A correction to the uniqueness of a partial
perfect locality over a Frobenius P-category
Lluis Puig
CNRS, Institut de Mathématiques de Jussieu, [email protected]
6 Av Bizet, 94340 Joinville-le-Pont, France
Abstract: Let p be a prime, P a finite p-group and ℱ a Frobenius P-category. In Existence, uniqueness and functoriality of the perfect locality over a Frobenius
P-category, Algebra Colloquium, 23(2016) 541-622, we also claimed the uniqueness of the partial
perfect locality ℒX over any up-closed set X of ℱ-selfcentralizing subgroups of P, but recently Bob Oliver exhibit some counter-examples, demanding
some revision of our arguments. In this Note we show that, up to replacing the perfect localities by the extendable perfect localities over any up-closed set X of ℱ-selfcentralizing subgroups of P, our arguments are correct, still proving the existence and the uniqueness of the perfect ℱsc-locality, since it is extendable.
We take advantage to simplify some of our arguments.
**1. Introduction **
1.1. Let p be a prime, P a finite p-group, F a Frobenius P-category [2]
and TP the category where the objects are the subgroups of P, the morphisms are defined by the
P-transporters and the composition is defined by the product in P.
Recall that, according to [3, 17.3], an F-locality L is a finite category where the objects are all the subgroups of P, endowed with two functors
[TABLE]
which are the identity on the set of objects, π being full, and such that the composition
π∘τ is induced by the conjugation in P; we say that L is divisible whenever it fulfills the following condition
1.1.2 If Q, R and T are subgroups of P, for any L-morphisms
x:R→Q and y:T→Q such that the image of πQ,T(y) is contained in
the image of πQ,R(x), there is a
unique L-morphism z:T→R such that x⋅z=y.
1.2. Then, it follows from [3, Proposition 18.4 and Theorem 18.6] that a
perfect F-locality, introduced in [3, 17.13], is a divisible F-locality P such that, for any subgroup Q of P fully normalized in F [3, 2.6] the finite group P(Q) endowed with the group homomorphims
[TABLE]
is the F-localizer of Q, introduced in [3, 18.5]. Actually, as we show in [3, Theorem 20.24] and, more carefully, in [5, Theorem 7.2], P is uniquely determined by
the full subcategory Psc over the set of F-selfcentralizing subgroups
of P, introduced in [3, 4.8].
1.3. More generally, in order to apply inductive arguments, for any nonempty set X of
F-selfcentralizing subgroups of P which contains any
subgroup of P admitting an F-morphism from some subgroup in X, we consider the full subcategory
FX of F over X as the set of objects and, replacing TP by its full
subcategory TPX over X, we may introduce the FX-localities
as the finite categories LX where the objects are the groups in X, endowed with two functors
[TABLE]
which are the identity on the set of objects, πX being full, and such that the composition πX∘τX is induced by the conjugation in P.
In particular, in [5, 2.8] we consider a perfect FX-locality PX
and in [5, 6.1] we claimed its existence and uniqueness.
1.4. But, recently, Bob Oliver exhibit some counter-examples to this uniqueness [1]; of course, these counter-examples demand a revision of our arguments
in [5]. Our purpose in this Note is to show that, up to restricting the perfect
FX-localities we consider, our arguments become correct and the
uniqueness of these restricted perfect FX-localities, called extendable, is true; naturally, our extendable perfect Fsc-localities
include Psc above. Moreover, we take advantage of this revision to simplify some arguments in [5]. Notations and terminology are the same as in [5] and the main
references come from [3].
2. Extendable perfect FX-localities
2.1. With the notation above, let us consider a perfect FX-locality
PX; that is to say, PX is a divisible FX-locality
such that, for any group Q in X fully normalized in F [3, 2.6], the finite group
PX(Q) endowed with the group homomorphims
[TABLE]
is the F-localizer of Q, introduced in [3, 18.5]; in particular, πQX
is surjective and, since Q is F-selfcentralizing, τQX is injective [3, Remark 18.7]. Moreover, note that condition 18.6.3 in [3, Theorem 18.6] implies that
Q fulfills equality 17.10.1 in [3, Proposition 17.10]; in particular, extending PX
as in [3, 17.4], it follows from [3, Proposition 17.10] that PX is a coherent
FX-locality [3, 17.9]; that is to say, we have
[TABLE]
for any pair of subgroups Q and R in X, any x∈PX(Q,R) and any
v∈R.
2.2. Actually, considering the
normalizer NF(Q) of Q in F [3, 2.14] which is a Frobenius NP(Q)-category
[3, Proposition 2.16], denoting by XQ the set of subgroups of NP(Q) belonging to
X and setting PQ=NP(Q)
and FQ=NF(Q), we can also consider
the normalizer NPX(Q) of Q in PX [3, 17.4 and 17.5]
and, setting FX,Q=(FQ)XQ, it is not difficult to see that
PX,Q=NPX(Q) is actually a
perfect FX,Q-locality.
2.3. Moreover, the FX,Q-locality PX,Q and the group
PX(Q) are related throughout the transporter of the p-subgroups of
PX(Q); explicitly, let us call
transporter TPX(Q) of PX(Q) the FQ-locality formed by the category where the objects are all the subgroups of PQ, where the morphisms are defined by the
elements of the PX(Q)-transporters of the corresponding
τQX-images, and where the composition is defined by the product in PX(Q), endowed with the obvious functors induced by τQX and by πQX. Then, denoting by TPX(Q)XQ the full subcategory of
TPX(Q) over XQ, we claim that we have an
FQ-locality equivalence [3, 2.9]
[TABLE]
Firstly, we need the following lemma which admits the same proof as in [3, Proposition 24.2].
Lemma 2.4. Any PX-morphism is a monomorphism and an epimorphism.
2.5. Now, we already know that any PX,Q-morphism x:T→R is induced by a
PX-morphism x^:T⋅Q→R⋅Q which stabilizes Q [3, 2.14.1];
then, it easily follows from the lemma above that x^ is uniquely determined by x, and the
divisibility of PX guarantees the existence of a unique
x^Q∈PX(Q) fulfilling
[TABLE]
moreover, from the coherence of PX (cf. 2.1.2), for any t∈T\iPQ we get
[TABLE]
so that from the lemma above we still get
[TABLE]
thus, the element x^Q belongs to the PX(Q)-transporter
{\cal T}_{{\cal P}^{{}^{{X}}}\!(Q)}\big{(}\tau^{{}_{X}}_{{}_{Q}}(R),\tau^{{}_{X}}_{{}_{Q}}(T)\big{)}
and it is not difficult to check that the correspondence sending the PX,Q-mor-phism
x:T→R to the TPX(Q)-morphism x^Q:T→R
defines a faithful FQ-locality functor PX,Q→TPX(Q)XQ [3, 2.9]. The “surjectivity” follows again from
condition 18.6.3 in [3, Theorem 18.6].
2.6. But, for any F-selfcentralizing subgroup W of P fully normalized in F,
we still have the normalizer FW=NF(W); let us set PW=NP(W);
if PW belongs to X, so that the set XW of subgroups of PW belonging to X is not empty, then we also
can consider the normalizer PX,W=NPX(W), which is again a
perfect FX,W-locality, and we always have the existence of the
F-localizer LFW of W [3, Theorem 18.6]; thus, we still can consider the
transporter TLFW of LFW as an FW-locality and the full subcategory TLFWXW of TLFW
over XW as the set of objects. Finally, we say that the perfect FX-locality PX is extendable whenever for any F-selfcentralizing subgroup W of P fully normalized in F such that PW∈X there exists an FX,W-locality isomorphism†††
In [5, 6.18], arguing by induction we claim such an equivalence but, with the notation there, if the group U is normal in ℱ then the induction argument cannot be applied!
[TABLE]
note that, according to 2.3.1, we may assume that W does not belong to X.
Proposition 2.7. If PX is an extendable perfect FX-locality
then, for any F-selfcentralizing subgroup V of P fully normalized in F such that
PV∈X,
PX,V is an extendable perfect FX,V-locality.
Proof: From our definition we have an FV-locality isomorphism
[TABLE]
which determines an NFV(W)-locality isomorphism
[TABLE]
where we identify PV with its image in LFV and, for any FV-selfcentralizing subgroup W of PV fully normalized in FV such that
NPV(W)∈XV, we denote by XV,W the set of subgroups of NPV(W) belonging to XV
But, it is not difficult to check that the normalizer NLFV(W), endowed with the group homomorphisms
[TABLE]
induced by the structural group homomorphisms of LFV, is the FV-localizer
of W. We are done.
3. A reduction procedure
3.1. With the notation above, recall that a basic P×P-set [3, 21,4] is a finite nonempty P×P-set Ω such that {1}×P acts freely on Ω, that we have
[TABLE]
where we denote by Ω∘ the P×P-set obtained by exchanging both factors,
and that, for any subgroup Q of P and any injective group homomorphism φ:Q→P such that Ω contains a P×P-subset isomorphic to (P×P)/Δφ(Q) where we set
Δφ(Q)={(φ(u),u)}u∈Q, we have a Q×P-set isomorphism
[TABLE]
3.2. Denoting by GΩ the group of automorphisms of the {1}×P-set
Res{1}×P(Ω), it is clear that we have an injective map from P×{1}
in GΩ; we identify its image with the p-group P so that, from now on, P is contained in GΩ and acts freely on Ω. Recall that the full subcategory of the
GΩ-transporter over the set of subgroups of P induces a
Frobenius P-category [3, Proposition 21.9] and we say that Ω is an
F-basic P×P-set if, for any pair of subgroups Q and R of P,
we have
[TABLE]
3.3. Actually, it follows from [3, Proposition 21.12] that an F-basic P×P-set
always exists; more precisely, we say that an F-basic P×P-set Ω is natural if it fulfills [5, 3.5]
[TABLE]
for any F-selfcentralizing subgroup Q of P and any φ∈F(P,Q), and if it is thick
[3, 21.7] outside of the set of F-selfcentralizing subgroups of P — namely the multiplicity of (P×P)/Δψ(R) is at least two if R is not F-selfcentralzing and ψ belongs to F(P,R). The existence of natural F-basic P×P-sets follows from [5, Proposition 3.4] together with [3, Proposition 21.12];
here, we are interested in the following form of [5, Proposition 3.7]
Proposition 3.4 Let Ω be a natural F-basic P×P-set, Q and T a pair of F-selfcentralizing subgroups of P and η an element of F(Q,T). The multiplicity of
(Q×P)/Δη(T) in ResQ×P(Ω) is at most one,
and if it is one then we have
[TABLE]
3.5. From now on, Ω is a natural F-basic P×P-set. For any subgroup Q of P, it is clear that CGΩ(Q) is just the group of automorphisms of the
Q×P-set ResQ×P(Ω) and it is clear that the correspondence
sending Q to CGΩ(Q) induces a contravariant functor CGΩ from F to the category Gr of finite groups. Let us denote by CGΩnsc(Q) the subgroup of
elements f∈CGΩ(Q) which act trivially
on all the Q×P-orbits of Ω isomorphic to (Q×P)/Δη(T) where T is
F-selfcentralizing; in particular, if Q is not F-selfcentralizing then we have
CGΩnsc(Q)=CGΩ(Q); in any case,
CGΩnsc(Q) is normal in CGΩ(Q) and, according to
Proposition 3.4, the quotient CGΩ(Q)/CGΩnsc(Q) is Abelian.
3.6. More generally, for any Q∈X denote by CGΩCX(Q) the subgroup of
elements f∈CGΩ(Q) which act trivially on all the Q×P-orbits
of Ω isomorphic to (Q×P)/Δη(T) where T belongs to X;
it is easily checked that the correspondence sending Q∈X to CGΩCX(Q) defines a subfunctor CGΩCX:FX→Gr of the restriction
of CGΩ to FX, and we consider the
quotient FX-locality TGΩX=TGΩX/CGΩCX — noted
Lˉn,X in [5, 5.1.2] — sending any pair of groups Q and R in X to
[TABLE]
here we are interested in the following form of [5, Corollaries 5.20 and 5.21].
Proposition 3.7. For any perfect FX-locality PX there is a unique naturally FX-isomorphic class of faithful FX-locality functors
λX:PX→TGΩX.
Moreover, if P′X is a perfect FX-locality which is
FX-locality isomorphic to PX then there is a commutative diagram of
FX-locality functors
[TABLE]
3.8. With the notation in 2.2 above, for any F-selfcentralizing subgroup W of P fully normalized in F such that PW∈X, it follows from [3, Proposition 21.11] that the subset of Ω
[TABLE]
is actually an FW-basic PW×PW-set; mutatis mutandi, denote by GΩW the group of {1}×PW-set automorphisms of
Res{1}×PW(ΩW) and identify PW with
PW×{1}; since the quotient NGΩ(W)/CGΩ(W) is isomorphic to F(W)
(cf. 3.2.1), it is clear that NGΩ(W) stabilizes ΩW and therefore we have a canonical group homomorphism from NGΩ(W) to GΩW; again, we are interested in the following form of [5, Proposition 6.15].
Proposition 3.9. With the notation above, for any pair of subgroups Q and R of
PW containing W and any element φ in FW(Q,R), there exists at most one Q×PW-orbit in ΩW isomorphic to
(Q×PW)/Δφ(R), ΩW is a natural
FW-basic PW×PW-set and, in particular, CGΩW(Q) is an Abelian p-group.
3.10. It follows from this proposition that, as in 3.6 above, if PW belongs to X
then we get the quotient FX,W-locality
TGΩWXW; actually, it follows from
Propositions 3.4 and 3.9 above that, with the notation in 2.2 and 2.6 above, the canonical group homomorphism from NGΩ(W) to GΩW induces an
FX,W-locality functor
[TABLE]
note that, according to Proposition 3.7 above, we have faithful FX,W-locality functors from PX,W=NPX(W) to both FX,W-localities NTGΩX(W) and TGΩWXW
and we may assume that they agree with gΩX,W.
3.11. On the other hand, let LFW be the F-localizer of W
[3, Theorem 18.6]; that is to say, LFW is a finite group endowed with an injective and a surjective group homomorphisms
[TABLE]
τFW(PW) is a Sylow p-subgroup of LFW, the composition
πFW∘τFW is defined by the conjugation in F(W) and we also have the exact sequence
[TABLE]
Below, we restate [5, Proposition 6.19].
Proposition 3.12. With the notation above, there is a unique
CGΩW(W)- conjugacy class of group homomorphisms
[TABLE]
compatible with the structural group homomorphisms from PW and to F(W).
3.13. As in 2.6 above, denote by TLFW the FW-locality determined by
τFW and by the transporter of the group LFW; it is clear that any group homomorphism λFW:LFW→NGΩW(W) in 3.12.1 above determines an FW-locality functor
[TABLE]
and two of them are naturally FW-isomorphic [5, 2.9]; moreover, if
PW∈X, it is not difficult to see that the full subcategory
TLFWXW of TLFW over XW as the set of objects is a
perfect FX,W-locality, and from 3.13.1 we get an
FX,W-locality functor
[TABLE]
4. Existence and uniqueness of an extendable perfect FX-locality
4.1. With the notation in 1.3 above, our main purpose is to prove that
Theorem. There exists an
extendable perfect FX-locality PX, which is unique up to
FX-locality isomorphisms.
The existence and the uniqueness of the F-localizer LFP of P [3, Theorem 18.6]
proves the existence and the uniqueness of the extendable perfect FX-locality whenever X={P}; indeed, LFP is actually a semidirect product P\mathchar9583K where
K≅F(P)/FP(P) is a p′-group and, for any F-selfcentralizing normal subgroup W of
P, the FX,W-locality equivalence 2.6.1 is obvious.
4.2. Thus, we may assume that X={P} and will argue by induction
on ∣X∣. Choose a minimal element U in X fully normalized
in F and set
[TABLE]
then, by the induction hypothesis, we may assume that there exists an
extendable perfect FY-locality PY, endowed with the
structural functors
[TABLE]
which is unique up to FY-locality isomorphisms. At this point, according to
Proposition 3.7 above, we may assume that PY is an
FY-sublocality of the FY-locality
TGΩY introduced in 3.6 above; then, denoting by
(TGΩX)Y the full subcategory of
TGΩX over Y as the set of objects, we have an obvious functor
(TGΩX)Y⟶TGΩY and we look to the pull-back
[TABLE]
which defines a coherent FY-locality MΩ,Y
[3, 17.9] endowed with obvious structural functors
[TABLE]
4.3. We extend MΩ,Y to a coherent
FX-sublocality MΩ,X of
TGΩX which contains MΩ,Y
as a full subcategory over Y and fulfills
[TABLE]
for any Q∈X and any V∈X−Y, and denote by
[TABLE]
the corresponding structural functors; finally, we consider the quotient
FX-lo-cality MˉΩ,X of MΩ,X defined by
[TABLE]
for any Q,R∈X,
together with the induced natural maps — denoted by υˉΩ,X and
ρˉΩ,X. Then, the proof of the Theorem above can be reduced to the proof of the following fact, that we prove in the next section
4.3.4. The structural functor ρˉΩ,X admits an
FX-locality functorial section.
4.4. Let us first prove this reduction. Choose an FX-locality functorial section
σˉΩ,X:FX→MˉΩ,X; for any
pair of groups Q and R in Y, we know that (cf. 2.1)
[TABLE]
and therefore, denoting by PΩ,Y(Q,R) the converse image of
\bar{\sigma}^{{}_{\Omega,{X}}}_{{}_{Q,R}}\big{(}{\cal F}^{{}^{{X}}}\!(Q,R)\big{)}
in MΩ,X(Q,R), it is clear that the canonical map
MΩ,X(Q,R)→PY(Q,R) induces a bijection
PΩ,Y(Q,R)≅PY(Q,R); that is to say, looking to the pull-back
[TABLE]
— which defines a coherent FX-locality PΩ,X
[3, 17.9] endowed with obvious structural functors
[TABLE]
— and denoting by (PΩ,X)Y the full subcategory of
PΩ,X over Y as the set of objects, it follows from those
bijections above that we have an FY-locality isomorphism
(PΩ,X)Y≅PY.
4.5. That is to say, for any Q∈Y fully normalized in F, we already know that
PΩ,X(Q) is an F-localizer of Q and,
for any V∈X−Y, it follows from the pull-back 4.4.2 above that we have the exact sequence
[TABLE]
and it is easily checked that the group PΩ,X(V), endowed with the group homomorphisms
[TABLE]
determined by the functors τΩ,X and πΩ,X,
is actually an FX-localizer of V whenever V is fully normalized in F;
consequently, it follows from 2.1 above that PΩ,X is a perfect FX-locality.
4.6. We claim that PΩ,X is actually an extendable perfect
FX-locality; indeed, let W be an F-selfcentralizing subgroup of P fully
normalized in F such that PW=NP(W) belongs to X; thus, if PW does not belong to Y then we have XW={PW} and PW is the unique object in both FX,W-localities NPΩ,X(W) and
TLFWXW; in this case, since
[TABLE]
where K≅FX,W(PW)/FPW(PW), it is clear that we get the equivalence 2.6.1.
Otherwise YW is not empty and, setting PΩ,X,W=NPΩ,X(W) and denoting by
PΩ,Y,W the full subcategory of PΩ,X,W over
YW, from 4.4 above we get an FY,W-locality isomorphism
[TABLE]
but, since PY is extendable, it follows from our definition in 2.6 above that we still get
an FY,W-locality isomorphism
[TABLE]
4.7. Always assuming that YW is not empty, note that in 3.10
above gΩY,W sends NPY(W) isomorphically to its image in TGΩWYW — still noted
PΩ,Y,W;
then, from this inclusion, mutatis mutandi we can define a coherent FY,W- locality MΩ,Y,W as in 4.2.3, and
coherent FX,W-localities MΩ,X,W\iTGΩWXW and MˉΩ,X,W as in 4.3; moreover, it is clear that σˉΩ,X induces an
FX,W-locality functorial section
σˉΩ,X,W:FX,W→MˉΩ,X,W and that we can define a coherent FX,W-locality PΩ,X,W as in 4.4.2 above which still fulfills
[TABLE]
we denote by τΩ,X,W:TPWXW→PΩ,X,W and by
πΩ,X,W:PΩ,X,W→FX,W
the structural functors.
4.8. On the other hand, since TLFWXW is a perfect
FX,W-locality (cf. 3.13), it follows from Proposition 3.7 (or from 3.13.2) that
TLFWXW is actually an FX,W-sublocality of
TGΩWXW; in particular, denoting by
(TGΩWXW)YW and by
(TLFWXW)YW the respective full subcategories
of TGΩWXW and of TLFWXW over YW, it is easily checked that the canonical functor
[TABLE]
sends (TLFWXW)YW isomorphically onto
TLFWYW\iTGΩWYW.
4.9. Moreover, from 4.4 we know that the canonical functor
[TABLE]
sends (PΩ,X)Y isomorphically onto PY;
but, it follows from our definition in 3.10 that, denoting by
(gΩX,W)Y the restriction of
gΩX,W to the normalizer in (TGΩX)Y of W, we have a commutative diagram of functors
[TABLE]
where the vertical arrows are defined by the functors 4.8.1 and 4.9.1; hence, since the functor 4.9.1 sends (PΩ,X)Y isomorphically onto PY
(cf. 4.4), this functor sends N(PΩ,X)Y(W) isomorphically onto
NPY(W) and we already know that gΩY,W sends NPY(W) isomorphically onto PΩ,Y,W (cf. 4.7), which is isomorphic to
TLFWYW (cf. 4.6.2).
4.10. At this point, it follows from Proposition 3.7 that there exist an
FY,W-locality functor lFY,W:TLFWYW→TGΩWYW
which sends TLFWYW isomorphically to PΩ,Y,W,
and that this functor is naturally FY,W-isomorphic to the inclusion
TLFWYW\iTGΩWYW
in 4.8 above; that is to say, according to our definition in [5, 2.9] and since the kernel of the structural group homomorphism from
TGΩWYW(PW) to FY,W(PW) is the image of CGΩW(PW)\iTGΩW(PW), there is
z∈CGΩW(PW) such that, denoting
by zQYW the image of z in
TGΩWYW(Q) for any Q∈YW, we get
[TABLE]
in TGΩWYW(Q,R), for any pair of groups
Q and R in YW.
4.11. But, we also can consider the images zQXW of z in
TGΩWXW(Q) for any
Q∈XW. Hence, up to replacing our choice of
TLFWXW as a FX,W-sublocality of
TGΩWXW by the choice
of zQXW⋅TLFWXW(Q,R)⋅(zRXW)−1 in TGΩWXW(Q,R), for any pair of groups Q and R in XW, in
TGΩWYW we actually may assume that we get
[TABLE]
In this situation, it follows from our definitions in 4.7 above that in
TGΩWXW the coherent
FX,W-sublocality MΩ,X,W contains
TLFWXW.
4.12. In particular, if XW=YW then we have
[TABLE]
so that we are done. Assume that XW=YW; then, by the very definition
of TGΩWXW
(cf. 3.6.1 and 4.3), for any V∈XW−YW we have
[TABLE]
and therefore, since p does not divide ∣F~W(PW,V)∣
[3, Proposition 6.7], we have a surjective group homomorphism
[TABLE]
mapping z=(zθ~)θ~∈F~W(PW,V) on
[TABLE]
4.13. At this point, considering the contravariant functor
[TABLE]
mapping any Q∈YW on {0} and any V∈XW−YW on
Ker(∇VΩ,X,W), and the quotient FX,W-locality
\overline{{\cal T}}^{{}^{{X}^{{}^{W}}}}_{\!G^{{}^{\Omega_{{}_{W}}}}}\big{/}{d}^{{}^{\Omega,{X},W}} [5, 2.10],
it is easily checked that the coherent FX,W-localities PΩ,X,W
(cf. 4.7) and TLFWXW have the same image in this quotient; indeed,
it follows from equalities 4.11.1 above that their images coincide over YW and, since for any
V∈XW−YW we have
[TABLE]
PΩ,X,W(V) and TLFWXW(V) map both isomorphically onto
\big{(}\overline{{\cal T}}^{{}^{{X}^{{}^{W}}}}_{\!G^{{}^{\Omega_{{}_{W}}}}}\big{/}{d}^{{}^{\Omega,{X},W}}\big{)}(V)\,.
In particular, we get PΩ,X,W≅TLFWXW since the functors
from PΩ,X,W and TLFWXW to the quotient \overline{{\cal T}}^{{}^{{X}^{{}^{W}}}}_{\!G^{{}^{\Omega_{{}_{W}}}}}\big{/}{d}^{{}^{\Omega,{X},W}} are faithful. This proves our claim in 4.6.
4.14. It remains to prove the uniqueness; thus, assume that PX and P′X
are two extendable perfect FX-localities; it follows from Proposition 3.7 that
we may assume that both are FX-sublocalities of the FX-locality
TGΩX introduced in 3.6 above. On the other hand, since the
respective full sub-categories PY of PX and P′Y of
P′X over Y as the set of objects are still two extendable perfect
FY-localities, it follows from our induction hypothesis that they are
FY-locality isomorphic. Consequently, considering the inclusions of PY and P′Y in TGΩY induced by the inclusions
[TABLE]
and by the canonical functor (TGΩX)Y→TGΩY (cf. 4.9.1), the existence of an
FY-locality isomorphism PY≅P′Y determines two
FY-locality functors from PY to
(TGΩX)Y; then, it follows again from Proposition 3.7 that the functors ainsi obtained are naturally
FY-isomorphic.
4.15. That is to say, as in 4.10 above, since the kernel of the structural group homomorphism from
TGΩY(P) to FY(P) is the image of
CGΩ(P) in TGΩY(P), there is
z∈CGΩ(P) such that, denoting by zQY the image of z in
TGΩY(Q) for any Q∈Y, in
TGΩY(Q,R) we get
[TABLE]
for any pair of groups Q and R in Y. As above, considering the images
zQX of z in TGΩX(Q) for any
Q∈X and modifying our choice of P′X as a FX-sublocality of
TGΩX by the choice
of zQX⋅P′X(Q,R)⋅(zRX)−1 in
TGΩX(Q,R) for any pair of groups Q and R in X, we actually may assume that in TGΩWYW we have
PY=P′Y.
4.16. Moreover, as in 4.12 above, by the very definition of TGΩX (cf. 3.6.1 and 4.3), for any V∈X−Y we have
[TABLE]
and therefore, since p does not divide ∣F~(P,V)∣
[3, Proposition 6.7], we have a surjective group homomorphism
[TABLE]
mapping z=(zθ~)θ~∈F~(P,V) on
[TABLE]
4.17. At this point, considering the contravariant Dirac functor
[TABLE]
mapping any Q∈Y on {0} and any V∈X−Y on
Ker(∇VΩ,X), and the quotient FX-locality
\overline{{\cal T}}^{{}^{{X}}}_{\!G^{{}^{\Omega}}}\big{/}{d}^{{}^{\Omega,{X}}} [5, 2.10],
it is easily checked that the coherent FX-localities PX
and P′X have the same image in this quotient; indeed,
it follows from 4.15 above that their images coincide over Y and, since for any
V∈X−Y we have
[TABLE]
PX(V) and P′X(V) map both isomorphically onto
\big{(}\overline{{\cal T}}^{{}^{{X}}}_{\!G^{{}^{\Omega}}}\big{/}{d}^{{}^{\Omega,{X}}}\big{)}(V)\,.
In particular, we get PX≅P′X since the functors
from PX and P′X to the quotient \overline{{\cal T}}^{{}^{{X}}}_{\!G^{{}^{\Omega}}}\big{/}{d}^{{}^{\Omega,{X}}} are faithful. This proves the uniqueness.
5. Existence and uniqueness of the sections from FX to
MˉΩ,X
5.1. With the hypothesis and notation in 4.3 above, our purpose in this section is to prove that
Theorem. The structural functor ρˉΩ,X:MˉΩ,X→FX admits an FX-locality functorial section
σˉΩ,X:FX→MˉΩ,X.
Actually, since we assume that U=P, we also have U=NP(U)=PU and therefore
PU belongs to YU; thus, this theorem is just the existence part of [5, Theorem 6.22] but we restate the proof in our new context; indeed, here we assume that PY is an extendable perfect
FY-locality and therefore the FY,U-locality isomorphism in [5, 6.18]
[TABLE]
follows from our definition in 2.6; in particular, as in 4.11 above, in
TGΩUYU we may assume that
PY,U=TLFUYU.
5.2. Since YU is not empty, as in 4.7 above we can define the
coherent FY,U-locality MΩ,Y,U via the
pull-back (cf. 4.2.3)
[TABLE]
and the coherent FX,U-localities MΩ,X,U\iTGΩWXU and MˉΩ,X,U as in 4.3, with the second structural functors
[TABLE]
Now recall that, denoting by F~X and F~X,U the respective
exterior quotients of FX and FX,U [3, 1.3], the coherency of
MˉΩ,X and MˉΩ,X,U determines contravariant functors [5, 2.8.3]
[TABLE]
as usual, the existence of σˉΩ,X depends on the vanishing of the cohomology class of a suitable Ker(ρˉΩ,X)-valued 2-cocycle and,
from the reduction procedure developed in section 3, we will move to the corresponding Ker(ρˉΩ,X,U)-va-lued 2-cocycle.
5.3. From the commutative diagram 4.2.3 we get the following commutative diagram of the
normalizers of U
[TABLE]
moreover, we are setting NPY(U)=PY,U and we have the commutative diagram 4.9.2 for W=U. Consequently, the FY,U- and FX,U-locality functors
(cf. 3.10.1)
[TABLE]
successively induce the new FY,U-locality functor (cf. 5.2.1)
[TABLE]
and, moreover, the FX,U-locality functors (cf. 4.3)
[TABLE]
Similarly, since we are assuming that PY,U=TLFUYU (cf. 4.11.1), the FX,U-lo-cality functor
(cf. 3.13.2)
[TABLE]
and the pull-back 5.2.1 above determine new FX,U-locality functors (cf. 4.3)
[TABLE]
5.4. At this point, denoting by ι~X,U:F~X,U→F~X the canonical functor, it is well-known that, for any n∈N, the restriction induces a group homomorphism (cf. 5.1.2)
[TABLE]
moreover, hˉΩX,U induces a natural map [5, 2.10.1]
[TABLE]
and therefore, for any n∈N, we also get a group homomorphism
[TABLE]
In [5, Proposition 6.9, 6.12.3 and 6.21.7] we prove that, for any n∈N, the composition
of the homomorphisms 5.4.1 and 5.4.3 determines an isomorphism
[TABLE]
5.5. Let us explicit the announced Ker(ρˉΩ,X)-valued 2-cocycle.
For any FX-morphism φ:R→Q, choose a lifting xφ
in MΩ,X(Q,R) (cf. 4.3) and denote by xˉφ the image of
xφ in MˉΩ,X(Q,R); actually, we can do our choice in such a way that we have (cf. 4.3)
[TABLE]
for any u∈Q, where κQX(u)∈FX(Q) denotes the conjugation by the image of u; indeed, if we have κQX(u)∘φ=φ then we get u=φ(z) for a suitable z∈Z(R); since
MˉΩ,X is coherent, in this case we obtain
[TABLE]
More precisely, if Q and R are contained in PU and φ:R→Q comes from an FX,U-morphism, it is quite clear that we may assume that xφ belongs to
\big{(}N_{\!{\cal M}^{{}^{\Omega,{X}}}}(U)\big{)}(Q,R) and then that
hΩX,U(xφ) belongs to the image
of TLFUXU(Q,R) via lFX,U, so that actually we have (cf. 5.3.6)
[TABLE]
5.6. Then, for any triple of subgroups Q, R and T in X,
and any pair of F-morphisms ψ:T→R and φ:R→Q, since
xφ⋅xψ and xφ∘ψ have the same image φ∘ψ in F(Q,T),
the divisibility of MΩ,X guarantees the existence and the uniqueness of
kφ,ψ∈Ker(ρTΩ,X) fulfilling
[TABLE]
Denote by kˉφ,ψ the image of kφ,ψ in
Ker(ρˉTΩ,X); since MˉΩ,X is
coherent, on the one hand for any u∈Q and any v∈R we get (cf. 5.5.1)
[TABLE]
hence, from the divisibility of MˉΩ,X we obtain
[TABLE]
That is to say, for any n∈N, setting [3, 1.5]
[TABLE]
we have obtained an element
kˉ={kˉq~}q~∈Fct(Δ2,F~X) in
{C}^{2}\big{(}\tilde{\cal F}^{{}^{\Omega,{X}}}\!,{K}{e}{r}(\bar{\rho}^{{}_{\Omega,{X}}})\big{)} where we set
kˉq~=kˉq~(1∙2),q~(0∙1)=kˉq(1∙2),q(0∙1) for some representative
q:Δ2→FX of q~.
5.7. We claim that kˉ is actually a 2-cocycle; explicitly, considering the usual differential map [3, A13.11]
[TABLE]
we claim that dˉ2Ω,X(kˉ)=0; indeed, with the notation above, for a third
FX-morphism η:W→T we get
[TABLE]
and the divisibility of MˉX forces
[TABLE]
since Ker(ρˉΩ,X) is Abelian, in the additive notation we obtain
[TABLE]
proving our claim.
5.8. Then, in order to prove the existence of a section σˉΩ,X,
it suffices to show that kˉ is a 2-coboundary and therefore, according to
isomorphism 5.4.4 above, it suffices to prove that the image via
νhˉΩX,U (cf. 5.4.2) of the restriction of kˉ to F~X,U is a
2-coboundary. But, for any pair of FX,U-morphisms
φ:R→Q and ψ:T→R, we have chosen xφ in
\big{(}N_{\!{\cal M}^{{}^{\Omega,{X}}}\!}(U)\big{)}(Q,R)\,, xψ in
\big{(}N_{\!{\cal M}^{{}^{\Omega,{X}}}\!}(U)\big{)}(R,T) and xφ∘ψ in
\big{(}N_{\!{\cal M}^{{}^{\Omega,{X}}}\!}(U)\big{)}(Q,T)\,, so that in equality 5.6.1 the element
kφ,ψ belongs to \big{(}N_{\!{\cal M}^{{}^{\Omega,{X}}}\!}(U)\big{)}(T) and therefore
we get (cf. 5.3.4)
[TABLE]
and therefore we still get
[TABLE]
so that equalities 5.5.3 force hˉΩX,U(kˉφ,ψ)=1;
that is to say, the image via νhˉΩX,U of the restriction of kˉ to F~X,U is just trivial, proving that kˉ is a 2-coboundary.
5.9. Thus, we have obtained a functorial section σˉΩ,X:FX→MˉΩ,X of ρˉΩ,X; actually, σˉΩ,X can be modified in order to get an FX-locality functorial section [5, 2.9]. Indeed, for any FPX-morphism
ζ:R→Q, choosing uζ in TP(R,Q) lifting ζ,
both MˉΩ,X-morphisms σˉQ,RΩ,X(ζ) and υˉQ,RΩ,X(uζ) (cf. 4.3)
lift ζ; once again, the divisibility of MˉΩ,X guarantees the existence and the uniqueness of mˉζ∈Ker(ρˉRΩ,X) fulfilling
[TABLE]
actually, it follows easily from 5.5.1 that mˉζ only depends on
ζ~∈F~P(Q,R) and, as above, we write mˉξ~ instead of mξ; moreover, for a second FPX-mor-phism ξ:T→R, we get
[TABLE]
5.10. Then, always the divisibility of MˉΩ,X forces
[TABLE]
and, since Ker(ρˉTΩ,X) is Abelian (cf. Proposition 3.4, 3.5 and 3.6), in the additive notation we obtain
[TABLE]
that is to say, denoting by ι~PX:F~PX\iF~X the obvious inclusion functor, the correspondence mˉ sending any
F~PX-morphism ζ~:R→Q
to mˉζ~ defines a 1-cocycle in {C}^{1}\big{(}\tilde{\cal F}^{{}^{{X}}}_{\!P},{K}{e}{r}(\bar{\rho}^{{}^{\Omega,{X}}})\circ\tilde{\iota}^{{}_{X}}_{P}\big{)}\,; but, since the category F~PX obviously
has a final object, we actually have [3, Corollary A4.8]
[TABLE]
consequently, we obtain mˉ=d0Ω,X(wˉ) for some element
wˉ=(wˉQ)Q∈X in
[TABLE]
In conclusion, equality 5.9.1 becomes
[TABLE]
thus, the new correspondence which, for any pair of subgroups Q and R in X, sends any φ∈F(Q,R) to wˉQ⋅σˉQ,RΩ,X(φ)⋅wˉR−1 defines an FX-locality functorial section
of ρˉX. We are done.
References
[1] Bob Oliver, A remark on the construction of centric linking systems,
arxiv.org/abs/1612.02132
[2]. Lluís Puig, Frobenius categories,
Journal of Algebra, 303(2006), 309-357.
[3]. Lluís Puig, “Frobenius categories versus Brauer blocks”, Progress in Math.
274(2009), Birkhäuser, Basel.
[4]. Lluís Puig, A criterion on trivial homotopy, arxiv.org/abs/1308.3765.
[5]. Lluís Puig, Existence, uniqueness and functoriality of the perfect locality over a Frobenius P-category, Algebra Colloquium, 23(2016) 541-622.