ) .
+
This implies that F1 ⊗ F2 ∈ DbIR−c (IkM∞ ×R∞ ). The proof is completed.
(4) This assertion follows from Propositions 3.6 (2) and Proposition 3.12.

3.3 Enhanced Subanalytic Sheaves


In this subsection, let us define enhanced subanalytic sheaves.
Let M∞ = (M, M̌ ) be a real analytic bordered space and set R∞ := (R, R) for R :=
R ⊔ {−∞, +∞}. Let us set

Eb (ksub b sub −1 b sub


M∞ ) := D (kM∞ ×R∞ )/π D (kM∞ )

and we shall call an object of Eb (ksub 7


M∞ ) an enhanced subanalytic sheaf on M∞ . The
category π −1 Db (kM∞ ) can be characterized as follows.

Lemma 3.13. For F ∈ Db (ksub


M∞ ×R∞ ), the following five conditions are equivalent

(i) F ∈ π −1 Db (ksub
M∞ ),

∼ +
(ii) F −→ kM ×R [1] ⊗ F ,

(iii) RIhom+,sub (kM ×R [1], F ) −→ F ,
+
(iv) k{t≥0} ⊕ k{t≤0} ⊗ F ≃ 0,

(v) RIhom+,sub k{t≥0} ⊕ k{t≤0} , F ≃ 0.

Proof. By using a distinguished triangle


+1
k{t≥0} ⊕ k{t≤0} −→ k{t=0} −→ kM ×R [1] −→
+
and the fact that k{t=0} ⊗ F ≃ F (resp. F ≃ RIhom+,sub (k{t=0} , F )), we have the condi-
tion (ii) (resp. (iii)) is equivalent to the condition (iv) (resp. (v)).
Let us prove (i), (ii), (iii) are equivalent. Let us assume the condition (ii) (resp. (iii))
∼ + ∼
is satisfied. Then we have F −→ kM ×R [1] ⊗ F (resp. RIhom+,sub (kM ×R [1], F ) −→ F ).
∼ ∼
By Corollary 3.11, F −→ π −1 Rπ!! [1]F (resp. π −1 Rπ∗ F ←− F ). Hence, the condition (i)
is satisfied.
7
It seems that a object of Db (ksub
M∞ ×R∞ ) is called an enhanced subanalytic sheaf, in [Kas16].

46
Let us assume the condition (i) is satisfied. By using Proposition 3.7 (2)(iii), we have
IM∞ ×R∞ F ∈ π −1 Db (IkM∞ ), and hence by [DK16, Lem. 4.4.3] we have

π −1 Rπ∗ IM∞ ×R∞ F −→ IM∞ ×R∞ F ,

IM∞ ×R∞ F −→ π ! Rπ!! IM∞ ×R∞ F .

By using Proposition 3.7 (3)(iv), (v) (resp. (2)(iv), (v)) and Proposition 3.6 (2), we have
∼ ∼
π −1 Rπ∗ F −→ F (resp. F −→ π ! Rπ!! F ).

By Corollary 3.11, this implies that the condition (ii) (resp. (iii)) is satisfied.
Therefore, the proof is completed.
Let us prove the quotient functor

Qsub b sub b sub


M∞ : D (kM∞ ×R∞ ) → E (kM∞ )

has fully faithful left and right adjoints. By Corollary 3.11, two functors

LE,sub b sub b sub


M∞ : E (kM∞ ) → D (kM∞ ×R∞ ),

RE,sub b sub b sub


M∞ : E (kM∞ ) → D (kM∞ ×R∞ )

which are defined by


+
LE,sub sub
M∞ (QM∞ (·)) := ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ (·),

RE,sub sub
M∞ (QM∞ (·)) := RIhom
+,sub
(ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), ·)

are well defined.

Lemma 3.14. The functors LE,sub E,sub


M∞ , RM∞ induce equivalences of categories

∼ + ∼
LE,sub b sub b sub
M∞ : E (kM∞ ) −→ {F ∈ D (kM∞ ×R∞ ) | ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ F −→ F },
∼ ∼
RE,sub b sub b sub
M∞ : E (kM∞ ) −→ {F ∈ D (kM∞ ×R∞ ) | F −→ RIhom
+,sub
(ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F )},

respectively.
E,sub E,sub
Moreover, the quotient functor admits a left (resp. right) adjoint LM ∞
(resp. RM ∞
).

Proof. By Corollary 3.11 and the fact that for any K ∈ Eb (ksub
M∞ ) there exist isomorphisms

+ ∼
ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ LE,sub E,sub
M∞ K −→ LM∞ K,
 

RE,sub
M∞ K −→ RIhom +,sub
ρ (k
M∞ ×R∞ ∗ {t≥0} ⊕ k {t≤0} ), R E,sub
M∞ K ,

functors
+ ∼
LE,sub b sub b sub
M∞ : E (kM∞ ) → {F ∈ D (kM∞ ×R∞ ) | ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ F −→ F },

RE,sub b sub b sub
M∞ : E (kM∞ ) → {F ∈ D (kM∞ ×R∞ ) | F −→ RIhom
+,sub
(ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F )}

47
are well defined. Let K ∈ Eb (ksub b sub
M∞ ). Then there exist F ∈ D (kM∞ ×R∞ ) such that
K ≃ Qsub
M∞ F . Let us prove that
 
+
QM∞ ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ F ≃ Qsub
sub
M∞ (F ),

∼ 
F −→ QsubM∞ RIhom
+,sub
(ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F ) .
+1
Since there exists a distinguished  ⊕ k{t≤0} −→ k{t=0} −→ kM ×R [1] −→,
 triangle k{t≥0}
+ 
it is enough to show that Qsub
M∞ kM ×R [1] ⊗ F ≃ Qsub
M∞ RIhom
+,sub
(kM ×R [1], F ) ≃ 0.
These assertions follows from Corollary 3.11. So that, we have
 
+
sub E,sub
QM∞ LM∞ K ≃ QM∞ ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ F ≃ Qsub
sub
M∞ (F ) ≃ K,

E,sub

Qsub sub
M∞ RM∞ K ≃ QM∞ RIhom
+,sub
(ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F ) ≃ Qsub
M∞ (F ) ≃ K.
E,sub E,sub
Hence we have Qsub
M∞ ◦ LM∞ ≃ id, Qsub
M∞ ◦ R M∞ ≃ id. Moreover, it is clear that
+ ∼
for any G1 ∈ {F ∈ Db (ksub
M∞ ×R∞ ) | ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ F −→ F } and any

G2 ∈ {F ∈ D (kM∞ ×R∞ ) | F −→ RIhom+,sub (ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F )}, we have
b sub

LE,sub sub E,sub sub


M∞ QM∞ (G1 ) ≃ G1 and RM∞ ◦ QM∞ (G2 ) ≃ G2 . Therefore, there exist equivalences of
categories
∼ + ∼
LE,sub b sub b sub
M∞ : E (kM∞ ) −→ {F ∈ D (kM∞ ×R∞ ) | ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ) ⊗ F −→ F },
∼ ∼
RE,sub b sub b sub
M∞ : E (kM∞ ) −→ {F ∈ D (kM∞ ×R∞ ) | F −→ RIhom
+,sub
(ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F )}.
E,sub
Let us prove that the quotient functor admits a left (resp. right) adjoint LM ∞

(resp. RE,sub
M∞ ). Let F ∈ D b sub
(kM∞ ×R∞ ) and K ∈ E b sub
(kM∞ ). Since functors LE,sub
M∞ , R E,sub
M∞
induce fully faithful functors
LE,sub b sub b sub
M∞ : E (kM∞ ) ֒→ D (kM∞ ×R∞ ),

RE,sub b sub b sub


M∞ : E (kM∞ ) ֒→ D (kM∞ ×R∞ ),

there exist isomorphisms


  
E,sub sub E,sub
HomEb (ksub
M ) Qsub
M∞ F , K ≃ HomDb (ksub
M ) LM ∞
QM∞ F , LM ∞
K ,
∞ ∞ ×R∞
  
E,sub E,sub sub
HomEb (ksub
M ) K, Qsub
M∞ F ≃ HomDb (ksub
M ) RM ∞
K, RM ∞
QM∞ F .
∞ ∞ ×R∞

Note that there exist isomorphisms in Db (ksub


M∞ ×R∞ )
 
E,sub E,sub
RIhom+,sub ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), LM ∞
K ≃ RM ∞
K,
+
ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} ⊗ RE,sub E,sub
M∞ K ≃ LM∞ K.

So that, we have
E,sub sub E,sub E,sub
HomDb (ksub
M ∞ ×R∞
) (LM∞ QM∞ F , LM∞ K) ≃ HomDb (ksub
M ∞ ×R∞
) (F , RM∞ K),
E,sub E,sub sub E,sub
HomDb (ksub
M ∞ ×R∞
) (RM∞ K, RM∞ QM∞ F ) ≃ HomDb (ksub
M ∞ ×R∞
) (LM∞ K, F ),

48
and hence there exist isomorphisms
E,sub
HomEb (ksub
M )
(Qsub
M∞ F , K) ≃ HomDb (ksub
M ) (F , RM∞ K),
∞ ∞ ×R∞
E,sub
HomEb (ksub
M )
(K, Qsub
M∞ F ) ≃ HomDb (ksub
M ) (LM∞ K, F ).
∞ ∞ ×R∞

E,sub E,sub
Therefore, the quotient functor admits a left (resp. right) adjoint LM ∞
(resp. RM ∞
).
E,sub E,sub
We sometimes denote Qsub
M∞ (resp. LM∞ , RM∞ ) by Q
sub
(resp. LE,sub , RE,sub) for
short. Let us set
E,sub
E≤0 (Csub b sub ≤0 sub
M∞ ) = {K ∈ E (CM∞ ) | LM∞ K ∈ D (CM∞ ×R∞ )},
E,sub
E≥0 (Csub b sub ≥0
M∞ ) = {K ∈ E (CM∞ ) | LM∞ K ∈ D (CM∞ ×R∞ )},
sub


where D≤0 (Csub ≥0 sub
M∞ ×R∞ ), D (CM∞ ×R∞ ) is
 the standard t-structure on Db (Csub M∞ ×R∞ ).
Note that a pair E≤0 (Csub
M∞ ), E ≥0
(C sub
M∞ ) is a standard t-structure on E b
(C sub
M∞ ). We
denote by
Hn : Eb (Csub 0
M∞ ) → E (CM∞ )
sub

the n-th cohomology functor, where we set E0 (Csub ≤0 sub ≥0


M∞ ) := E (CM∞ ) ∩ E (CM∞ ).
sub

By Proposition 3.9, the convolution functors can be lifted to the triangulated category
+
Eb (ksub
M∞ ). We denote them by the same symbols ⊗, RIhom
+,sub
. Namely, we obtain
functors
+
(·) ⊗ (·) : Eb (ksub b sub b sub
M∞ ) × E (kM∞ ) → E (kM∞ ),
RIhom+,sub (·, ·): Eb (ksub
M∞ )
op
× Eb (ksub b sub
M∞ ) → E (kM∞ )

which are defined by


+ 
Qsub sub sub −1 −1
M∞ (F ) ⊗ QM∞ (G) := QM∞ Rµ!! (p1 F ⊗ p2 G) ,

RIhom+,sub (Qsub sub sub
M∞ (F ), QM∞ (G)) := QM∞ Rp1∗ RIhom
sub −1
(p2 F , µ! G) ,

for F , G ∈ Db (ksub
M∞ ×R∞ ). Moreover, by Proposition 3.4 (4), for a morphism f : M∞ → N∞
of real analytic bordered spaces, the following functors are well defined

Ef∗ : Eb (ksub b sub sub sub


M∞ ) → E (kN∞ ), QM∞ F 7−→ QN∞ (RfR∞ ∗ F )

Ef −1 : Eb (ksub
N∞ ) → E b sub
(kM∞ ), Qsub
N∞ G −
7 → Qsub
M∞ f −1
R∞ G ,
Ef!! : Eb (ksub b sub sub sub
M∞ ) → E (kN∞ ), QM∞ F 7−→ QN∞ (RfR∞ !! F ) ,

Ef ! : Eb (ksub b sub sub sub !
N∞ ) → E (kM∞ ), QN∞ G 7−→ QM∞ fR∞ G .

Let us define external hom functors

RIhomE,sub (·, ·) : Eb (ksub


M∞ )
op
× Eb (ksub b sub
M∞ ) → D (kM∞ ),
RHomE,sub (·, ·) : Eb (ksub
M∞ )
op
× Eb (ksub b
M∞ ) → D (kM ),

RHomE,sub (·, ·) : Eb (ksub


M∞ )
op
× Eb (ksub b
M∞ ) → D (k),

49
by

RIhomE,sub Qsub sub
M∞ F1 , QM∞ F2 := Rπ∗ RIhom(F1 , F2 ),
 
RHomE,sub Qsub sub
M∞ F1 , QM∞ F2 := ρM∞ ∗ RIhom
E
Qsub sub
M∞ F1 , QM∞ F2 ,
 
RHomE,sub Qsub sub E sub sub
M∞ F1 , QM∞ F2 := RΓ M; RHom (QM∞ F1 , QM∞ F2 ) ,

for F1 , F2 ∈ Eb (ksub b sub


M∞ ×R∞ ). Note that for any K1 , K2 ∈ E (kM∞ ), we have

HomEb (ksub
M )
(K1 , K2 ) ≃ H0 RHomE,sub (K1 , K2 ) .

Moreover, for F0 ∈ Db (Csub b sub


M∞ ) and F ∈ E (CM∞ ×R∞ ), the objects

π −1 F0 ⊗ Qsub sub −1
M∞ F := QM∞ π F0 ⊗ F ,
 
RIhomsub π −1 F0 , Qsub sub
M∞ F := QM∞ RIhom
sub
(π −1 F0 , Qsub
M∞ F

are well defined and hence the following functors are well defined

π −1 (·) ⊗ (·) : Db (ksub b sub b sub


M∞ ) × E (kM∞ ) → E (kM∞ ),

RIhomsub π −1 (·), · : Db (ksub
M∞ )
op
× Eb (ksub b sub
M∞ ) → E (kM∞ ).

At the end of this subsection, let us prove that these functors have several properties
as similar to the classical sheaves.
Proposition 3.15. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces.

(1) (i) For any K1 , K2 , K3 ∈ Eb (ksub


M∞ ), one has
 
+,sub
+ 
RIhom K1 ⊗ K2 , K3 ≃ RIhom+,sub K1 , RIhom+,sub (K2 , K3 ) ,
 
E,sub
+ 
RIhom K1 ⊗ K2 , K3 ≃ RIhomE,sub K1 , RIhom+,sub (K2 , K3 ) ,
 
E,sub
+ 
RHom K1 ⊗ K2 , K3 ≃ RIhomE,sub K1 , RIhom+,sub (K2 , K3 ) ,
 
E,sub
+ 
RHom K1 ⊗ K2 , K3 ≃ RHomE,sub K1 RIhom+,sub (K2 , K3 ) ,
 
+
+,sub

HomEb (ksub
M ) K 1 ⊗ K 2 , K 3 ≃ Hom E b (ksub ) K1 , RIhom
N
(K 2 , K 3 ) .
∞ ∞

(ii) For any K ∈ Eb (ksub b sub


M∞ ) and any L ∈ E (kN∞ ), one has

Ef∗ RIhom+,sub Ef −1L, K ≃ RIhom+,sub (L, Ef∗ K) ,

Rf∗ RIhomE,sub Ef −1L, K ≃ RIhomE,sub (L, Ef∗ K) ,

Rf∗ RHomE,sub Ef −1L, K ≃ RIhomE,sub (L, Ef∗ K) ,

RHomE,sub Ef −1L, K ≃ RHomE,sub (L, Ef∗ K) ,
−1

HomEb (ksub
M )
Ef L, K ≃ HomEb (ksub
N )
(L, Ef∗ K) .
∞ ∞

50
(iii) For any K ∈ Eb (ksub b sub
M∞ ) and any L ∈ E (kN∞ ), one has

RIhom+,sub (Ef!! K, L) ≃ Ef∗ RIhom+,sub K, Ef ! L ,

RIhomE,sub (Ef!! K, L) ≃ Rf∗ RIhomE,sub K, Ef ! L ,

RHomE,sub (Ef!! K, L) ≃ Rf∗ RHomE,sub K, Ef !L ,

HomE,sub (Ef!! K, L) ≃ HomE,sub K, Ef !L ,
!

HomEb (ksub
N ) (Ef!! K, L) ≃ HomEb (ksub ) K, Ef L,
M
.
∞ ∞

(2) For any K, K1, K2 ∈ Eb (ksub b


M∞ ) and any L, L1 , L2 ∈ E (kN∞ ), one has,
 
+ +
−1
Ef K1 ⊗ K2 ≃ Ef −1 K1 ⊗ Ef −1 K2 ,
 
+ +
−1
Ef!! K ⊗ Ef L ≃ Ef!! K ⊗ L,

Ef ! RIhom+,sub (L1 , L2 ) ≃ RIhom+,sub Ef −1L1 , Ef ! L2

(3) For a cartesian diagram


′ f′ ′
M∞ / N∞
g′ g
 
M∞ f
/ N∞

and any K ∈ Eb (ksub


M∞ ), one has

Eg −1Ef!! K ≃ Ef!!′ Eg ′−1K,


Eg ! Ef∗ K ≃ Ef∗′ Eg ′! K.

(4) (i) For any F ∈ Db (ksub b sub


M∞ ) and any K1 , K2 ∈ E (kM∞ ), one has
 
RIhom+,sub π −1 F ⊗ K1 , K2 ≃ RIhom+,sub K1 , RIhomsub (π −1 F , K2

≃ RIhomsub π −1 F , RIhom+,sub (K1 , K2 ,
 
RIhomE,sub π −1 F ⊗ K1 , K2 ≃ RIhomE,sub K1 , RIhomsub (π −1 F , K2

≃ RIhomsub F , RIhomE,sub (K1 , K2 ,
 
RHomE,sub π −1 F ⊗ K1 , K2 ≃ RHomE,sub K1 , RIhomsub (π −1 F , K2

≃ RHomsub F , RIhomE,sub (K1 , K2 ,
 
RHomE,sub π −1 F ⊗ K1 , K2 ≃ RHomE,sub K1 , RIhomsub (π −1 F , K2

≃ RHomsub F , RIhomE,sub (K1 , K2 ,
−1
 sub −1

HomEb (ksub
M∞ ) π F ⊗ K 1 , K 2 ≃ HomEb (ksub ) K1 , RIhom
M∞
(π F , K 2
E,sub

≃ HomDb (ksub
M )
F , RIhom (K 1 , K 2 ,

51
(ii) For any F ∈ Db (ksub b sub b sub
M∞ ), any G ∈ D (kN∞ ), any K ∈ E (kM∞ ) and any L ∈
Eb (ksub
N∞ ), one has

Ef −1 π −1 F ⊗ L ≃ π −1 f −1 F ⊗ Ef −1L,

Ef!! π −1 F ⊗ Ef −1L ≃ π −1 Rf!! F ⊗ L,

Ef!! π −1 f −1 F ⊗ K ≃ π −1 F ⊗ Ef!! K,
 
Ef ! RIhom+,sub π −1 G, L ≃ RIhom+,sub π −1 f −1 G, Ef ! L .

(iii) For any F ∈ Db (ksub b sub


M∞ ) and any K, L ∈ E (kM∞ ), one has
 
+ +
π F ⊗ K ⊗ L ≃ π −1 F ⊗ K ⊗ L,
−1

 
RIhomsub π −1 F , RIhom+,sub (K, L) ≃ RIhom+,sub π −1 F ⊗ K, L

≃ RIhom+,sub K, RIhomsub (π −1 F , L ,
 
sub
+ 
Rπ∗ RIhom K ⊗ L, F ≃ Rπ∗ RIhomsub K, RIhom+,sub (L, F ) .

Proof. (1)(i) The first (resp. second) assertion follows from Proposition 3.8 (resp. 3.4 (2)).
The third (resp. fourth, fifth, sixth) assertion follows from the second (resp. third, fourth,
fifth) one.
(ii) The first (resp. second) assertion follows from Proposition 3.8 (resp. 3.4 (2)). The third
(resp. fourth, fifth, sixth) assertion follows from the second (resp. third, fourth, fifth) one.

(iii) The first (resp. second) assertion follows from Proposition 3.8 (resp. 3.4 (2)). The
third (resp. fourth, fifth, sixth) assertion follows from the second (resp. third, fourth,
fifth) one.
(2) The three assertions follow from Proposition 3.8.
(3) The three assertions follow from Proposition 3.4 (4).
(4)(i) By the definition we have
  
E,sub E,sub
RIhom+,sub π −1 F ⊗ K1 , K2 ≃ Qsub
M∞ RIhom+,sub
π −1
F ⊗ LM∞ K 1 , R M∞ K 2

and hence by Proposition 3.9 there exist isomorphisms


 
E,sub E,sub
Qsub
M∞ RIhom +,sub
π −1
F ⊗ L M∞ K 1 , R M∞ K 2
 
E,sub E,sub
≃ Qsub
M∞ RIhom +,sub
LM∞ K 1 , RIhom sub
(π −1
F , R M∞ K 2

≃ RIhom+,sub K1 , RIhomsub (π −1 F , K2
and
 
E,sub E,sub
Qsub
M∞ RIhom+,sub
π −1
F ⊗ LM∞ K 1 , R M∞ K 2
 
E,sub E,sub
≃ Qsub
M∞ RIhom sub
π −1
F , RIhom +,sub
(LM∞ K 1 , R M∞ K 2 )

≃ RIhomsub π −1 F , RIhom+,sub(K1 , K2 .

52
By the definition we have
  
E,sub E,sub
RIhomE,sub π −1 F ⊗ K1 , K2 ≃ Rπ∗ RIhomsub π −1 F ⊗ LM ∞
K 1 , LM∞ K 2

and hence by Proposition 3.4 (2) there exist isomorphisms


 
Rπ∗ RIhomsub π −1 F ⊗ LE,sub M∞ K 1 , L E,sub
M∞ K 2
  
≃Rπ∗ RIhomsub LE,subM∞ K 1 , RIhom sub
π −1
F , LE,sub
M∞ K 2

≃RIhomE,sub K1 , RIhomsub (π −1 F , K2

and
 
Rπ∗ RIhomsub π −1 F ⊗ LE,sub
M∞ K 1 , L E,sub
M∞ K 2
  
sub −1 sub E,sub E,sub
≃Rπ∗ RIhom π F , RIhom LM∞ K1 , LM∞ K2
  
E,sub E,sub
≃RIhomsub F , Rπ∗ RIhomsub LM ∞
K 1 , L M∞ K 2
sub E,sub

≃RIhom F , RIhom (K1 , K2 .

The third (resp. fourth, fifth) assertion follows from the second (resp. third, fourth) one.
The three assertions of (ii) follow from Proposition 3.4 (3), (4).
The three assertions of (iii) follow from Proposition 3.9.
Remark 3.16. One can consider enhanced subanalytic sheaves with ring actions and
Proposition 3.15 can be extended to the ones with ring actions. We shall skip the expla-
nation of them.

3.4 Relation between Enhanced Ind-Sheaves and Enhanced


Subanalytic Sheaves
In this subsection, we shall explain a relation between enhanced subanalytic sheaves and
enhanced ind-sheaves. Theorem 3.18 and 3.23 are main results of this paper.
Let M∞ = (M, M̌ ) be a real analytic manifold. Let us consider a quotient category

EbIR−c (IkM∞ ) := DbIR−c (IkM∞ ×R∞ )/π −1 DbIR−c (IkM∞ ).

Note that this is a full triangulated subcategory of Eb (IkM∞ ) by using π −1 DbIR−c (IkM∞ ) =
π −1 Db (IkM∞ ) ∩ DbIR−c (IkM∞ ×R∞ ) and [KS90, Prop. 1.6.10]. Note also that kEM∞ ∈
EbIR−c (IkM∞ ). Moreover, Propositions 3.17 below follows from Lemma 3.5 and Propo-
sition 3.12 (3).
Proposition 3.17. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces
associated with a morphism fˇ: M̌ → Ň of real analytic manifolds. The functors below
are well defined:

53
(1) eM∞ : DbIR−c (IkM ) → EbIR−c (IkM∞ ),
+
(2) (·) ⊗ (·) : EbIR−c (IkM∞ ) × EbIR−c (IkM∞ ) → EbIR−c (IkM∞ ),

(3) Ef −1 : EbIR−c (IkN∞ ) → EbIR−c (IkM∞ ),

(4) Ef!! : EbIR−c (IkM∞ ) → EbIR−c (IkN∞ ),

(5) Ef ! : EbIR−c (IkN∞ ) → EbIR−c (IkM∞ ).

By Proposition 3.7 (2)(iii), (3)(v), the following functors are well defined
E
IM ∞
: Eb (ksub b sub
M∞ ) → E (IkM∞ ), QM∞ F 7→ QM∞ IM∞ ×R∞ F ,
E
JM ∞
: Eb (IkM∞ ) → Eb (ksub sub
M∞ ), QM∞ F 7→ QM∞ RJM∞ ×R∞ F.

Theorem 3.18. Let M∞ = (M, M̌ ) be a real analytic bordered space. Then we have
E E ∼
(1) a pair (IM ∞
, JM ∞
) is an adjoint pair and there exists a canonical isomorphism id −→
E E
J M ∞ ◦ IM ∞ ,

(2) there exists an equivalence of triangulated categories:


E
IM ∞
b /
E (ksub
M∞ ) o ∼ EbIR−c (IkM∞ ).
E
JM ∞

Proof. (1) For any K ∈ Eb(ksub b


M∞ ) and any L ∈ E (IkM∞ ), there exist isomorphisms

  
E E,sub
HomEb (IkM∞ ) IM∞ K, L ≃ HomEb (IkM∞ ) QM∞ IM∞ ×R∞ LM∞ K, L

≃ HomEb (ksub
M∞ )
K, Qsub E
M∞ RJM∞ ×R∞ RM∞ L
sub

≃ HomEb (ksub
M ) K, J M ∞
L ,

where in the second isomorphism we used Proposition 3.6 (1) and Lemma 3.14. This
E E
implies that a pair (IM ∞
, JM ∞
) is an adjoint pair. Moreover, by Proposition 3.6 (1) it is
∼ E E
clear that id −→ JM∞ ◦ IM∞ .
(2) Since the functor IM∞ ×R∞ : Db (ksub b
M∞ ×R∞ ) → DIR−c (IkM∞ ×R∞ ) is well defined, for any
K ∈ Eb (ksub
M∞ ) we have

E
IM ∞
K = QM∞ IM∞ ×R∞ LE,sub b
M∞ K ∈ EIR−c (IkM∞ ).

Let L ∈ EbIR−c (IkM∞ ). Then there exists G ∈ DbIR−c (IkM∞ ×R∞ ) such that L ≃ QM∞ G and
hence
E E
IM ∞
RJM ∞
L ≃ QM∞ IM∞ ×R∞ RJM∞ ×R∞ G ≃ QM∞ G ≃ L,
where in the second isomorphism we used Proposition 3.6 (2). Therefore, the proof is
completed.

54

We will denote by λEM∞ : EbIR−c (IkM∞ ) −→ Eb (ksub
M∞ ) the inverse functor of
E b sub ∼ b
IM∞ : E (kM∞ ) −→ EIR−c (IkM∞ ). There exists a commutativity between the various
functors and functors I E , J E , λE as below.

Proposition 3.19. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces


associated with a morphism fˇ: M̌ → Ň of real analytic manifolds.

(1) For any K, K1, K2 ∈ Eb (ksub b


M∞ ) and any L ∈ E (IkM∞ ), we have
 
+ +
E E E
(i) IM∞ K1 ⊗ K2 ≃ IM ∞
K1 ⊗ I M ∞
K2 ,
E
(ii) JM ∞
RIhom+ (IM
E

K, L) ≃ RIhom+,sub (K, JM
E

L)
(iii) RJM∞ RIhomE (IM
E

K, L) ≃ RIhomE,sub (K, JM
E

L).

(2) For any K ∈ Eb (ksub b sub


M∞ ) and any L ∈ E (kN∞ ), we have

E
(i) IM ∞
Ef −1 L ≃ Ef −1 INE∞ L,
E
(ii) Ef!! IM ∞
K ≃ INE∞ Ef!! K,
E
(iii) IM ∞
Ef ! L ≃ Ef ! INE∞ L.

(3) For any K ∈ Eb (IkM∞ ) and any L ∈ Eb (IkN∞ ), we have


E
(i) Ef∗ JM ∞
K ≃ JNE∞ Ef∗ K,
E
(ii) JM ∞
Ef ! L ≃ Ef ! JNE∞ L.

(4) For any K ∈ EbIR−c (IkM∞ ) and any L ∈ EbIR−c (IkN∞ ), we have

(i) λEM∞ Ef −1 L ≃ Ef −1 λEN∞ L,


(ii) Ef!! λEM∞ K ≃ λEN∞ Ef!! K,
 
+ +
(iii) λM∞ K1 ⊗ K2 ≃ λEM∞ K1 ⊗ λEM∞ K2 .
E

Proof. Let us denote by fR∞ : M∞ × R∞ → N∞ × R∞ the morphism f × idR∞ of real


analytic bordered spaces.
(1) Let K, K1 , K2 ∈ Eb (ksub b b
M∞ ), L ∈ E (IkM∞ ), then there exist F , F1 , F2 ∈ D (kM∞ ×R∞ ),
b sub sub sub
G ∈ D (IkM∞ ×R∞ ) such that K ≃ QM∞ F , K1 ≃ QM∞ F1 , K2 ≃ QM∞ F2 , L ≃ QM∞ G.
(i) By Proposition 3.12 (1), we have
   
+ +
E
IM∞ K1 ⊗ K2 ≃ QM∞ IM∞ ×R∞ F1 ⊗ F2
 
+
≃ QM∞ IM∞ ×R∞ F1 ⊗ IM∞ ×R∞ F2
+
E E
≃ IM ∞
K1 ⊗ I M ∞
K2 .

55
(ii) By Proposition 3.12 (2), we have
E
JM ∞
RIhom+ (IM
E

K, L) ≃ Qsub +
M∞ RJM∞ ×R∞ RIhom (IM∞ ×R∞ F , G)
≃ Qsub +
M∞ RIhom (F , RJM∞ ×R∞ G)
≃ RIhom+,sub (K, JM
E

L).

(iii) By Proposition 3.7 (1), (3)(iv), we have

RJM∞ RIhomE (IM


E

K, L) ≃ RJM∞ π∗ RIhom(IM∞ ×R∞ F , G)
≃ π∗ RJM∞ ×R∞ RIhom(IM∞ ×R∞ F , G)
≃ π∗ RIhomsub (F , RJM∞×R∞ G)
≃ RIhomE,sub (K, JM
E

L).

(2) Let K ∈ Eb (ksub b sub


M∞ ), L ∈ E (kN∞ ). Then there exist F ∈ Db (ksub
M∞ ×R∞ ), G ∈
b sub sub sub
D (kN∞ ×R∞ ) such that K ≃ QM∞ F , L ≃ QN∞ G.
(i) By Proposition 3.7 (2)(iii), we have
E
IM ∞
Ef −1 L ≃ QM∞ IM∞ ×R∞ fR−1

G ≃ QM∞ fR−1 I
∞ N∞ ×R∞
G ≃ Ef −1INE∞ L.

(ii) By Proposition 3.7 (2)(iv), we have


E
Ef!! IM ∞
K ≃ QN∞ fR∞ !! IM∞ ×R∞ F ≃ QN∞ IN∞ ×R∞ fR∞ !! F ≃ Ef!! INE∞ K.

(iii) By Proposition 3.7 (2)(v), we have


E
IM ∞
Ef ! L ≃ QM∞ IM∞ ×R∞ fR! ∞ G ≃ QM∞ fR! ∞ IN∞ ×R∞ G ≃ Ef ! INE∞ L.

(3) Let K ∈ Eb (IkM∞ ), L ∈ Eb (IkM∞ ). Then there exist F ∈ Db (IkM∞ ×R∞ ), G ∈


Db (IkM∞ ×R∞ ) such that K ≃ QM∞ F, L ≃ QM∞ G.
(i) By Proposition 3.7 (3)(iv), we have
E
Ef∗ JM ∞
K ≃ Qsub sub E
N∞ RfR∞ ∗ RJM∞ ×R∞ F ≃ QN∞ RJN∞ ×R∞ RfR∞ ∗ F ≃ JN∞ Ef∗ K.

(ii) By Proposition 3.7 (3)(v), we have


E
JM ∞
Ef ! L ≃ Qsub ! sub ! ! E
M∞ RJM∞ ×R∞ fR∞ G ≃ QM∞ fR∞ RJN∞ ×R∞ G ≃ Ef JN∞ L.

(4) Let K ∈ EbIR−c (IkM∞ ), L ∈ EbIR−c (IkN∞ ).


(i) By using (2)(i) and Theorem 3.18 (2), we have

λEM∞ Ef −1 L ≃ λEM∞ Ef −1INE∞ λEM∞ L ≃ λEM∞ IM


E

Ef −1 λEM∞ L ≃ Ef −1 λEN∞ L.

(ii) By using (2)(ii) and Theorem 3.18 (2), we have

Ef!! λEM∞ K ≃ λEN∞ INE∞ Ef!! λEM∞ K ≃ λEN∞ Ef!! IM


E
λE K ≃ λEN∞ Ef!! K.
∞ M∞

56
(iii) By using (1)(i) and Theorem 3.18 (2), we have
   
+ +
E E E E E E
λ M ∞ K 1 ⊗ K2 ≃ λ M ∞ I M ∞ λ M ∞ K 1 ⊗ I M ∞ λ M ∞ K2
 
+
E E E E
≃ λ M ∞ I M ∞ λ M ∞ K 1 ⊗ λ M ∞ K2
+
≃ λEM∞ K1 ⊗ λEM∞ K2 .

E E
Let us prove that the functors IM ∞
, JM ∞
are preserve the R-constructability. Let us
recall that an enhanced ind-sheaf K on M∞ is R-constructible if for any open subset U
of M which is subanalytic and relatively compact in M̌ there exists F ∈ DbR−c (kU∞ ×R∞ )
+
such that Ei−1 E b
U∞ K ≃ kU∞ ⊗ QU∞ ιU∞ ×R∞ F . We denote by ER−c (IkM∞ ) the category of
R-constructible enhanced ind-sheaves.
We shall set
!
kE,sub sub
M∞ := QM∞ lim ρM∞ ×R∞ ∗ k{t≥a} ∈ Eb (ksub
M∞ ).
−→
a→+∞

E
Lemma 3.20. There exist an isomorphism IM kE,sub ≃ kEM∞ in Eb (IkM∞ ) and an iso-
∞ M∞
E
morphism JM kE ≃ kE,sub
∞ M∞
b sub
M∞ in E (kM∞ ).
!
−1
Proof. By the definition of kEM∞ , we have kEM∞ ≃ QM∞ jM “ lim ” ιM̌ ×R k{t≥a} .
∞ ×R∞ −→
a→+∞
Since IM̌ ×R ◦ ρM̌ ×R∗ ≃ ιM̌ ×R and the fact that a functor I commutes with the inductive
limits, there exist isomorphisms in Db (IkM̌ ×R ):
!
“ lim ” ιM̌ ×R k{t≥a} ≃ IM̌ ×R lim ρM̌ ×R∗ k{t≥a} .
−→ −→
a→+∞ a→+∞

Hence, we have
!
−1
kEM∞ ≃ QM∞ jM I lim ρM̌ ×R∗ k{t≥a}
∞ ×R∞ M̌ ×R −→
a→+∞
!
E
≃ IM Qsub −1
lim jM ρ k
∞ ×R∞ M̌ ×R∗ {t≥a}
∞ M∞ −→
a→+∞
!
E
≃ IM Qsub lim ρM∞ ×R∞ ∗ k{t≥a}
∞ M∞ −→
a→+∞
E
≃ IM kE,sub .
∞ M∞

The second assertion follows from the first assertion and Theorem 3.18 (1).

57
Proposition 3.21. The triangulated category EbR−c (IkM∞ ) is a full triangulated subcate-
gory of EbIR−c (IkM∞ ).
E
Proof. Let K ∈ EbR−c (IkM∞ ). Since a pair (IM ∞
E
, JM ∞
) is an adjoint pair, we have a mor-
E E b
phism IM∞ JM∞ K → K in E (IkM∞ ). Since K is R-constructible, for any open subset U
of M which is subanalytic and relatively compact in M̌ there exists F U ∈ DbR−c (kU∞ ×R∞ )
+
such that Ei−1 E U
U∞ K ≃ kU∞ ⊗ QU∞ ιU∞ ×R∞ F . By Proposition 3.7 (4)(i), Proposition 3.19
(1)(i) and Lemma 3.20, we have
 
+ E,sub +
E U E sub U
kU∞ ⊗ QU∞ ιU∞ ×R∞ F ≃ IU∞ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ F ∈ EbIR−c (IkU∞ ).

This implies that for any open subset U of M which is subanalytic and relatively com-
pact in M̌ , Ei−1 b
U∞ K ∈ EIR−c (IkU∞ ). Hence, by Proposition 3.19 (2)(i), (3)(ii) there exist
isomorphisms
E E

IM ∞
J M ∞
K |U∞ ≃ IUE∞ JUE∞ (K|U∞ ) ≃ K|U∞

for any open subset U of M which is subanalytic and relatively compact in M̌ . This
E ∼
implies that IM J E K −→ K and hence K ∈ EbIR−c (IkM∞ ).
∞ M∞

Definition 3.22. Let M∞ be a real analytic bordered space. We say that an enhanced
subanalytic sheaf K is R-constructible if for any open subset U of M which is subanalytic
and relatively compact in M̌ there exists F ∈ DbR−c (kU∞ ×R∞ ) such that

E,sub +
Ei−1 sub
U∞ K ≃ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ F .

Let us denote by EbR−c (ksub


M∞ ) the category of R-constructible enhanced subanalytic
sheaves.
E E
Theorem 3.23. Let M∞ be a real analytic bordered space. Then the functors IM ∞
, JM ∞
induce an equivalence of categories
E
IM ∞
/
EbR−c (ksub
M∞ ) o ∼ EbR−c (IkM∞ ).
E
JM ∞

Proof. It is enough to show that the following functors are well defined:
E
IM ∞
: EbR−c (ksub b
M∞ ) → ER−c (IkM∞ ),
E
JM ∞
: EbR−c (IkM∞ ) → EbR−c (ksub
M∞ ).

E
By Proposition 3.21, a functor IM ∞
: EbR−c (ksub b
M∞ ) → ER−c (IkM∞ ) is well defined.
b
Let K ∈ ER−c (IkM∞ ) and U an open subset of M which is subanalytic and relatively
+
compact in M̌ . Then there exists F U ∈ DbR−c (kU∞ ×R∞ ) such that Ei−1 E
U∞ K ≃ kU∞ ⊗

58
QM∞ ιM∞ ×R∞ F U . By Propositions 3.7 (3)(i), 3.19 (4)(i), (iii) and Lemma 3.20, there exist
isomorphisms

Ei−1 E E −1
U∞ λM∞ K ≃ λU∞ (EiU∞ K)
 
+
E E U
≃ λU∞ kU∞ ⊗ QU∞ ιU∞ ×R∞ F
+
≃ λEU∞ kEU∞ ⊗ λEU∞ QU∞ ιU∞ ×R∞ F U
+
≃ kE,sub sub U
U∞ ⊗ QU∞ ρU∞ ×R∞ ∗ F .

Therefore, we have λEM∞ K ∈ EbR−c (ksub


M∞ ).
The proof is completed.
Let us summarize results of Theorems 3.18 and 3.23 in the following commutative
diagram:
 I M∞
/
Eb (ksub
M∞ ) o Eb (IkM∞ )
E
JM
❘❘ ∞
S i❘❘❘❘❘❘❘❘❘IM S
❘❘❘ ❘❘❘∞
❘❘❘∼ ❘❘❘
❘❘❘ ❘❘❘
λE
❘❘❘ ❘)
M∞ ❘❘❘
EbR−c (ksub
M∞ ) EbIR−c (IkM∞ ).
❘❘
i❘❘❘❘❘❘❘❘❘IM∞
❘❘❘ ❘❘❘
❘❘∼ S
❘❘❘ ❘❘❘❘❘
E ❘❘❘ ❘)
λM∞ ❘❘❘

EbR−c (IkM∞ ).

At the end of this subsection, let us consider an embedding functor from Db (ksub
M∞ ) to
E (ksub
b
M∞ ) and a Verdier duality functor for enhanced subanalytic sheaves.
Let us define a functor
E,sub
esub b sub b sub −1
M∞ : D (kM∞ ) → E (kM∞ ), F 7→ kM∞ ⊗ π F .

By Proposition 3.24 below, we have a commutative diagram:

esub
M∞
Db (ksub
M∞ )
/ Eb (ksub
M∞ )

I M∞ E
IM ∞
 
Db (IkM∞ ) / Eb (IkM∞ ).
esub
M∞

Proposition 3.24. For any F ∈ Db (ksub b


M∞ ) and any F ∈ DIR−c (IkN∞ ), we have

E
IM esub F ≃ eM∞ IM∞ F ,
∞ M∞

λEM∞ eM∞ F ≃ esub


M∞ λM∞ F.

Moreover, the functor esub b sub b sub


M∞ : D (kM∞ ) → E (kM∞ ) is fully faithful.

59
Proof. Let F ∈ Db (ksub
M∞ ). By Proposition 3.7 (2)(iii), (vi) and Lemma 3.20, there exist
isomorphisms in Eb (IkM∞ ):
E
IM esub F ≃ IM
∞ M∞
E
kE,sub ⊗ π −1 IM∞ F ≃ kEM∞ ⊗ π −1 IM∞ F ≃ eM∞ IM∞ F .
∞ M∞

Let F ∈ DbIR−c (IkM∞ ). By Proposition 3.7 (4)(ii), (iv) and Lemma 3.20, there exist
isomorphisms in Eb (ksub
M∞ ):

λEM∞ eM∞ F ≃ λEM∞ kEM∞ ⊗ π −1 λM∞ F ≃ kE,sub −1 sub


M∞ ⊗ π λM∞ F ≃ eM∞ λM∞ F.

Let F1 , F2 ∈ Db (ksub b sub b


M∞ ). By Proposition 3.6, the functor IM∞ : D (kM∞ ) → D (IkM∞ ) is
fully faithful and hence there exists an isomorphism

HomDb (ksub
M )
(F1 , F2 ) ≃ HomDb (IkM∞ ) (IM∞ F1 , IM∞ F2 ).

Since the functor eM∞ : Db (IkM∞ ) → Eb (IkM∞ ) is fully faithful, we have an isomorphism

HomDb (IkM∞ ) (IM∞ F1 , IM∞ F2 ) ≃ HomEb (IkM∞ ) (eM∞ IM∞ F1 , eM∞ IM∞ F2 ).

Moreover, by the first assertion, we have


E
HomEb (IkM∞ ) (eM∞ IM∞ F1 , eM∞ IM∞ F2 ) ≃ HomEb (IkM∞ ) (IM ∞ M∞
E
esub F1 , IM esub F2 ).
∞ M∞

By Theorem 3.18, the functor IM∞ : Db (ksub b


M∞ ) → E (IkM∞ ) is fully faithful, and hence

E
HomEb (IkM∞ ) (IM esub F1 , IM
∞ M∞
E
esub F2 ) ≃ HomEb (ksub
∞ M∞ M )
(esub sub
M∞ F1 , eM∞ F2 ).

Therefore, we have

HomDb (ksub
M )
(F1 , F2 ) ≃ HomEb (ksub
M )
(esub sub
M∞ F1 , eM∞ F2 ).
∞ ∞

This implies that the functor esub b sub b sub


M∞ : D (kM∞ ) → E (kM∞ ) is fully faithful.

The functor esub commutes with several functors as below.

Proposition 3.25. For a morphism f : M∞ → N∞ of real analytic bordered spaces, any


F , F1, F2 ∈ Db (ksub b sub
M∞ ) and any G ∈ D (kN∞ ), we have

+
esub sub sub
M∞ (F1 ⊗ F2 ) ≃ eM∞ F1 ⊗ eM∞ F2 ,
Ef!! esub sub
M∞ F ≃ eN∞ Rf!! F ,
Ef −1 esub sub −1
N∞ G ≃ eM∞ f G,
Ef ! esub sub !
N∞ G ≃ eM∞ f G.

60
E,sub + E,sub E,sub
Proof. By Proposition 3.17 (4)(iii) and the fact that kM ∞
⊗ kM ∞
≃ kM ∞
, we have
E,sub
esub −1
M∞ (F1 ⊗ F2 ) ≃ kM∞ ⊗ π (F1 ⊗ F2 )
 
E,sub + E,sub
≃ kM∞ ⊗ kM∞ ⊗ (π −1 F1 ⊗ π −1 F2 )
 + 
E,sub −1 E,sub −1
≃ kM∞ ⊗ π F1 ⊗ kM∞ ⊗ π F2
+
≃ esub sub
M∞ F1 ⊗ eM∞ F2 .

E,sub
The last three assertion follow from Proposition 3.15 (4)(ii) and the fact that Ef −1 kN ∞

E,sub
kM∞ .
Let i0 : M∞ → M∞ × R∞ be the morphism of real analytic bordered spaces induced
by x 7→ (x, 0). We set
E,sub
shsub ! b sub b sub
M∞ := i0 ◦ RM∞ : E (kM∞ ) → D (kM∞ )

and call it the sheafification functor for enhanced subanalytic sheaves on real analytic
bordered spaces. It has the following properties.

Proposition 3.26.
sub
(1) A pair (esub
M∞ , shM∞ ) is an adjoint pair.


sub sub
(2) For any F ∈ Db (ksub
M∞ ), one has F −→ shM∞ eM∞ F .

(3) For any K ∈ Eb (IkM∞ ), one has RJM∞ IshM∞ K ≃ shsub E


M∞ JM∞ K.

(4) For a morphism f : M∞ → N∞ of real analytic bordered spaces, any K ∈ Eb (ksub


M∞ )
and any L ∈ Eb (ksub
N∞ ), one has

Rf∗ shsub sub


M∞ K ≃ shN∞ Ef∗ K,

f ! shsub sub !
M∞ L ≃ shM∞ Ef L.

Proof. First, let us prove the assertion (3). Let K ∈ Eb (IkM∞ ). Then there exists
F ∈ Db (IkM∞ ×R∞ ) such that K ≃ QM∞ (F ). Then we have

RJM∞ IshM∞ K ≃ RJM∞ i!0 REM∞ K


≃ i!0 RJM∞ ×R∞ REM∞ K
≃ i!0 RJM∞ ×R∞ RIhom+ (ιM∞ ×R∞ (k{t≥0} ⊕ k{t≤0} ), F )
≃ i!0 RJM∞ ×R∞ RIhom+ (IM∞ ×R∞ ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), F )
≃ i!0 RIhom+,sub (ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), RJM∞ ×R∞ F )
≃ i!0 RE,sub E
M ∞ JM ∞ K
≃ shsub E
M∞ JM∞ K,

61
where in the second (resp. fourth, fifth) isomorphism we used Proposition 3.7 (3)(v)
(resp. (4)(i), Proposition 3.12 (2)).
(1) Let F ∈ Db (ksub b sub
M∞ ) and K ∈ E (kM∞ ). Then we have

sub
 E sub E

HomEb (ksub
M∞ ) eM ∞
F , K ≃ HomEb (Ik
M ∞ ) I M ∞
eM ∞
F , I M ∞
K
E

≃ HomEb (IkM∞ ) eM∞ IM∞ F , IM K

E

≃ HomDb (ksub
M ) F , RJM∞ IshM∞ IM ∞
K


≃ HomDb (ksub
M ) F , shsub E E
M ∞ J M ∞ IM ∞ K


≃ HomDb (ksub
M ) F , shsub
M∞ K

where in the first (resp. second) isomorphism we used Theorem 3.18 (resp. Proposition
3.24), in the third isomorphism we used the fact that (eM∞ , IshM∞ ) is an adjoint pair and
Proposition 3.6 (1), and the fourth isomorphism follows from (3). This implies that a pair
sub
(esub
M∞ , shM∞ ) is an adjoint pair.
sub sub
(2) Let F ∈ Db (ksub
M∞ ). By (1), there exists a canonical morphism F → shM∞ eM∞ F .
Then we have

shsub sub sub E E sub


M∞ eM∞ F ≃ shM∞ JM∞ IM∞ eM∞ F ≃ RJM∞ IshM∞ eM∞ IM∞ F ≃ F

where in the second isomorphism we used Theorem 3.18 (1), in the second isomorphism
we used (3) and Proposition 3.24, and in the last isomorphism we used the fact that
IshM∞ ◦ eM∞ ≃ id and Proposition 3.6 (1).
(4) Let K ∈ Eb (ksub b sub
M∞ ). For any F ∈ D (kN∞ ), we have
 
HomDb (ksub
N )
F , Rf∗shsub
M∞ K ≃ HomDb (ksub
M ) esub
M∞ f
−1
F, K
∞ ∞

≃ HomDb (ksub
M ) Ef −1 esub
N∞ F , K


≃ HomDb (ksub
M ) F , shsub
N∞ Ef∗ K ,

where in the first and last isomorphisms we used the fact that a pair (f −1 , Rf∗ ) is an ad-
joint pair, (1) and Proposition 3.15 (1)(ii), and in the second isomorphism we used Propo-
sition 3.25. This implies that there exists an isomorphism Rf∗ shsub sub
M∞ K ≃ shN∞ Ef∗ K.
Let L ∈ Eb (ksub b sub
N∞ ). Then there exists G ∈ D (kN∞ ×R∞ ) such that L ≃ QN∞ G. We
shall denote by fR∞ : M∞ × R∞ → N∞ × R∞ the morphism f × idR∞ of real analytic
bordered spaces. By Proposition 3.15 (2), we have
E,sub
f ! shsub ! !
M∞ L ≃ i0 fR∞ RN∞ L
≃ i!0 fR! ∞ RIhom+,sub (ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), G)
≃ i!0 RIhom+,sub (fR−1 ρ
∞ N∞ ×R∞ ∗
(k{t≥0} ⊕ k{t≤0} ), fR! ∞ G)
≃ i!0 RIhom+,sub (ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} ), fR! ∞ G)
≃ i!0 RE,sub sub !
M∞ QM∞ fR∞ G
≃ shsub !
M∞ Ef L.

62
Let us set
E,sub
ωM ∞
:= esub b sub
M∞ (ρM∞ ∗ ωM ) ∈ E (kM∞ )

where ωM ∈ Db (kM∞ ) ( ≃ Db (CM )) is the dualizing complex, see [KS90, Def. 3.1.16 (i)]
−1
for the details. Note that since ωM ≃ jM ωM̌ , we have ωM ∈ DbR−c (kM∞ ). We shall define
a functor

DE,sub b sub op
M∞ : E (kM∞ ) → Eb (ksub
M∞ ), K 7→ RIhom
+,sub E,sub
(K, ωM ∞
).

E
Lemma 3.27. There exist an isomorphism IM ω E,sub ≃ ωM
∞ M∞
E

in Eb (IkM∞ ) and an iso-
E E,sub
morphism JM ω E ≃ ωM
∞ M∞ ∞
in Eb (ksub
M∞ ).

Proof. Since ωM is R-constructible, there exists an isomorphism ιM∞ ωM ≃ IM∞ ρM∞ ∗ ωM


in Db (IkM∞ ). Hence, we have
E
ωM ∞
E
≃ eM∞ (IM∞ ρM∞ ∗ ωM ) ≃ IM ∞ M∞
E
esub (ρM∞ ∗ ωM ) ≃ IM ω E,sub ,
∞ M∞

where in the second isomorphism we used Proposition 3.24.


The second assertion follows from the first assertion and Theorem 3.18 (1).

Proposition 3.28. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces.

(1) For any K ∈ Eb (ksub b sub


M∞ ) and any L ∈ E (kN∞ ), one has

Ef ! DE,sub E,sub
N∞ L ≃ DM∞ Ef
−1
L,
Ef∗ DE,sub E,sub
M∞ K ≃ DN∞ Ef!! K,
E E,sub
JM ∞
E
DEM∞ IM ∞
K ≃ DM ∞
K.

E,sub E,sub E,sub


(2) For any K ∈ EbR−c (ksub b sub
M∞ ), we have DM∞ K ∈ ER−c (kM∞ ) and DM∞ DM∞ K ≃ K.
In particular, there exists an equivalence of categories:

DE,sub b sub op
M∞ : ER−c (kM∞ ) −→ EbR−c (ksub
M∞ ).

E,sub E,sub
Proof. (1) First, let us prove that Ef ! ωN ∞
≃ ωM ∞
. By Proposition 3.25 and the fact
that f ! ωN ≃ ωM , we have
E,sub
Ef ! ωN ∞
≃ Ef ! eE,sub sub ! sub E,sub
N∞ (ρN∞ ∗ ωN ) ≃ eM∞ (ρM∞ ∗ f ωN ) ≃ eM∞ (ρM∞ ∗ ωM ) ≃ ωM∞ .

Let K ∈ Eb (ksub b sub


M∞ ) and L ∈ E (kN∞ ). By Proposition 3.15 (2), we have

Ef ! DE,sub !
N∞ L ≃ Ef RIhom
+,sub E,sub
(L, ωN ∞
)
E,sub
≃ RIhom+,sub (Ef −1 L, Ef ! ωN ∞
)
E,sub
≃ RIhom+,sub (Ef −1 L, ωM ∞
)
≃ DE,sub
M∞ Ef
−1
L.

63
By Proposition 3.15 (1)(iii), we have

Ef∗ DE,sub
M∞ K ≃ Ef∗ RIhom
+,sub E,sub
(K, ωM ∞
)
E,sub
≃ Ef∗ RIhom+,sub (K, Ef ! ωN ∞
)
E,sub
≃ RIhom+,sub (Ef!! K, ωN ∞
)
≃ DE,sub
N∞ Ef!! K.

By Proposition 3.19 (1)(ii) and Lemma 3.27, we have


E
JM ∞
E
DEM∞ IM ∞
E
K ≃ JM ∞
RIhom+ (IM
E

E
K, ωM ∞
)
≃ RIhom+,sub (K, JM
E
ωE )
∞ M∞
E,sub
≃ RIhom+,sub (K, ωM ∞
)
≃ DE,sub
M∞ K.

(2) Let K ∈ EbR−c (ksub E b


M∞ ). By Theorem 3.23 we have IM∞ K ∈ ER−c (IkM∞ ). Since
E
DEM∞ : EbR−c (IkM∞ )op → EbR−c (IkM∞ ) (see, [DK19, Prop. 3.3.3 (ii)]), we have DEM∞ IM ∞
K∈
b E E E b sub
ER−c (IkM∞ ). Hence, by Theorem 3.23, we have JM∞ DM∞ IM∞ K ∈ ER−c (kM∞ ). Since
JME

E
DEM∞ IM ∞
K ≃ DE,sub E,sub b
M∞ K, this implies that DM∞ K ∈ ER−c (kM∞ ).
sub
E
Moreover, since IM J E DEM∞ IM
∞ M∞
E

K ≃ DEM∞ IM E

K we have

DE,sub E,sub E E E E E E
M ∞ D M ∞ K ≃ J M ∞ D M ∞ IM ∞ J M ∞ D M ∞ IM ∞ K
E
≃ JM ∞
DEM∞ DEM∞ IM
E

K
E
≃ JM IE K
∞ M∞

≃K

where the third (resp. last) isomorphism follows from [DK19, Prop. 3.3.3 (ii)])
(resp. Theorem 3.18 (1)).
Several operations preserve the R-constructability as below. Let us recall that a mor-
phism f : (M, M̌ ) → (N, Ň) of real analytic bordered spaces is called semi-proper if the
second projection M̌ × Ň → Ň is proper on the closure Γf of the graph Γf of f in M̌ × Ň .

Proposition 3.29. Let f : M∞ → N∞ be a morphism of real analytic bordered spaces


associated with a morphism fˇ: M̌ → Ň of real analytic manifolds. The functors below
are well defined:

(1) esub b b
M∞ ρM∞ ∗ : DR−c (kM∞ ) → ER−c (IkM∞ ),

(2) if f is semi-proper, Ef∗ : EbR−c (ksub b sub


M∞ ) → ER−c (kN∞ ),

(3) Ef −1 : EbR−c (ksub b sub


N∞ ) → ER−c (kM∞ ),

(4) if f is semi-proper, Ef!! : EbR−c (ksub b sub


M∞ ) → ER−c (kN∞ ),

(5) Ef ! : EbR−c (ksub b sub


N∞ ) → ER−c (kM∞ ).

64
Proof. (1) Let F ∈ DbR−c (kM∞ ) and U be an open subset of M which is subanalytic and
relatively compact in M̌ . We set F U := k{t=0} ⊗ π −1 F |U . By Propositions 3.4 (6) and
3.25, we have
 E,sub
Ei−1 sub sub −1
U∞ eM∞ ρM∞ ∗ F ≃ eU∞ ρU∞ ∗ (F |U ) ≃ kU∞ ⊗ π ρU∞ ∗ (F |U ).

+
Since kE,sub E,sub
U∞ ≃ kU∞ ⊗ ρU∞ ×R∞ ∗ k{t=0} , there exist isomorphisms
 
E,sub E,sub +
kU∞ ⊗ π ρU∞ ∗ (F |U ) ≃ kU∞ ⊗ ρU∞ ×R∞ ∗ k{t=0} ⊗ ρU∞ ×R∞ ∗ π −1 F |U
−1

+ 
≃ kE,sub −1
U∞ ⊗ ρU∞ ×R∞ ∗ k{t=0} ⊗ π F |U
+
≃ kE,sub U
U∞ ⊗ ρU∞ ×R∞ (F ),

where in the first isomorphism we used Proposition 3.15 (4)(iii). Since F is R-


constructible, we have F U ∈ DbR−c (kU∞ ). This implies that esub
M∞ ρM∞ ∗ F is R-constructible.

(2) Let K ∈ EbR−c (ksub


M∞ ). By Proposition 3.19 (3)(i), we have isomorphisms

Ef∗ K ≃ Ef∗ λEM∞ IM∞ K ≃ λEN∞ Ef∗ IM


E

K.
E
Since K is R-constructible we have IM ∞
K ∈ EbR−c (IkM∞ ) by Theorem 3.23, and hence,
E
by [DK19, Prop. 3.3.3 (iv)] we have Ef∗ IM ∞
K ∈ EbR−c (IkN∞ ). This implies that Ef∗ K ≃
E
λEN∞ Ef∗ IM ∞
K is R-constructible by Theorem 3.23.
(3) Let L ∈ EbR−c (ksub
N∞ ). By Proposition 3.19 (4)(i), we have isomorphisms

Ef −1K ≃ Ef −1 λEN∞ IN∞ L ≃ λEM∞ Ef −1 INE∞ L.


Since L is R-constructible we have INE∞ L ∈ EbR−c (IkN∞ ) by Theorem 3.23, and hence, by
[DK19, Prop. 3.3.3 (iii)] we have Ef −1INE∞ L ∈ EbR−c (IkM∞ ). This implies that Ef −1 L ≃
λEM∞ Ef −1INE∞ L is R-constructible by Theorem 3.23.
(4) Let K ∈ EbR−c (ksub
M∞ ). Then we have

Ef!! K ≃ Ef!! DE,sub E,sub E,sub E,sub


M∞ DM∞ K ≃ DN∞ Ef∗ DM∞ K

by Proposition 3.28. This implies that Ef!! K is R-constructible by the assertion (2).
(5) Let L ∈ EbR−c (ksub
N∞ ). Then we have

Ef ! K ≃ Ef ! DE,sub E,sub E,sub


N∞ DN∞ L ≃ DM∞ Ef
−1 E,sub
D N∞ L
by Proposition 3.28. This implies that Ef ! K is R-constructible by the assertion (3).
Moreover, convolution functors preserve the R-constructability as below.
Proposition 3.30. (1) The functors
+
(·) ⊗ (·) : EbR−c (ksub b sub b sub
M∞ ) × ER−c (kM∞ ) → ER−c (kM∞ ),
RIhom+,sub (·, ·) : EbR−c (ksub b sub b sub
M∞ ) × ER−c (kM∞ ) → ER−c (kM∞ )
op

are well defined.

65
(2) For any K, L ∈ EbR−c (ksub
M∞ ), one has
 
+
(i) DE,sub
M∞ K ⊗ L ≃ RIhom+,sub (K, DE,sub
M∞ L),

+
(ii) DE,sub
M∞ RIhom
+,sub
(K, L) ≃ K ⊗ DE,sub L,
M∞ 
+,sub +,sub E,sub E,sub
(iii) RIhom (K, L) ≃ RIhom DM∞ L, DM∞ K ,
 
(iv) RIhomE,sub (K, L) ≃ RIhomE,sub DE,sub M∞ L, D E,sub
M∞ K ,
 
(v) RHomE,sub (K, L) ≃ RHomE,sub DE,sub E,sub
M∞ L, DM∞ K .

Proof. (1) Let K, L ∈ EbR−c (ksub


M∞ ) and U be an open subset of M which is subanalytic
and relatively compact in M̌ . Then there exist F , G ∈ DbR−c (kU∞ ×R∞ ) such that
E,sub + E,sub +
Ei−1 sub −1 sub
U∞ K ≃ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ F , EiU∞ L ≃ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ G.

Hence ,we have


 
+ +
−1
EiU∞ K ⊗ L ≃ Ei−1U∞ K ⊗ Ei−1
U∞ L
   
E,sub + sub
+ E,sub + sub
≃ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ F ⊗ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ G
 
E,sub + sub
+
≃ kU∞ ⊗ QU∞ ρU∞ ×R∞ ∗ F ⊗ ρU∞ ×R∞ ∗ G

where in the first isomorphism we used Proposition 3.15 (2) and in the last isomorphism
+
we used kE,sub
U∞ ⊗ kU∞
E,sub
≃ kE,sub b
U∞ . Since F , G ∈ DR−c (kU∞ ×R∞ ), there exists an isomor-
+ 
phism ρU∞ ×R∞ ∗ F ⊗ ρU∞ ×R∞ ∗ G ≃ ρU∞ ×R∞ ∗ µ!! (p−1 −1 −1 −1
1 F ⊗ p2 G) and µ!! (p1 F ⊗ p2 G) ∈
+
DbR−c (kU∞ ×R∞ ). Therefore, K ⊗ L is R-constructible.
Moreover, by using the assertion (2)(i), (iii), there exist isomorphisms
 
+,sub +,sub E,sub E,sub
RIhom (K, L) ≃ RIhom DM∞ L, DM∞ K
 
E,sub E,sub +
≃ DM ∞ DM ∞ L ⊗ K
 
E,sub + E,sub
≃ DM ∞ K ⊗ DM ∞ L ,

and hence by Proposition 3.28 (2) and the first assertion of (1), RIhom+,sub (K, L) is
R-constructible.
(2)(i) By Proposition 3.15 (1)(i), we have
   
E,sub + + E,sub
+,sub
DM∞ K ⊗ L ≃ RIhom K ⊗ L, ωM∞
 
+,sub +,sub E,sub
≃ RIhom K, RIhom (L, ωM∞ )
 
≃ RIhom+,sub K, DE,sub M∞ L .

66
E,sub E,sub
(ii) Since L is R-constructible, we have an isomorphism L ≃ DM ∞
DM∞ L by Proposition
3.28 (2). Hence, by using the assertion (2)(i), we have

DE,sub
M∞ RIhom
+,sub
(K, L) ≃ DE,sub
M∞ RIhom
+,sub
(K, DME,sub E,sub
D L)
 ∞
 M∞
+
≃ DE,sub
M∞ D E,sub
M∞ K ⊗ E,sub
DM ∞
L
+
≃ K ⊗ DE,sub
M∞ L,

where in the last isomorphism we used Proposition 3.28 (2) and the assertion (1).
(iii) By Proposition 3.15 (1)(i), we have
   
E,sub + E,sub + E,sub E,sub
+,sub
DM∞ K ⊗ DM∞ L ≃ RIhom K ⊗ DM∞ L, ωM∞
 
≃ RIhom+,sub DE,sub
M∞ L, RIhom +,sub
(K, ω E,sub
M∞ )
 
≃ RIhom+,sub DE,sub E,sub
M∞ L, DM∞ K .

 
+
Since RIhom +,sub
(K, L) ≃ DE,sub
M∞
E,sub
K ⊗ DM∞ L , we have
 
E,sub E,sub
RIhom+,sub (K, L) ≃ RIhom+,sub DM ∞
L, D M∞ K .

+
(iv) First, let us prove kE,sub E,sub E E
M∞ ⊗ K ≃ K. Since kM∞ ≃ λM∞ kM∞ and K is R-constructible,
we have
 
E,sub + + +
E E
kM∞ ⊗ K ≃ λM∞ kM∞ ⊗ λM∞ IM∞ K ≃ λM∞ kM∞ ⊗ IM∞ K ≃ λEM∞ IM
E E E E E E

K ≃ K,

where in the second and last isomorphisms (resp. second isomorphism) we used Theorem
3.18 (1) (resp. Proposition 3.19 (4)(iii)), and in the third isomorphism we used the fact
+
that kEM∞ ⊗ K ′ ≃ K ′ for any K ′ ∈ EbR−c (IkM∞ ).
Hence, there exist isomorphisms
+
RIhomE,sub (K, L) ≃ RIhomE,sub (kE,sub ⊗ K, L)
 M∞ 
≃ RIhomE,sub kE,sub
M∞ , RIhom +,sub
(K, L)
 
≃ RIhomE,sub kE,sub
M∞ , RIhom +,sub
(D E,sub
M∞ L, D E,sub
M∞ K)
 
+
≃ RIhomE,sub kE,sub E,sub
M∞ ⊗ DM∞ L, DM∞ K
E,sub

 
≃ RIhomE,sub DE,sub
M∞ L, D E,sub
M∞ K ,

where in the second and fourth isomorphisms we used Proposition 3.15 (1)(i) and in the
third isomorphism we used the assertion (iii).
The assertion (v) follows from the assertion (iv).

67
3.5 Irregular Riemann–Hilbert Correspondence and Enhanced
Subanalytic Sheaves
In this subsection, we will explain a relation between [DK16, Thm. 9.5.3] and [Kas16,
Thm. 6.2]. Theorem 3.40 and 3.41 are main results of this paper.

3.5.1 Main Results of [DK16] and [Kas16]


Let X be a complex manifold and denote by XR the underlying real analytic manifold
of X. We denote by OX and DX the sheaves of holomorphic functions and holomorphic
differential operators on X, respectively. Let Db (DX ) be the bounded derived category
of left DX -modules. Moreover we denote by Dbcoh (DX ), Dbhol (DX ) and Dbrh (DX ) the full
triangulated subcategories of Db (DX ) consisting of objects with coherent, holonomic and
regular holonomic cohomologies, respectively. For a morphism f : X → Y of complex
D ∼
manifolds, denote by ⊗, Df∗ , Df ∗ , DX : Dbcoh (DX )op −→ Dbcoh (DX ) the standard opera-
tions for D-modules. The classical solution functor is defined by
SolX : Db (DX )op → Db (CX ), M 7−→ RHomDX (M, OX ).

Let M be a real analytic manifold of dimension n. We denote by CM the sheaf of

complex functions of class C on M and by DbM the sheaf of Schwartz’s distributions M.
Definition 3.31 ([KS01, Def. 7.2.3]). Let U be an open subset of M.

(1) One say that f ∈ CM (U) has polynomial growth at p ∈ M if for a local coordinate
system (x1 , . . . , xn ) around p, there exist a sufficiently small compact neighborhood
K of p and a positive integer N such that
sup dist(x, K \ U)N |f (x)| < +∞.
x∈K∩U


A function f ∈ CM (U) is said to be tempered at p ∈ M if all its derivatives have
polynomial growth at p, and is said to be tempered if tempered at any point of M.
∞,t ∞
Let us denote by CM (U) the subset of CM (U) consisting of functions which are
tempered.
(2) Let us defined DbtM (U) by the image of the restriction map
Γ(M; DbM ) → Γ(U; DbM ).
∞,t
Note that subanalytic presheaves U 7→ CM (U) and U 7→ DbtM (U) are subanalytic
sheaves, see [KS01, §7.2], also [Pre08, §3.3].
Definition 3.32 ([KS01, §7.3], [DK16, §5.2] and also [Pre08, §3.3]). Let us denote b X c
t
the complex conjugate manifold of X. An object OX ∈ Db (Csub
X ) is defined by
t ∞,t
OX := RIhomsub sub t
ρX !DX c (ρX c ! OX c , CXR ) ≃ RIhomρX !DX c (ρX c ! OX c , DbXR )

and is call the subanalytic sheaf of tempered holomorphic functions on X.


Moreover, the tempered solution functor is defined by
SoltX : Db (DX )op → Db (ICX ), M 7−→ RIhomβX DX (βX M, IM OX
t
).

68
t
Note that an ind-sheaf IX OX is nothing but the ind-sheaf of tempered holomorphic
t
functions on X which is denoted by OX in [KS01, §7.3]. Note also that there exist
isomorphisms ρX OX ≃ αX IX OX ≃ OX and hence we have αX SoltX (M) ≃ SolX (M) for
−1 t t

any M ∈ Dbcoh (DX ).


Definition 3.33 ([DK16, Def. 8.1.1] and [KS16b, Def. 7.2.1]). Let e k M : M × R∞ → M ×
1
P R be the natural morphism of real analytic bordered spaces and t be a coordinate of
R, where P1 R is the real projective line. An object DbtM ×R∞ ∈ Db (ICM ×R∞ ) is defined by

DbtM ×R∞ := e!
kM IM ×P1 R DbtM ×P1 R ≃ IM ×R∞ e!
kM DbtM ×P1 R

and DbTM ∈ Db (ICM ×R∞ ) is defined by the complex, concentrated in −1 and 0


∂ −1
DbtM ×R∞ −−−
t
−→ DbtM ×R∞ .

Moreover, we set DbEM := QM (DbTM ) ∈ Eb (ICM ) and call it the enhanced ind-sheaf of
tempered distributions.
Note that we have Hi (DbTM ) = 0 for any i 6= −1 and hence there exists an isomorphism
DbTM ≃ Ker(∂t − 1)[1] in Db (ICM ×R∞ ).
Definition 3.34 ([DK16, Def. 8.2.1] and [KS16b, Def. 7.2.3]). Let ei : X × R∞ → X × P1 C
be the natural morphism of bordered spaces and τ ∈ C ⊂ P1 C the affine coordinate such
that t = τ |R , where P1 C is the complex projective line. An object OX E
∈ Eb (ICX ) is
defined by
E −1 T
OX := QX RIhomπ−1c βX c DX c (πX c βX c OX c , DbX )
R
X

≃ QXei! RIhomp−1 βP1 C DP1 C (p−1 βP EC|P


τ t
1 C , IX×P1 C OX×P1 C )[2],

τ
where EC|P is the meromorphic connection associated to d + dτ , p : X × P1 C → P1 C is the
projection and πX c : X c × R∞ → X c is the morphisms of bordered spaces associated with
the projection X c × R → X c .
It is called the enhanced ind-sheaf of tempered holomorphic functions on X.
Moreover, the enhanced solution functor is defined by

SolEX : Db (DX )op → Eb (ICX ), M 7−→ RIhomπ−1 βX DX (π −1 βX M, OX


E
),

where π : X × R∞ → X is the morphism of bordered spaces associated with the first


projection X × R → X.
E
Note that OX is isomorphic to the enhanced ind-sheaf induced by the Dolbeault com-
plex with coefficients in DbTXR [−1]:
∂ ∂ ∂
DbTXR [−1] −−→ Ω1X c ⊗OX c DbTXR [−1] −−→ · · · −−→ ΩdXXc ⊗OX c DbTXR [−1],

where ΩpX c is the sheaf of p-differential forms with coefficients in OX c and dX is the
dimension of X.
Note that IshX OX E
≃ IX OXt
and hence there exists an isomorphism IshX SolEX (M) ≃
t b
SolX (M) for any M ∈ Dcoh (DX ).
Let us recall the main results of [DK16].

69
Theorem 3.35 ([DK16, Thm. 9.5.38 and 9.6.1]). The enhanced solution functor induces
an embedding functor
SolEX : Dbhol (DX )op ֒→ EbR−c (ICX ).
Moreover for any M ∈ Dbhol (DX ) there exists an isomorphism

M −→ RHEX SolEX (M),

where RHEX (K) := RHomE (K, OX


E
)

Let us recall the main results of [Kas16].


T,sub
Definition 3.36 ([Kas16, §§5.3, 5.4]). An object DbM ∈ Db (Csub
M ×R∞ ) is defined by the
complex, concentrated in −1 and 0

e ∂t −1 e!
k ! DbtM ×P1 R −−− −→ k DbtM ×P1 R
T,sub
and OX ∈ Db (Csub
X×R∞ ) is defined by the Dolbeault complex with coefficients in
T,sub
DbXR [−1]:

∂ ∂ ∂
DbT,sub −→ Ω1X c ⊗OX c DbT,sub
XR [−1] − −→ · · · −−→ ΩdXXc ⊗OX c DbT,sub
XR [−1] − XR [−1].

Moreover, we set

SolT,sub
X : Db (DX )op → Db (Csub sub −1 T,sub
X×R∞ ), M 7−→ RIhomπ −1 ρX! DX (π ρX! M, OX ).

Note that we have Hi (DbT,sub


M ) = 0 for any i 6= −1 and hence there exists an isomor-
phism DbT,sub
M ≃ Ker(∂ t − 1)[1] in Db (Csub T
M ×R∞ ). Remark that the notion Db in [Kas16,
§5.2] is equal to DbT,sub [−1] in our notion. Note also that there exists an isomorphism in
Db (ksub
X×R∞ )  
T,sub −1 T,sub
OX ≃ RIhomsub
π −1c ρX c ! DX c
πX c ρX c ! OX c , DbX
R
[−1] .
X

Theorem 3.37 ([Kas16, Thms. 6.2 and 6.39 ]). There exists an embedding functor

SolT,sub
X : Dbhol (DX )op ֒→ Db (Csub
X×R∞ ).

Moreover for any M ∈ Dbhol (DX ) there exists an isomorphism in Db (DX ):


 

M −→ RHomE,sub SolT,sub X (M), O T,sub
X .
8
In [DK16], although Theorem 9.5.3 was stated by using the enhanced de Rham functor, we can obtain
a similar statement by using the enhanced solution functor.
9
In [Kas16], although Theorem 6.3 was stated by using the enhanced de Rham functor, we can obtain
a similar statement by using the enhanced solution functor.

70
3.5.2 Relation between [DK16, Thm. 9.5.3] and [Kas16, Thm. 6.2]
Let us explain a relation between [DK16, Thm. 9.5.3] and [Kas16, Thm. 6.2].

Definition 3.38. Let us define


 
E,sub T,sub
OX := Qsub
X OX [1] ∈ Eb (Csub
X )

and set

SolE,sub
X : Db (DX )op → Eb (Csub sub −1 E,sub
X ), M 7−→ RIhomπ −1 ρX! DX (π ρX! M, OX ).

By the definition, it is clear that


 
SolE,sub
X (M) ≃ QX Sol T,sub
X (M) [1]

for any M ∈ Db (DX ).


T,sub
Lemma 3.39. (1) There exists an isomorphism DbTM ≃ IM ×R∞ DbM in Db (ICM ×R∞ ).
E,sub
(2) There exists an isomorphism OX ≃ JXE OX
E
in Eb (Csub
X ).

(3) For any M ∈ Db (DX ), there exists an isomorphism in Eb (Csub


X ) :

SolE,sub
X (M) ≃ JXE SolEX (M).

(4) For any M ∈ Db (DX ), there exists an isomorphism in Db (Csub


X×R∞ ) :

SolT,sub
X (M)[1] ≃ RE,sub
X
E,sub
SolX (M).

Proof. (1) Since the functor IM ×R∞ is exact, we have isomorphisms



DbTM ≃ Ker ∂t − 1 : DbtM ×R∞ → DbtM ×R∞
 
e! t e ! t
≃ Ker ∂t − 1 : IM ×R∞ kM DbM ×P1 R → IM ×R∞ kM DbM ×P1 R
 
≃ IM ×R∞ Ker ∂t − 1 : ekM!
DbtM ×P1 R → e!
kM DbtM ×P1 R
≃ IM ×R∞ DbT,sub
M ,
T,sub
see Definition 3.33 for the details of DbTM , Definition 3.36 for the details of DbM .
−1 −1
(2) Since βX c ≃ IX c ◦ρX c ! and πX c ◦IX c ≃ IX c ×R∞ ◦πX c (see §2.5 for the first isomorphism,

Proposition 3.7 (2)(iii) for the second isomorphism), we have isomorphisms


−1

E
JXE OX ≃ QX RJX×R∞ RIhomπ−1c βX c ! DX c πX c βX c OX c , DbX
T
R
X
−1 T

≃ QX RJX×R∞ RIhomπ−1c βX c ! DX c πX c IX c ρX c ! OX c , DbX
R
X
−1 T

≃ QX RJX×R∞ RIhomπ−1c βX c ! DX c IX c ×R∞ πX c ρX c ! OX c , DbX
R
.
X

71
Moreover, there exist isomorphisms
−1 T

QX RJX×R∞ RIhomπ−1c βX c ! DX c IX c ×R∞ πX c ρX c ! OX c , DbX
R
sub
X
−1 T

≃ QX RIhomπ−1c ρ c D c πX c ρX c ! OX c , RJXR ×R∞ DbXR
X ! X
X
 
sub −1 T,sub
≃ QX RIhomπ−1c ρ c D c πX c ρX ! OX , DbXR
c c
X X ! X
E,sub
≃ OX

where in the second isomorphism we used the assertion (1) and Proposition 3.6 (1).
(3) Let M ∈ Db (DX ). By the fact that βX ≃ IX ◦ ρX! and the assertion (2), there exist
isomorphisms

JXE SolEX (M) ≃ JXE RIhomπ−1 βX DX (π −1 βX M, OX


E
)
≃ JXE RIhomπ−1 βX DX (π −1 IX ρX! M, OX
E
)
≃ JXE RIhomπ−1 βX DX (IX×R∞ π −1 ρX! M, OX
E
)
≃ RIhomsub −1 E E
π −1 βX DX (π ρX! M, JX OX )
E,sub
≃ RIhomsub −1
π −1 βX DX (π ρX! M, OX )
≃ SolEX (M).

(4) First let us prove that


  
T,sub T,sub
RIhom+,sub ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbX R
≃ DbX R
.

By the assertion (1) and Proposition 3.19 (1)(ii),


  
T,sub
RIhom+,sub ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbX R

+,sub
 
≃ RIhom ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , RJX×R∞ DbTXR
 
≃ RJX×R∞ RIhom+ IX×R∞ ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbTXR
 
≃ RJX×R∞ RIhom+ ιM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbTXR .
 
Moreover, by using the fact that RIhom+ ιM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbTXR ≃ DbTXR ,
see e.g. [DK16, Prop. 8.1.3], we have
  T,sub
RJX×R∞ RIhom+ ιM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbTXR ≃ RJX×R∞ DbTXR ≃ DbX R
.
  
+,sub T,sub T,sub
Hence, we proved RIhom ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbXR ≃ DbX R
.
Next, we shall prove that
T,sub T,sub
RIhom+,sub (ρM∞ ×R∞ ∗ (k{t≥0} ⊕ k{t≤0} , OX )) ≃ OX .

72
 
T,sub −1 T,sub
By the fact that OX ≃ RIhomsub
π −1 ρ c D c
πX c ρX c ! OX c , Db
XR [−1] , we have
Xc X ! X

  
T,sub
RIhom+,sub ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , OX
   
+,sub sub −1 T,sub
≃ RIhom ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , RIhomπ−1c ρX c ! DX c πX c ρX c ! OX c , DbXR [−1]
X
   
T,sub
≃ RIhomsub
πX−1
ρ c
c X ! X D c
πX
−1
c ρX c ! OX c , RIhom
+,sub
ρ M ∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , DbXR [−1]
 
−1 T,sub
≃ RIhomsub
π −1c ρX c ! DX c
πX c ρX c ! OX c , DbX
R
[−1]
X
T,sub
≃ OX .

By the definition, it is clear that

RE,sub
X SolE,sub
X (M)
  
+,sub sub −1 T,sub
≃ RIhom ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , RIhomπ−1 ρX! DX (π ρX! M, OX [1]
  
T,sub
≃ RIhomsub −1
π −1 ρX! DX (π ρX! M, RIhom
+,sub
ρM∞ ×R∞ ∗ k{t≥0} ⊕ k{t≤0} , OX [1]
 
T,sub
≃ RIhomsub −1
π −1 ρX! DX π ρX! M, OX [1]
≃ SolT,sub
X (M)[1].

Theorem 3.40. The functor SolE,sub


X induces an embedding functor

SolE,sub
X : Dbhol (DX )op ֒→ EbR−c (Csub
X ).

Moreover for any M ∈ Dbhol (DX ) there exists an isomorphism in Db (DX ) :



M −→ RHE,sub
X SolE,sub
X (M),

where RHE,sub
X
E,sub
(K) := RHomE,sub (K, OX ).
Proof. First, let us prove that SolE,sub
X (M) ∈ EbR−c (Csub b
X ) for any M ∈ Dhol (DX ). Let
E
M ∈ Dbhol (DX ). By Theorem 3.35, we have SolX (M) ∈ EbR−c (ICX ) and hence by Theorem
E,sub
3.23 we have JXE SolEX (M) ∈ EbR−c (Csub
X ). This implies that SolX (M) ∈ EbR−c (Csub
X ) by
E,sub b op b sub
Lemma 3.39 (3). Hence, a functor SolX : Dhol (DX ) → ER−c (CX ) is well defined.
For any M, N ∈ Dbhol (DX ), there exist isomorphisms

HomDb (DX ) (M, N ) ≃ HomEbR−c (ICX ) SolEX (N ), SolEX (M)
hol
E E E E

≃ HomEbR−c (Csub
X ) J X Sol X (N ), J X Sol X (M)
 
E,sub E,sub
≃ HomEbR−c (Csub
X )
SolX (N ), SolX (M) ,

where in the first (resp. second, third) isomorphism we used Theorem 3.35 (resp. Theorem
E,sub
3.23, Lemma 3.39 (3)). This implies that the functor SolX : Dbhol (DX )op → EbR−c (Csub
X )
is fully faithful.

73
Let M ∈ Dbhol (DX ). By using the adjoontness, there exist isomorphisms
 
HomDb (DX ) M, RHE,sub X Sol E,sub
X (M)
hol
 
≃ HomDb (DX ) M, RHomE,sub (SolE,sub X (M), O E,sub
X )
hol
 
−1 sub E,sub E,sub
≃ HomDb (DX ) M, ρX Rπ∗ RIhom (SolX (M), OX )
hol
 
≃ HomDb (DX ) π −1 ρX! M, RIhomsub (SolE,sub X (M), O E,sub
X )
hol
 
E,sub sub −1 E,sub
≃ HomEbR−c (Csub
X )
Sol X (M), RIhom π −1 ρX! DX (π ρ X! M, O X )
 
≃ HomEbR−c (Csub
X )
SolE,sub
X (M), SolE,sub
X (M) .

Hence, there exists a canonical morphism

M → RHE,sub
X SolE,sub
X (M)

which is given by the identity map idM of M. Let us prove that it is isomorphism.
By Lemma 3.39 (2), we have isomorphisms

RHE,sub
X SolE,sub
X (M) ≃ RHomE,sub (SolE,sub
X
E,sub
(M), OX E,sub
) ≃ RHomE,sub (SolX (M), JXE OX
E
)

and by Proposition 3.19 (1)(iii) we have

RHomE,sub (SolE,sub
X (M), JXE OX
E E,sub
) ≃ ρ−1 RIhomE,sub (SolX (M), JXE OX
E
)
E,sub
≃ ρ−1 RJX RIhomE (IX
E
SolX E
(M), OX ).

By Proposition 3.7 (3)(ii) and Lemma 3.39 (3), there exists an isomorphism in Db (CX )

ρ−1 RJX RIhomE (IX


E
SolE,sub
X
E
(M), OX ) ≃ αX RIhomE (IX
E E,sub
SolX E
(M), OX )
E,sub
≃ RHomE (IX
E
SolX E
(M), OX )
≃ RHomE (IX JX SolEX (M), OX
E E E
).

Since M ∈ Dbhol (DX ), SolEX (M) is R-constructible by the first assertion and hence there
exists an isomorphism IXE E
JX SolEX (M) ≃ SolEX (M) by Theorem 3.23. By Theorem 3.35,
we have
RHomE (SolEX (M), OX E
) ≃ RHEX SolEX (M) ≃ M.

Therefore, there exists an isomorphism M −→ RHE,sub
X
E,sub
SolX (M).

Theorem 3.41. (1) For any M ∈ Dbhol (DX ), there exists an isomorphism in Eb (ICX ) :
E,sub
SolEX (M) ≃ IX
E
SolX (M).

(2) For any M ∈ Db (DX ), there exists an isomorphism in Db (Csub


X×R∞ ) :

sub
SolT,sub
X (M)[1] ≃ REX JXE, SolEX (M).

74
Moreover, there exists a commutative diagram:

 SolT,sub
Dbhol (DX )  Db (Csub
X
X×R∞ )
/
 ❘v
❘❘❘
❘❘❘ O
❘❘❘ RE,sub
X
E
JX
SolE
X
❘❘❘
( ?
EbR−c (ICX ).

Proof. (1) Let M ∈ Dbhol (DX ). Since SolEX (M) is R-constructible, there exists an isomor-
phism IX JX SolEX (M) ≃ SolEX (M) by Theorem 3.23 and Lemma 3.39 (3).
E E

(2) This follows from Lemma 3.39 (3), (4).


Let us summarize results of Theorems 3.18, 3.23, 3.40 and 3.41 in the following com-
mutative diagram:

Db (Csub
X×R∞ )
❣❣❣3 O
T,sub ❣❣❣❣❣❣❣
SolX ❣❣❣❣❣❣ ❣❣
❣❣ ❣❣❣❣❣ RE,sub
X
❣ ❣❣ ❣❣
❣❣❣❣ ?
%  ❣❣❣❣❣❣
Dbhol (DX ) t
 ❖ SolE,sub
/ Eb (Csub )
R−c X ⊂ E (Csub
b
X ) g❖
❖❖❖ X O O  ❖t ❖❖❖❖❖ E
❖❖❖ ❖❖❖ ❖❖JX
❖❖❖ E ≀ JE E ≀ JE
❖❖❖ ❖❖❖
❖❖❖ IX X
IX X
❖❖❖ ❖❖❖
SolE ❖❖❖ ❖❖❖
IE
X
❖❖❖ ❖
❖'   X ❖❖❖
'
EbR−c (ICX ) ⊂ EbIR−c (ICX ) ⊂ Eb (ICX ).

Remark 3.42. One can consider C-constructability for enhanced subanalytic sheaves as
similar to [Ito20, Def. 3.19]. We shall skip the explanation of it.

References
[BBD] Alexander A. Beilinson, Joseph Bernstein and Pierre Deligne, Faisceaux pervers,
in: Analysis and topology on singular spaces, Luminy, 1981, Astérisque 100, Soc.
Math. France, Paris, 5–171, 1982.

[Be] Joseph Bernstein, Algebraic Theory of D-Modules, unpublished notes.

[BM88] Edward Bierstone, Pierred Milman, Semianalytic and subanalytic sets, Publica-
tions mathèmatiques de l’ I.H.È.S., tome 67,5–42, 1988

[Bjö93] Jan-Erik Björk, Analytic D-modules and applications, Mathematics and its Ap-
plications, 247, Kluwer Academic Publishers Group, 1993.

[Bor] Armand Borel et al., Algebraic D-Modules, Perspectives in Mathematics, 2, Aca-


demic Press, 1987.

75
[Bry86] Jean-Luc Brylinski, Transformations canoniques, dualité projective, théorie de
Lefschetz, transformations de Fourier et sommes trigonométriques, Astérisque, 140–
141, 3–134, 1986, Géométrie et analyse microlocales.

[DK16] Andrea D’Agnolo and Masaki Kashiwara, Riemann–Hilbert correspondence for


holonomic D-modules, Publ. Math. Inst. Hautes Études Sci., 123(1), 2016, 69–197.

[DK17] Andrea D’Agnolo and Masaki Kashiwara, A microlocal approach to the enhanced
Fourier-Sato transform in dimension one, Adv. in Math., 339, 1–59, 2019.

[DK19] Andrea D’Agnolo and Masaki Kashiwara, Enhanced perversities, J. Reine Angew.
Math. (Crelle’s Journal), 751, 185–241, 2019.

[DK21] Andrea D’Agnolo and Masaki Kashiwara, On a topological counterpart of regu-


larization for holonomic D-modules, Journal de l’École polytechnique Mathématiques,
8, 27–55, 2021

[Gab68] Andrei M. Gabrielov, Projections of semi-analytic sets, Functional Anal. Appl.,


2, 282–291, 1968.

[Har75] Robert M. Hardt, Stratification of real analytic mappings and images, Invent.
Math., 28, 193–208, 1975.

[Har77] Robert M. Hardt, Triangulation of subanalytic sets and proper light subanalytic
maps, Invent. Math., 38, 207–217, 1977.

[Hiro73a] Heisuke Hironaka, Subanalytic sets, Number Theory, Algebraic Geometry and
Commutative Algebra, Tokyo, Kinokuniya, 1973, 453–493.

[Hiro73b] Heisuke Hironaka, Introduction to real-analytic sets and real-analytic maps,


Istituto Matematico “L. Tonelli”, Pisa, 1973.

[HTT08] Ryoshi Hotta, Kiyoshi Takeuchi, and Toshiyuki Tanisaki, D-modules, perverse
sheaves, and representation theory, Progress in Mathematics, 236, Birkhäuser, 2008.

[Ito20] Yohei Ito, C-Conctructible Enhanced Ind-Sheaves, Tsukuba journal of Mathemat-


ics, 44(1), 155–201, 2020.

[Ito21] Yohei Ito, Note on Algebraic Irregular Riemann–Hilbert Correspondence, Rend.


Sem. Mat. Univ. Padova., 2021, in press, arXiv:2004.13518, 52 pages..

[Kas84] Masaki Kashiwara, The Riemann–Hilbert problem for holonomic systems, Publ.
Res. Inst. Math. Sci., 20(2), 319–365, 1984.

[Kas16] Masaki Kashiwara, Riemann–Hilbert correspondence for irregular holonomic D-


modules, Japan J. Math., 11, 13–149, 2016.

[KS90] Masaki Kashiwara and Pierre Schapira, Sheaves on manifolds, Grundlehren der
Mathematischen Wissenschaften, 292, Springer-Verlag, 1990.

[KS01] Masaki Kashiwara and Pierre Schapira, Ind-sheaves, Astérisque, 271, 2001.

76
[KS06] Masaki Kashiwara and Pierre Schapira, Categories and Sheaves, Grundlehren der
Mathematischen Wissenschaften, 332 , Springer-Verlag, 2006.

[KS16a] Masaki Kashiwara and Pierre Schapira, Irregular holonomic kernels and Laplace
transform, Selecta Math., 22(1), 55–109, 2016.

[KS16b] Masaki Kashiwara and Pierre Schapira, Regular and irregular holonomic D-
modules, London Mathematical Society Lecture Note Series, 433, Cambridge Univer-
sity Press, 2016.

[Ked10] Kiran S. Kedlaya, Good formal structures for flat meromorphic connections, I:
surfaces, Duke Math. J., 154(2), 343–418, 2010.

[Ked11] Kiran S. Kedlaya, Good formal structures for flat meromorphic connections, II:
excellent schemes, J. Amer. Math. Soc., 24(1), 183–229, 2011.

[Kuwa18] Tatsuki Kuwagaki, Irregular perverse sheaves, arXiv:1808.02760, preprint.

[Pre08] Luca Prelli, Sheaves on subanalytic sites, Rend. Semin. Mat. Univ. Padova, 120,
167–216, 2008.

[Pre13] Luca Prelli, Microlocalization of subanalytic sheaves, Mém. Soc. Math. Fr., 135,
2013.

[SGA4] S-G-A 4, Sém. Géom. Algébrique (1963-64) by M. Artin, A. Grothendieck and


J-L. Verdier, Théorie des topos et cohomologie étale des schémas, Lecture Notes in
Math., 269, 270, 305, Springer-Verlag, 1972/73.

[Ler50] Jean Leray, L’anneau spectral et l’anneau filtré d’homologie d’un espace locale-
ment compact et d’une application continue, J. Math. Pures et Appl., 29, 1–139,
1950.

[Loj59] Stanislaw Lojasiewicz, Sur Ie probleme de la division, Studio Math., 8, 87–136,


1959.

[Loj64a] Stanislaw Lojasiewicz, Triangulation of semi-analytic sets, Ann. Scuola Norm.


Sup. Pisa (3), 18, 449-474, 1964.

[Loj64b] Stanislaw Lojasiewicz, Ensembles semi-analytiques, Inst. Hautes Études Sci.,


Bures-sur-Yvette, 1964.

[Moc09] Takuro Mochizuki, Good formal structure for meromorphic flat connections on
smooth projective surfaces, In Algebraic analysis and around, Adv. Stud. Pure Math.,
54, 223–253, 2009.

[Moc11] Takuro Mochizuki, Wild harmonic bundles and wild pure twistor D-modules,
Astérisque, 340, 2011.

[Moc16] Takuro Mochizuki, Curve test for enhanced ind-sheaves and holonomic D-
modules, arXiv:1610.08572, preprint.

77

You might also like