Quantization of Poisson Groups
Extracted source text ยท arXiv q-alg/9511022
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% %%
%% QUANTIZATION OF POISSON GROUPS %%
%% %%
%% by %%
%% %%
%% Fabio GAVARINI %%
%% %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\input amstex
\documentstyle{amsppt}
\magnification=\magstep1
\hsize=6.5truein
\vsize=9truein
\document
\baselineskip=.15truein
%\baselineskip= 13pt
\font \cs=eufm10
\font \cal=cmsy10
\font \fnum=cmbx10
\def \aij {a_{ij}}
\def \unon {1,\dots,n}
\def \N {\Bbb N}
\def \Z {\Bbb Z}
\def \Q {\Bbb Q}
\def \C {\Bbb C}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% %%%
%% WARNING: in case your TeX fonts do not include the %%
%% \Bbb font, use the following alternative definitions %%
%%% %%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% \def \N{\hbox{\rm I\hskip -2pt N}}
% \def \Z{\hbox{\rm Z\hskip -4,5pt Z}}
% \def \Q{\hskip0.5mm \hbox{\rm l\hskip -4.5pt Q\/}}
% \def \C{\hskip0.5mm \hbox{\rm l\hskip -4.5pt C\/}}
%
\def \qm {q^{-1}}
\def \kq {k(q)}
\def \kqqm {k[q,\qm]}
\def \ehat {\widehat E}
\def \fhat {\widehat F}
\def \ebar {\overline E}
\def \fbar {\overline F}
\def \rbar {\overline r}
\def \otimeshat {\,\widehat\otimes\,}
\def \gerE {\hbox{{\cs E}}}
\def \germ {\hbox{{\cs m}}}
\def \gerg {\hbox{{\cs g}}}
\def \gerh {\hbox{{\cs h}}}
\def \gerk {\hbox{{\cs k}}}
\def \gerK {\hbox{{\cs K}}}
\def \gert {\hbox{{\cs t}}}
\def \gern {\hbox{{\cs n}}}
\def \gerb {\hbox{{\cs b}}}
\def \ug {U(\gerg)}
\def \ugtau {U(\gerg^\tau)}
\def \uh {U(\gerh)}
\def \uhtau {U(\gerh^\tau)}
\def \uqPg {U_q^P(\gerg)}
\def \uqPgtilde {\widetilde{U}_q^P(\gerg)}
\def \uqQg {U_q^Q(\gerg)}
\def \uqQghat {\widehat{U}_q^Q(\gerg)}
\def \uunoQghat {\widehat{U}_1^Q(\gerg)}
\def \uepsilonQghat {\widehat{U}_\varepsilon^Q (\gerg)}
\def \uepsilonPgtilde {\widetilde{U}_\varepsilon^P (\gerg)}
\def \uqPt {U_q^P(\gert)}
\def \uqPttilde {\widetilde{U}_q^P(\gert)}
\def \uqQt {U_q^Q(\gert)}
\def \uqQthat {\widehat{U}_q^Q(\gert)}
\def \uqPbm {U_q^P(\gerb_-)}
\def \uqPbmtilde {\widetilde{U}_q^P(\gerb_-)}
\def \uqPbam {U_q^Q(\gerb_-)}
\def \uqQbm {U_q^Q(\gerb_-)}
\def \uqQbmhat {\widehat{U}_q^Q(\gerb_-)}
\def \uqQbmtilde {\widetilde{U}_q^Q(\gerb_-)}
\def \uqPbp {U_q^P(\gerb_+)}
\def \uqPbptilde {\widetilde{U}_q^P(\gerb_+)}
\def \uqQbp {U_q^Q(\gerb_+)}
\def \uqQbphat {\widehat{U}_q^Q(\gerb_+)}
\def \uqMbm {U_q^M(\gerb_-)}
\def \uqhahat {\widehat{U}_q^Q(\gerh)}
\def \uqhstilde {\widetilde{U}_q^P(\gerh)}
\def \uunohstilde {\widetilde{U}_1^P(\gerh)}
\def \uepsilonhstilde {\widetilde{U}_\varepsilon^P(\gerh)}
\def \uqinfh {U_q^\infty(\gerh)}
\def \uqMinfh {U_q^{M,\infty}(\gerh)}
\def \uqinfhs {U_q^\infty^P(\gerh)}
\def \uqinfha {U_q^\infty^Q(\gerh)}
\def \uqinfhhat {\widehat{U}_q^\infty(\gerh)}
\def \uqinfhshat {\widehat{U}_q^{P,\infty}(\gerh)}
\def \uqinfhahat {\widehat{U}_q^{Q,\infty}(\gerh)}
\def \uunoinfhahat {\widehat{U}_1^{Q,\infty}(\gerh)}
\def \uepsiloninfhahat {\widehat{U}_\varepsilon^{Q,\infty}
(\gerh)}
\def \uqQhqinf {{U_q^Q(\gerh)}_{(q-1),\infty}}
\def \uqQhhatqinf {{\widehat{U}_q^Q (\gerh)}_{(q-1),\infty}}
\def \uunoQhhatqinf {{\widehat{U}_1^Q (\gerh)}_{(q-
1),\infty}}
\def \uepsilonQhhatqinf {{\widehat{U}_\varepsilon^Q
(\gerh)}_{(q-1),\infty}}
\def \uqQhhat {\widehat{U}_q^Q(\gerh)}
\def \uunoQhhat {\widehat{U}_1^Q(\gerh)}
\def \uepsilonQhhat {\widehat{U}_\varepsilon^Q (\gerh)}
\def \uqinfhtilde {\widetilde{U}_q^\infty(\gerh)}
\def \uqPinfhtilde {\widetilde{U}_q^{P,\infty}(\gerh)}
\def \uunoPinfhtilde {\widetilde{U}_1^{P,\infty}(\gerh)}
\def \uepsilonPinfhtilde {\widetilde{U}_\varepsilon^{P,\infty}
(\gerh)}
\def \fg {F[G]}
\def \fh {F[H]}
\def \finfg {F^\infty[G]}
\def \finfgtau {F^\infty[G^\tau]}
\def \finfh {F^\infty[H]}
\def \finfhtau {F^\infty[H^\tau]}
\def \fqQg {F_q^Q[G]}
\def \fqQinfg {F_q^{Q,\infty} [G]}
\def \fqQgqinf {{F_q^Q[G]}_{(q-1),\infty}}
\def \fqQgtilde {{\widetilde F}_q^Q[G]}
\def \fqQinfgtilde {{\widetilde F}_q^{Q,\infty} [G]}
\def \fqQgtildeqinf {{{\widetilde F}_q^Q[G]}_{(q-1),\infty}}
\def \funoQgtilde {\widetilde{F}_1^Q[G]}
\def \funoQinfgtilde {\widetilde{F}_1^{Q,\infty} [G]}
\def \fepsilonQgtilde {\widetilde{F}_\varepsilon^Q [G]}
\def \fepsilonQinfgtilde {\widetilde{F}_\varepsilon^{Q,\infty}
[G]}
\def \fqPg {F_q^P [G]}
\def \fqPinfg {F_q^{P,\infty} [G]}
\def \fqPghat {{\widehat F}_q^P [G]}
\def \fqPinfghat {{\widehat F}_q^{P,\infty} [G]}
\def \funoPghat {\widehat{F}_1^P [G]}
\def \fepsilonPghat {\widehat{F}_\varepsilon^P [G]}
\def \fqPbptilde {{\widetilde F}_q^P [B_+]}
\def \fqPbphat {{\widehat F}_q^P [B_+]}
\def \fqQbm {F_q^Q [B_-]}
\def \fqPbm {F_q^P [B_-]}
\def \fqPbp {F_q^P [B_+]}
\def \fqQbmtilde {{\widetilde F}_q^Q [B_-]}
\def \fqPbmtilde {{\widetilde F}_q^P [B_-]}
\def \fqQbmhat {{\widehat F}_q^Q [B_-]}
\def \fqPbmhat {{\widehat F}_q^P [B_-]}
\def \aPhat {\widehat{A}^P}
\def \aQtilde {\widetilde{A}^Q}
\def \aPinf {A^P_\infty}
\def \aPinfhat {\widehat{A}^P_\infty}
\def \aQinf {A^Q_\infty}
\def \aQinftilde {\widetilde{A}^Q_\infty}
\def \aQqinftilde {\widetilde{A}^Q_{(q-1),\infty}}
\def \frhatgerg {{\widehat{{\Cal F}r}}_{\gerg}}
\def \frhatgerh {{\widehat{{\Cal F}r}}_{\gerh}}
\def \frtildegerg {{\widetilde{{\Cal F}r}}_{\gerg}}
\def \frtildegerh {{\widetilde{{\Cal F}r}}_{\gerh}}
\def \frhatG {{\widehat{{\Cal F}r}}_G}
\def \frtildeH {{\widetilde{{\Cal F}r}}_H}
\def \uqphiMg {U_{q,\varphi}^M(\gerg)}
\def \uqphiPg {U_{q,\varphi}^P(\gerg)}
\def \uqphiPgtilde {\widetilde{U}_{q,\varphi}^P(\gerg)}
\def \uqphiQg {U_{q,\varphi}^Q(\gerg)}
\def \uqphiQinfg {U_{q,\varphi}^{Q,\infty} (\gerg)}
\def \uqphiQghat {\widehat{U}_{q,\varphi}^Q(\gerg)}
\def \uunophiQghat {\widehat{U}_{1,\varphi}^Q(\gerg)}
\def \uepsilonphiQghat {\widehat{U}_{\varepsilon,\varphi}^Q
(\gerg)}
\def \uqphiQinfghat {\widehat{U}_{q,\varphi}^{Q,\infty}
(\gerg)}
\def \uunophiPgtilde {\widetilde{U}_{1,\varphi}^P(\gerg)}
\def \uepsilonphiPgtilde {\widetilde{U}_{\varepsilon,\varphi}^P
(\gerg)}
\def \uqphiPt {U_{q,\varphi}^P(\gert)}
\def \uqphiPttilde {\widetilde{U}_{q,\varphi}^P(\gert)}
\def \uqphiQt {U_{q,\varphi}^Q(\gert)}
\def \uqphiQthat {\widehat{U}_{q,\varphi}^Q(\gert)}
\def \uqphiPbm {U_{q,\varphi}^P(\gerb_-)}
\def \uqphiPbmtilde {\widetilde{U}_{q,\varphi}^P(\gerb_-)}
\def \uqphiQbm {U_{q,\varphi}^Q(\gerb_-)}
\def \uqphiQbmhat {\widehat{U}_{q,\varphi}^Q(\gerb_-)}
\def \uqphiQbmtilde {\widetilde{U}_{q,\varphi}^Q(\gerb_-)}
\def \uqphiPbp {U_{q,\varphi}^P(\gerb_+)}
\def \uqphiPbptilde {\widetilde{U}_{q,\varphi}^P(\gerb_+)}
\def \uqphiQbp {U_{q,\varphi}^Q(\gerb_+)}
\def \uqphiQbphat {\widehat{U}_{q,\varphi}^Q(\gerb_+)}
\def \uqphiMbm {U_{q,\varphi}^M(\gerb_-)}
\def \uqphiQh {U_{q,\varphi}^Q(\gerh)}
\def \uqphiQhhat {\widehat{U}_{q,\varphi}^Q(\gerh)}
\def \uqphiPhtilde {\widetilde{U}_{q,\varphi}^P(\gerh)}
\def \uunophiPhtilde {\widetilde{U}_{1,\varphi}^P(\gerh)}
\def \uepsilonphiPhtilde {\widetilde{U}_{\varepsilon,\varphi}^P
(\gerh)}
\def \uqphiinfh {U_{q,\varphi}^\infty (\gerh)}
\def \uqphiMinfh {U_{q,\varphi}^{M,\infty} (\gerh)}
\def \uqphiPinfh {U_{q,\varphi}^{P,\infty} (\gerh)}
\def \uqphiQinfh {U_{q,\varphi}^{Q,\infty} (\gerh)}
\def \uqphiinfhhat {\widehat{U}_{q,\varphi}^\infty (\gerh)}
\def \uqphiPinfhhat {\widehat{U}_{q,\varphi}^{P,\infty}
(\gerh)}
\def \uqphiQinfhhat {\widehat{U}_{q,\varphi}^{Q,\infty}
(\gerh)}
\def \uunophiQinfhhat {\widehat{U}_{1,\varphi}^{Q,\infty}
(\gerh)}
\def \uepsilonphiQinfhhat {\widehat{U}_{\varepsilon,\varphi}^{Q,
\infty} (\gerh)}
\def \uqphiQhqinf {{U_{q,\varphi}^Q (\gerh)}_{(q-1),\infty}}
\def \uqphiQhhatqinf {{\widehat{U}_{q,\varphi}^Q (\gerh)}_{(q-
1),\infty}}
\def \uunophiQhhatqinf {{\widehat{U}_{1,\varphi}^Q
(\gerh)}_{(q-1),\infty}}
\def \uepsilonphiQhhatqinf {{\widehat{U}_{\varepsilon,\varphi}^Q
(\gerh)}_{(q-1),\infty}}
\def \uunophiQhhat {\widehat{U}_{1,\varphi}^Q (\gerh)}
\def \uepsilonphiQhhat {\widehat{U}_{\varepsilon,\varphi}^Q
(\gerh)}
\def \uqphiinfhtilde {\widetilde{U}_{q,\varphi}^\infty
(\gerh)}
\def \uqphiPinfhtilde {\widetilde{U}_{q,\varphi}^{P,\infty}
(\gerh)}
\def \uunophiPinfhtilde {\widetilde{U}_{1,\varphi}^{P,\infty}
(\gerh)}
\def \uepsilonphiPinfhtilde
{\widetilde{U}_{\varepsilon,\varphi}^{P,\infty} (\gerh)}
\def \fgtau {F[G^\tau]}
\def \fhtau {F[H^\tau]}
\def \finfgtau {F^\infty[G^\tau]}
\def \finfhuau {F^\infty[H\tau]}
\def \fqphiMg {F_{q,\varphi}^M[G]}
\def \fqphiMinfg {F_{q,\varphi}^{M,\infty} [G]}
\def \fqphiQg {F_{q,\varphi}^Q[G]}
\def \fqphiQinfg {F_{q,\varphi}^{Q,\infty} [G]}
\def \fqphiQgqinf {{F_{q,\varphi}^Q[G]}_{(q-1),\infty}}
\def \fqphiQgtilde {{\widetilde F}_{q,\varphi}^Q [G]}
\def \fqphiQinfgtilde {{\widetilde F}_{q,\varphi}^{Q,\infty}
[G]}
\def \fqphiQgtildeqinf {{{\widetilde F}_{q,\varphi}^Q
[G]}_{(q-1),\infty}}
\def \funophiQgtilde {\widetilde{F}_{1,\varphi}^Q [G]}
\def \funophiQinfgtilde {\widetilde{F}_{1,\varphi}^{Q,\infty}
[G]}
\def \fepsilonphiQgtilde {\widetilde{F}_{\varepsilon,\varphi}^Q
[G]}
\def \fepsilonphiQinfgtilde
{\widetilde{F}_{\varepsilon,\varphi}^{Q,\infty} [G]}
\def \fqphiPg {F_{q,\varphi}^P [G]}
\def \fqphiPinfg {F_{q,\varphi}^{P,\infty} [G]}
\def \fqphiPghat {{\widehat F}_{q,\varphi}^P [G]}
\def \fqphiPinfghat {{\widehat F}_{q,\varphi}^{P,\infty} [G]}
\def \funophiPghat {\widehat{F}_{1,\varphi}^P[G]}
\def \fepsilonphiPghat {\widehat{F}_{\varepsilon,\varphi}^P
[G]}
\def \fqphiPbp {F_{q,\varphi}^P [B_+]}
\def \fqphiPbptilde {{\widetilde F}_{q,\varphi}^P [B_+]}
\def \fqphiPbphat {{\widehat F}_{q,\varphi}^P [B_+]}
\def \fqphiPbm {F_{q,\varphi}^P [B_-]}
\def \fqphiPbmtilde {{\widetilde F}_{q,\varphi}^P [B_-]}
\def \fqphiPbmhat {{\widehat F}_{q,\varphi}^P [B_-]}
\def \fqphiQbm {F_{q,\varphi}^Q [B_-]}
\def \fqphiQbp {F_{q,\varphi}^Q [B_+]}
\def \fqphiQbmtilde {{\widetilde F}_{q,\varphi}^Q [B_-]}
\def \fqphiQbmhat {{\widehat F}_{q,\varphi}^Q [B_-]}
\def \aphiPhat {\widehat{A}^{\varphi,P}}
\def \aphiQtilde {\widetilde{A}^{\varphi,Q}}
\def \aphiPinf {A^{\varphi,P}_\infty}
\def \aphiPinfhat {\widehat{A}^{\varphi,P}_\infty}
\def \aphiQinf {A^{\varphi,Q}_\infty}
\def \aphiQinftilde {\widetilde{A}^{\varphi,Q}_\infty}
\def \aphiQqinftilde {\widetilde{A}^{\varphi,Q}_{(q-
1),\infty}}
\def \aphiMinf {A^{\varphi,M}_\infty}
\def \frhatgergtau {{\widehat{{\Cal F}r}}_{\gerg^\tau}}
\def \frhatgerhtau {{\widehat{{\Cal F}r}}_{\gerh^\tau}}
\def \frtildegergtau {{\widetilde{{\Cal F}r}}_{\gerg^\tau}}
\def \frtildegerhtau {{\widetilde{{\Cal F}r}}_{\gerh^\tau}}
\def \frhatGtau {{\widehat{{\Cal F}r}}_{G^\tau}}
\def \frtildehtau {{\widetilde{{\Cal F}r}}_{H^\tau}}
%\NoBlackBoxes
\abstract
The quantization of well-known pairs of Poisson groups of
a wide class is studied by means of Drinfeld's double
construction and dualization via formal Hopf algebras; new
quantized enveloping algebras $ \uqphiMinfh $ and quantum
formal groups $ F_{q,\varphi}^{M,\infty}[G] $ are
introduced (both in one-parameter and multi-parameter case),
and their specializations at roots of 1 (in particular, their
classical limits) are studied: a new insight into the
classical Poisson duality is thus obtained.
\endabstract
\topmatter
\title
Quantization of Poisson groups
\endtitle
\author
F. Gavarini
\endauthor
\affil
Dipartimento di Matematica, Istituto G. Castelnuovo, \\
Universit\`a degli studi di Roma "La Sapienza" \\
\endaffil
\address\hskip-\parindent
Department of Mathematics \newline
University of Rome "La Sapienza" \newline
Piazzale Aldo Moro 5 \newline
00185 Rome --- ITALY \newline
e-mail: \ gavarini\@mat.uniroma1.it
\endaddress
\endtopmatter
\footnote""{ Revised version of \ January
$ 29^{\text{th}} $ 1996 }
\vskip0,8truecm
\centerline{ \bf Introduction }
\vskip10pt
\hfill \hbox{\vbox{ \hbox{\it \ "Dualitas dualitatum }
\hbox{\it \ \ \;\, et omnia dualitas" }
\vskip4pt
\hbox{\sl N.~Barbecue, "Scholia" } } \hskip1truecm }
\vskip8pt
Let $ G $ be a semisimple, connected, and simply
connected affine algebraic group over an algebraically closed field of
characteristic zero; we consider a family of structures of Poisson group
on $ G $, indexed by a multiparameter $ \tau $, which generalize the
well-known Sklyanin-Drinfeld one (cf.~e.~g.~[DP], \S 11). Then every such
Poisson group $ G^\tau $ has a well defined corresponding dual Poisson
group $ H^\tau $, and $ \, \gerg^\tau := Lie(G^\tau) \, $ and $ \,
\gerh^\tau := Lie(H^\tau) \, $ are Lie bialgebras dual of each other.
\par
In 1985 Drinfeld ([Dr]) and Jimbo ([Ji]) provided
a quantization $ \uqQg $ of $ \, \ug = U(\gerg^0) \, $,
namely a Hopf algebra $ \uqQg $ over $ \kq $, presented by
generators and relations with a $\kqqm$-form $ \uqQghat $
which for $ \, q \rightarrow 1 \, $ specializes to
$ \ug $ as a Poisson Hopf coalgebra. This has been extended to general parameter $ \tau $ introducing multiparameter quantum groups $ \uqphiQg $ (cf.~[R], and [CV-1], [CV-2]). Dually, by means of a Peter-Weyl type axiomatic trick one constructs a Hopf algebra $ \fqPg $ of matrix coefficients of $ \uqQg $ with a $\kqqm$-form
$ \fqPghat $ which specializes to $ \fg $, as a Poisson Hopf algebra, for $ \, q \rightarrow 1 \, $; in particular $ \fqPghat $ is nothing but the Hopf subalgebra of "functions" in $ \fqPg $ which take values in $ \kqqm $ when "evaluated" on $ \uqQghat $ (in a word, the $\kqqm$-integer valued functions on $ \uqQghat $). This again extends to general $ \tau $ (cf.~[CV-2]).
\par
So far the quantization procedure only dealt with the Poisson group $ G^\tau $; the dual group $ H $ is involved defining a different $ \kqqm $-form
$ \uqPgtilde $ (of a quantum group $ \uqPg $) which specializes to $ \fh $ (as a Poisson Hopf algebra) for
$ \, q \rightarrow 1 \, $ (cf.~[DP]), with generalization to the general case possible again. Here a new phenomenon of "crossing dualities" (duality among enveloping and function algebra and duality among Poisson groups) occurs which was described (in a formal setting) by Drinfel'd (cf.~[Dr], \S 7). This leads to consider the following: let $ \fqQg $ be the quantum function algebra dual of
$ \uqPg $, and look at the "dual" of $ \uqPgtilde $ within $ \fqQg $, call it $ \fqQgtilde $, which should be the Hopf algebra of $\kqqm$-integer valued functions on $ \uqPgtilde $; then this should specialize to $ \uh $ (as a Poisson Hopf coalgebra) for $ \, q \rightarrow 1
\, $; the same conjecture can be formulated in the general case too.
\par
The starting aim of the present work was to achieve this
goal, i.~e.~to construct $ \fqQg $ and its $ \kqqm $-form
$ \fqQgtilde $, and to prove that $ \fqQgtilde $ is a
deformation of the Poisson Hopf coalgebra $ \uh $. This goal is succesfully attained by performing a suitable dualization of Drinfeld's quantum double; but by the way, this leads to discover a new "quantum group" (both one-parameter and multiparameter), which we call $ \uqMinfh $, which is to $ \uh $ as $ U_q^M(\gerg) $ is to $ \ug $; in particular it has an integer form $ \uqQhhat $ which is a quantization of $ \uh $, and an integer form
$ \uqPinfhtilde $ which is a quantization of $ \finfg $ (the function algebra of the formal Poisson group associated to $ G \, $); this is also produced in the shape
of a {\it quantum formal group\/} $ \fqPinfghat $.
Moreover, we exhibit a quantum analog of the Hopf pairings
$ \, \fh \otimes \uh \rightarrow k \, $, $ \, \fg \otimes
\ug \rightarrow k \, $, $ \, \finfg \otimes \ug \rightarrow
k \, $, and also a quantum analog of the Lie bialgebra
pairing $ \, \gerh \otimes \gerg \rightarrow k \, $, which
are strictly related with each other and with the already
known quantum pairing $ \, \fqPg \otimes \uqQg \rightarrow
\kq \, $: all this is stressed by introducing the definition
of {\it quantum Poisson pairing\/}. Once again all this extend to the
multiparameter case. Thus in particular we provide a quantization for a
wide class of Poisson groups (the $ H^\tau $'s); now, in the
summer of 1995 (when the present work was already
accomplished) a quantization of any Poisson group was
presented in [EK-1] and [EK-2], but greatest generality
implies lack of concreteness: in contrast, our construction
is extremely concrete, by hands; moreover, it allows
specialization at roots of 1, construction of quantum
Frobenius morphisms, and so on (like for $ \uqQg $ and
$ \uqPg $), which is not possible in the approach of [EK-1], [EK-2].
\par
Finally, a brief sketch of the main ideas of the paper. Our aim being to study the "dual" of a quantum group
$ U_{q,\varphi}^M(\gerg) $ ($ M $ being a lattice of
weights), we proceed as follows. First, we select as operation of "dualization"
(of a Hopf algebra $ \, H \, $) the most na\"\i{}ve one, namely taking the {\it full linear dual\/} (i.~e.~$ \, H^* \, $, rather than the usual Hopf dual $ \, H^\circ \, $), the latter being a {\it formal\/} Hopf algebra (rather than a common Hopf algebra). Second, we remark that $ U_{q,\varphi}^M (\gerg) $ is a quotient, as a Hopf algebra, of a quantum double
$ \, D_{q,\varphi}^M (\gerg) := D \left( \uqphiQbm,
U_{q,\varphi}^M(\gerb_+), \pi^\varphi_+ \right) \, $ (cf.~\S 3), hence its linear dual $ {U_{q,\varphi}^M (\gerg)}^* $ embeds into the formal Hopf algebra
$ {D_{q,\varphi}^M (\gerg)}^* $. Third, since $ \, D_{q,\varphi}^M (\gerg) \cong U_{q,\varphi}^M (\gerb_+) \otimes \uqphiQbm \, $ (as coalgebras) we have
$ \, {D_{q,\varphi}^M (\gerg)}^* \cong {U_{q,\varphi}^M
(\gerb_+)}^* \otimeshat {\uqphiQbm}^* \, $ (as algebras), where $ \, \otimeshat \, $
denotes topological tensor product. Fourth, DRT pairings among quantum Borel algebras
yield natural embeddings $ \, U_{q,\varphi}^M (\gerb_\pm) \hookrightarrow
{U_{q,\varphi}^{M'} \! (\gerb_\mp)}^* \, $ ($ M' $ being the dual lattice of $ M \, $). Fifth, from previous remarks we locate a special formal Hopf subalgebra of $ \, \left( U_{q,\varphi}^{M'} \!(\gerb_-) \otimeshat \uqphiPbp \right) \subseteq {D_{q,\varphi}^M (\gerg)}^* \, $. Seventh, we take out of this an axiomatic construction of a formal Hopf algebra
$ U_{q,\varphi}^{M'}(\gerh) $, by definition isomorphic to the one of step six and therefore naturally paired with
$ U_{q,\varphi}^M (\gerg) $: this is the object we were looking for, sort of completion of $ F_{q,\varphi}^{M'}
[G] $, and from its very construction the whole series of results we stated above will be proved with no serious trouble.
\par
Sections 1 to 4 introduce the already known material; the original part of the paper is sections 5 to 8. The Appendix describes in full detail the example $ \, G = SL(2) \, $.
\vskip15pt
\centerline { \bf Acknoledgements }
\vskip4pt
The author wishes to thank C.~Procesi for several fruitful
conversations, and C.~De Concini for some useful talk; he is also endebted with M.~Costantini, M.~Varagnolo, and I.~Damiani for many helpful discussions.
\vskip2truecm
\centerline{ \bf \S~1 The classical objects }
\vskip10pt
{\bf 1.1 Cartan data.} \ Let $ A:= {(\aij)}_{i,j=\unon} $ be a $ n \times n $ symmetrizable
Cartan matrix; thus we have $ \, \aij \in \Z \, $ with
$ \, a_{ii} = 2 $ and $ a_{ij} \leq 0 \, $ if $ i \neq
j $, and there exists a vector $ (d_1, \ldots, d_n) $
with relatively prime positive integral entries $ d_i $
such that $ \, (d_i \aij)_{i,j=\unon} \, $ is a symmetric
positive definite matrix. Define the {\it weight lattice\/} $ P $ to be the lattice with basis
$ \{ \omega_1, \dots, \omega_n \} $ (the {\it fundamental weights\/}); let \ $ P_+ = \sum^n_{i=1} \N \omega_i $ \ be the subset of {\it dominant integral weights\/}, let \ $ \alpha_j = \sum^n_{i=1} \aij \omega_i \; (j=\unon) $ \ be the {\it simple roots\/}, and \ $ Q = \sum^n_{j=1} \Z \alpha_j \; (\,\subset P \,) $ \ the {\it root lattice\/}, \ $ Q_+ = \sum^n_{j=1} \N \alpha_j $ \ the {\it positive root lattice\/}. Let $ W $ be the Weyl group associated to $ A $, with generators $ s_1, \ldots, s_n $, and let $ \, \Pi := \{ \alpha_1, \dots, \alpha_n \} \, $: then
$ \, R := W \Pi \, $ is the set of {\it roots\/}, $ \, R^+ = R \cap Q_+ \, $ the set of {\it positive roots\/}; finally, we set $ N:= \#(R^+) \; (\, = \vert W \vert) $.
\par
Define bilinear pairings $ \, \langle \ \vert \ \rangle
\colon Q \times P \rightarrow \Z \, $ and $ \, (\ \vert \
) \colon Q \times P \rightarrow \Z \, $ by $ \langle
\alpha_i \vert \omega_j \rangle = \delta_{ij} $ and
$ (\alpha_i \vert \omega_j) = \delta_{ij} d_i $. Then
$ (\alpha_i \vert \alpha_j) = d_i \aij $, giving a
symmetric $ \Z $-valued $ W $-invariant bilinear form on
$ Q $ such that $ (\alpha \vert \alpha) \in 2 \Z $. We
can also extend the $ \Z $-bilinear pairing $ \, (\ \vert
\ ) \colon Q \times P \rightarrow \Z \, $ to a (non-degenerate) pairing $ \, (\ \vert \ ) \colon (\Q \otimes_\Z
Q) \times (\Q \otimes_\Z P) \rightarrow \Z \, $ of $ \Q $-vector spaces by scalar extension: restriction gives a
pairing $ \, (\ \vert \ ) \colon P \times P \rightarrow \Q
\, $ (looking at $ P $ as a sublattice of $ \, \Q
\otimes_\Z Q \, $), which takes values in $ \Z[d^{-1}] $,
where $ \, d:= det \left( {(\aij)}_{i,j=1}^n \right) \, $.
Given a pair of lattices $ (M,M') $, with $ \, Q \leq M, M' \leq P \, $, we say that they are {\it dual of
each other\/} if
$$ M' = \{\, y \in P \mid (M,y) \subseteq \Z \,\} \quad ,
\qquad M = \{\, x \in P \mid (x,M') \subseteq \Z \,\} $$
the two conditions being equivalent; then for any lattice $
M $ with $ \, Q \leq M \leq P \, $ there exists a unique
dual lattice $ M' $ such that $ \, Q \leq M' \leq P \, $.
\vskip7pt
{\bf 1.2 The Poisson groups $ G $ and $ H $.} \ Let $ G $ be a connected simply-connected semisimple affine algebraic group over an algebraically closed field $ k $ of charachteristic 0. Fix a maximal torus $ T \leq G $ and opposite Borel subgroups $ B_\pm $, with unipotent subgroups $ \, U_\pm \, $, such that $ \, B_+ \cap B_- = T \, $, and let $ \, \gerg := Lie(G) \, $, $ \, \gert := Lie(T) \, $, $ \, \gerb_\pm := Lie(B_\pm) \, $, $ \, \gern_\pm := Lie(U_\pm) \, $; fix also
$ \, \tau := (\tau_1, \ldots, \tau_n) \in Q^n \, $ such that $ \, (\tau_i, \alpha_j) = - (\tau_j, \alpha_i) \, $ for all $ \, i, j= \unon \, $: when $ \, \tau = (0, \ldots, 0) \, $ we shall simply skip it throughout. Set
$ \, K = G \times G \, $, define $ \, G^\tau:= G \, $ embedded in $ K $ as the diagonal subgroup, and define a second subgroup
$$ H^\tau := \{\, (u_- t_-, t_+ u_+) \mid u_\pm \in U_\pm , t_\pm \in T , t_- t_+ \in \exp(\gert^\tau) \,\} \; (\, \leq B_- \times B_+ \leq K) $$
where $ \, \gert^\tau := \sum_{i=1}^n k \cdot h_{-\alpha_i
+ 2 \tau_i} \oplus h_{\alpha_i + 2 \tau_i} \leq \gert \oplus
\gert \leq \gerg \oplus \gerg = \gerk := Lie(K) \, $; thus
$$ \gerh^\tau := Lie(H^\tau) = (\gern_-,0) \oplus
\gert^\tau \oplus (0,\gern_+) \, . $$
\indent Thus we have a triple $ (K,G^\tau,H^\tau) $; this is an algebraic Manin triple (i.~e.~its "tangent triple" $ (\gerk, \gerg^\tau, \gerh^\tau) $ is a Manin triple), when a non-degenerate symmetric invariant bilinear form on $ \gerk $ is defined as follows: first rescale the Killing form $ \, (\ ,\ ) \colon \gerg \otimes \gerg \rightarrow k \, $ so that short roots of $ \gerg $ have square length 2; then define the form on $ \, \gerk = \gerg \oplus \gerg \, $ by
$$ \langle x_1 \oplus y_1 , x_2 \oplus y_2 \rangle := {1
\over 2} (y_1,y_2) - {1 \over 2} (x_1,x_2) \, . $$
\indent In general, if $ (\gerk', \gerg', \gerh') $ is any Manin triple, the bilinear form on $ \gerk' $ gives by restriction a non-degenerate pairing $ \, \langle \ , \ \rangle \colon \gerh' \otimes \gerg' \rightarrow k \, $ which is a pairing of Lie bialgebras, that is
$$ \langle x, [y_1,y_2] \rangle = \langle \delta_{\gerh}
(x), y_1 \wedge y_2 \rangle \, , \quad \langle [x_1,x_2],
y \rangle = \langle x_1 \wedge x_2, \delta_{\gerg} (y)
\rangle $$
for all $ x, x_1, x_2 \in \gerh' $, $ y, y_1, y_2 \in
\gerg' $, where $ \delta_{\gerh'} $,
resp.~$ \delta_{\gerg'} $, is the Lie cobracket of
$ \gerh' $, resp.~$ \gerg' $; we shall call such pairing
also {\it Poisson pairing\/}, and denote it by $ \,
\pi_{\Cal P}(h,g) := \langle h, g \rangle \, $ for all
$ \, h \in \gerh' $, $ g \in \gerg' $. In the present case the Poisson pairing is described by formulas
$$ \matrix
\langle \text{f}^\tau_i, f_j \rangle = 0 & \langle
\text{f}^\tau_i, h_j \rangle = 0
& \langle \text{f}^\tau_i, e_j \rangle = - {1 \over 2}
\delta_{ij} d_i^{-1} \\
\langle \text{h}^\tau_i, f_j \rangle = 0 & \langle
\text{h}^\tau_i, h_j
\rangle = a_{ij} d_j^{-1} = a_{ji} d_i^{-1} & \langle
\text{h}^\tau_i, e_j \rangle
= 0 \\
\langle\text{e}^\tau_i, f_j \rangle = {1 \over 2}
\delta_{ij} d_i^{-1} & \langle
\text{e}^\tau_i, h_j \rangle = 0 & \langle
\text{e}^\tau_i, e_j \rangle = 0 \\
\endmatrix \eqno (1.1-a) $$
where the $ \, \text{f}^\tau_s $, $ \text{h}^\tau_s $,
$ \text{e}^\tau_s $ ($ s=\unon $), resp.~$ \, f_s $,
$ h_s $, $ e_s $ ($ s=\unon $), are Chevalley-type generators of $ \gerh^\tau $, resp.~$ \gerg^\tau $, embedded inside $ \, \gerk = \gerg \oplus \gerg \, $, namely $ \, \text{f}^\tau_s = f_s \oplus 0 \, $, $ \, \text{h}^\tau_s = h_{-\alpha_s + 2 \tau_s} \oplus h_{\alpha_s + 2 \tau_s} \, $, $ \, \text{e}^\tau_s
= 0 \oplus e_s \, $, and $ \, f_s = f_s \oplus f_s \, $, $ \, h_s = h_s \oplus h_s \, $, $ \, e_s = e_s \oplus e_s \, $ (see below); moreover we also record that
$$ \langle\text{e}^\tau_\alpha, f_\beta \rangle = {1 \over 2} \delta_{\alpha \beta} d_\alpha^{-1} \; , \qquad \langle \text{f}^\tau_\alpha, e_\beta \rangle = - {1 \over 2} \delta_{\alpha \beta} d_\alpha^{-1} \eqno (1.1-b) $$
for all $ \, \alpha, \beta \in R^+ \, $, where $ \,
\text{e}^\tau_\alpha, \text{f}_\alpha, e_\beta, f_\beta \, $ are Chevalley-type generators of
$ \, {(\gerh^\tau)}_\alpha $, $ {(\gerh^\tau)}_{-\alpha} $, $ {(\gerg^\tau)}_\beta $, $ {(\gerg^\tau)}_{-\beta} \, $ and $ \, d_\alpha := {(\alpha, \alpha) \over 2} \, $ for all $ \, \alpha \in R^+ \, $ (in particular
$ \, d_{\alpha_i} = d_i \, $ $ \, \forall \, i=\unon
\, $).\footnote{{\it Warning:\/} beware, in particular, of the normalization we chose for the symmetric form of $ \gerk $, which is different (by a coefficient $ 1 \over
2 $) from the one fixed in [DP]. }
\vskip7pt
{\bf 1.3 The Poisson Hopf coalgebra $ \ugtau $.} \ It
is known that the universal enveloping algebra $ \, \ugtau = \ug \, $ can be presented as the associative \hbox{$k$-alge}bra with 1 generated by elements $ , f_i $, $ h_i $, $ e_i $ ($ i=\unon $) (the {\it Chevalley generators\/}) satisfying the well-known Serre's relations (cf.~for instance [Hu], ch.~V). As an enveloping algebra of a Lie algebra, $ \ugtau $ has a canonical structure of Hopf algebra, given by $ \, \Delta(x) = x \otimes 1 + 1 \otimes x $, $ S(x) = -x $, $ \varepsilon(x) = 0 \, $ for all $ \, x \in \gerg \, $; finally, the Poisson structure of
$ G^\tau $ reflects into a Lie coalgebra
structure\break
$ \, \delta = \delta_{\gerg^\tau} \colon \gerg^\tau
\longrightarrow \gerg^\tau \otimes \gerg^\tau \, $ of
$ \gerg^\tau $, extending to a co-Poisson structure $ \, \delta \colon \ugtau \longrightarrow \ugtau \otimes \ugtau \, $ (compatible with the Hopf structure) given by (cf.~[DL], \S 8, and [DKP], \S 7.7)
$$ \eqalign{
\delta (f_i) & = {(\alpha_i + 2 \tau_i \vert\alpha_i +
2 \tau_i) \over 2} \, h_{\alpha_i + 2 \tau} \otimes f_i - {(\alpha_i + 2 \tau_i \vert\alpha_i + 2 \tau_i)
\over 2} \, f_i \otimes h_{\alpha_i + 2 \tau} \cr
\delta (h_i) & = 0 \cr
\delta (e_i) & = {(\alpha_i - 2 \tau_i \vert\alpha_i -
2 \tau_i) \over 2} \, h_{\alpha_i - 2 \tau} \otimes e_i -
{(\alpha_i - 2 \tau_i \vert\alpha_i - 2 \tau_i) \over 2} \, e_i
\otimes h_{\alpha_i - 2 \tau} \; . \cr } \eqno (1.2) $$
\vskip7pt
{\bf 1.4 The Poisson Hopf coalgebra $ \uhtau $.} \ We already remarked that $ \, \gerh^\tau \cong \gern_- \oplus \gert \oplus \gern_+ \cong \gern_- \oplus \gert^\tau \oplus \gern_+ \, $ as vector spaces: then the Lie structure of $ \gerh^\tau $ is uniquely determined by the following constraints: \, {\it (a)\/} $ \, \gern_- $,
$ \gert^\tau $, $ \gern_+ $ are Lie subalgebras of
$ \gerh^\tau $; \, {\it (b)\/} $ \, [n_+,n_-] = 0 \;
\forall \, n_+ \in \gern_+, n_- \in \gern_- \, $; \, {\it
(c)\/} $ \, [\text{h}^\tau_i, n_-] = - \langle \alpha_i - 2
\tau_i , \beta \rangle \, n_- \; \forall \, i=\unon, n_- \in
{(\gern_-)}_\beta , \beta \in R^- \, $;\break
{\it (d)\/} $ \, [\text{h}^\tau_i,n_+] = \langle \alpha_i + 2 \tau_i , \beta \rangle \, n_+ \; \forall \, i=\unon, n_+ \in {(\gern_+)}_\alpha , \alpha \in R^+ \, $.
\par
Thus if $ \text{f}^\tau_i $, $ \text{h}^\tau_i $,
$ \text{e}^\tau_i $ ($ i = \unon $) are Chevalley-type generators of $ \gerg $ (thought of as elements of
$ \gern_- $, $ \, \gert \cong \gert^\tau \, $,
$ \gern_+ $), $ \uhtau $ can be presented as the associative \hbox{$k$-alge}bra with 1 generated by
$ \text{f}^\tau_i $, $ \text{h}^\tau_i $,
$ \text{e}^\tau_i $ ($ i=\unon $) with relations
$$ \eqalign {
\text{h}^\tau_i \text{h}^\tau_j - \text{h}^\tau_j
\text{h}^\tau_i = 0 \, , & \qquad \text{e}^\tau_i
\text{f}^\tau_j - \text{f}^\tau_j \text{e}^\tau_i = 0 \cr
\text{h}^\tau_i \text{f}^\tau_j - \text{f}^\tau_j
\text{h}^\tau_i = \langle \alpha_i - 2 \tau_i ,
\alpha_j \rangle \text{f}^\tau_j \, , & \qquad
\text{h}^\tau_i \text{e}^\tau_j - \text{e}^\tau_j
\text{h}^\tau_i =
\langle \alpha_i + 2 \tau_i , \alpha_j \rangle
\text{e}^\tau_j \cr
\sum_{k=0}^{1-\aij} {(-1)}^k {1 - \aij \choose k} &
{(\text{f}^\tau_i)}^{1-\aij-k} \text{f}^\tau_j
{(\text{f}^\tau_i)}^k = 0 \qquad (i \neq j) \cr
\sum_{k=0}^{1-\aij} {(-1)}^k {1 - \aij \choose k} &
{(\text{e}^\tau_i)}^{1-\aij-k} \text{e}^\tau_j
{(\text{e}^\tau_i)}^k = 0 \qquad (i \neq j) \cr }
\eqno (1.3) $$
for all $ \, i, j= \unon \, $; its natural Hopf structure is given by
$$ \eqalign {
\Delta (\text{f}^\tau_i) = \text{f}^\tau_i \otimes 1
+ 1 \otimes \text{f}^\tau_i \, , &
\quad S(\text{f}^\tau_i) = - \text{f}^\tau_i \, , \quad
\varepsilon(\text{f}^\tau_i) = 0 \cr
\Delta (\text{h}^\tau_i) = \text{h}^\tau_i \otimes 1
+ 1 \otimes \text{h}^\tau_i \, , &
\quad S(\text{h}^\tau_i) = - \text{h}^\tau_i \, , \quad
\varepsilon(\text{h}^\tau_i) = 0 \cr
\Delta (\text{e}^\tau_i) = \text{e}^\tau_i \otimes 1
+ 1 \otimes \text{e}^\tau_i \, , &
\quad S(\text{e}^\tau_i) = - \text{e}^\tau_i \, , \quad
\varepsilon(\text{e}^\tau_i) = 0 \; . \cr }
\eqno (1.4) $$
and the co-Poisson structure $ \, \delta = \delta_{\gerh^\tau}
\colon \uhtau \longrightarrow \uhtau \otimes \uhtau \, $ by
$$ \eqalign {
\delta(\text{f}^\tau_i) & = d_i \text{h}^\tau_i
\otimes \text{f}^\tau_i - d_i \text{f}^\tau_i
\otimes \text{h}^\tau_i \cr
\delta(\text{h}^\tau_i) & = 4 \, d_i^{-1} \cdot
\sum_{\gamma \in R^+} d_\gamma (\gamma \vert \alpha_i) \cdot
(\text{e}^\tau_\gamma \otimes \text{f}^\tau_\gamma -
\text{f}^\tau_\gamma \otimes \text{e}^\tau_\gamma) \cr
\delta(\text{e}^\tau_i) & = d_i \text{e}^\tau_i
\otimes \text{h}^\tau_i - d_i \text{h}^\tau_i \otimes
\text{e}^\tau_i \; . \cr } \eqno (1.5) $$
\vskip2truecm
\centerline{ \bf \S~2 Quantum Borel algebras and DRT
pairings }
\vskip10pt
{\bf 2.1 Notations.} \ We shall introduce our quantum
groups using the construction of [DL] and [CV-1], [CV-2]. For all $ \, s, n \in \N \, $, let
$ \, {(n)}_q := {q^n - 1 \over q - 1} \; (\in k[q]) \, $, $ \, {(n)}_q! := \prod_{r=1}^n {(r)}_q $,
$ {({n \atop s})}_q := {{(n)}_q! \over {(s)}_q! {(n-s)}_q! } \; (\in k[q]) \, $, and $ \, {[n]}_q :=
{q^n - q^{-n} \over q - \qm} \; (\in \kqqm) \, $, $ \, {[n]}_q! := \prod_{r=1}^n {[r]}_q $,
$ {[{n \atop s}]}_q := $\break
$ := {{[n]}_q! \over {[s]}_q! {[n-s]}_q! } \; (\in \kqqm) \, $; let
$ \, q_\alpha := q^{d_\alpha} \, $ for all $ \, \alpha \in R^+ \, $, and $ \, q_i :=
q_{\alpha_i} \, $. Let $ Q $, $ P $ be as in \S 1; we fix an endomorphism $ \varphi $
of the $ \Q $-vector space $ \, \Q P := \Q \otimes_\Z P \, $ which is antisymmetric --- with respect to $ (\ \mid \ ) $ --- and satisfies the conditions
\hfill\break
\centerline{ $ \varphi(Q) \subseteq Q \, , \quad {1 \over 2} (\varphi(P) \mid P) \subseteq \Z \, , \quad 2 A Y A^{-1} \in M_n(\Z) $ }
where, letting $ \, \tau_i := {1 \over 2} \varphi(\alpha_i)
= \sum_{j=1}^n y_{ji} \alpha_j \, $, we set $ \, Y :=
{(y_{ij})}_{i,j= \unon;} \, $, $ M_n(\Z) $ is the set of
$ n \times n $-matrices with integer entries, and $ A $
is our Cartan matrix. For later use, we also define
$$ \tau_\alpha := {1 \over 2} \varphi(\alpha) \, , \quad
\tau_i := \tau_{\alpha_i} \eqno (2.1) $$
for all $ \, \alpha \in R \, $, $ \, i= \unon \, $. It is
proved in [CV-1] that $ \, (\text{id}_{\Q P} + \varphi)
\, $ and $ \, (\text{id}_{\Q P} - \varphi) \, $ are isomorphisms, adjoint of each other with respect to $ \, (\ \vert \ ) \, $: then we define $ \, r := {(\text{id}_{\Q P} + \varphi)}^{-1} \, , \quad \rbar := {(\text{id}_{\Q P} - \varphi)}^{-1} \, $.
\vskip7pt
{\bf 2.2 Quantum Borel algebras.} \ Let $ M $ be any lattice such that $ \, Q \leq
M \leq P \, $; then (cf.~[DL] and [CV-1]) $ U_{q,\varphi}^M (\gerb_-) $ (resp.
$ U_{q,\varphi}^M (\gerb_+) $) is the associative \hbox{$\kq$-alge}bra with 1
generated by \ $ \{\, L_\mu \mid \mu \in M \,\} \cup \{\, F_i \mid i=\unon \,\} $ \
(resp. \ $ \{\, L_\mu \mid \mu \in M \,\} \cup \{\, E_i \mid i=\unon \,\} $~) \ with
relations
$$ \eqalign{
L_0 = 1 \, , & \quad L_\mu L_\nu = L_{\mu+\nu}
\, , \cr
L_\mu F_j = q^{-( \mu | \alpha_j )} F_j L_\mu \, , &
\quad \sum_{p+s=1-\aij} {(-1)}^s {\left[ {1 - \aij \atop s} \right]}_{q_i} F_i^p F_j F_i^s = 0 \cr
\hbox{(resp. \ \ } L_\mu E_j = q^{( \mu | \alpha_j )}
E_j L_\mu \, , & \quad \sum_{p+s=1-\aij} {(-1)}^s {\left[ {1 - \aij \atop s} \right]}_{q_i} E_i^p E_j E_i^s = 0 \, \hbox{)}
\cr } $$
for all $ \, i, j= \unon $ and $ \, \mu, \nu \in M \, $;
furthermore, it is proved in [CV-1] that $ U_{q,\varphi}^M
(\gerb_-) $ and $ U_{q,\varphi}^M (\gerb_-) $ can be
endowed with a Hopf algebra structure
given by\footnote{ Actually, we use here a different
normalization than in [CV-1]: the results therein coincide
with the ones we list below up to change $ \, q
\leftrightarrow \qm \, $, $ \, L_\lambda \leftrightarrow
L_{-\lambda} \, $ (or $ \, K_\lambda \leftrightarrow K_{-
\lambda} \, $). }
$$ \eqalign {
\Delta_\varphi (F_i) = F_i \otimes L_{-\alpha_i - \tau_i}
+ L_{\tau_i} \otimes F_i \, , & \quad
\varepsilon_\varphi (F_i) = 0 \, , \quad S_\varphi (F_i) =
- F_i L_{\alpha_i} \cr
\Delta_\varphi (L_\mu) = L_\mu \otimes L_\mu \, , &
\quad \varepsilon_\varphi (L_\mu) = 1 \, ,
\quad S_\varphi (L_\mu) = L_{-\mu} \cr
\Delta_\varphi (E_i) = E_i \otimes L_{\tau_i} +
L_{\alpha_i - \tau_i} \otimes E_i \, , & \quad
\varepsilon_\varphi (E_i) = 0 \, , \quad S_\varphi (E_i) =
- L_{-\alpha_i} E_i \cr } $$
for all $ \, i=\unon \, $, $ \, \mu \in M \, $. In
particular we use notation $ K_\alpha := L_\alpha \;
\forall \, \alpha \in Q \, $. We shall also consider the
subalgebras (of quantum Borel algebras)
$ U_{q,\varphi}^M (\gert) $ (generated by the $ L_\mu $'s),
$ U_{q,\varphi} (\gern_+) $ (generated by the $ E_i $'s), and
$ U_{q,\varphi}(\gern_-) $ (generated by the $ F_i $'s).
\par
When $ \, \varphi = 0 \, $ we could skip the
superscript $ \varphi $, our quantum algebras then
conciding with the one-parameter ones, i.~e.~those of, say,
[Lu-1], [DP], etc.
\vskip7pt
{\bf 2.3 DRT pairings.} \ From now on, if $ H $
is any Hopf algebra, then $ H^{op} $ will denote the same coalgebra as $ H $ with the opposite multiplication, while $ H_{op} $ will denote the same algebra as $ H $ with the opposite comultiplication.
\par
From [DL], \S 2 (for the 1-parameter case), and [CV-1],
\S 3 (for the multiparameter case) we recall the existence
of perfect (i.~e.~non-degenerate) pairings of Hopf algebras
among quantum Borel algebras
$$ \eqalign{
\pi_-^\varphi \colon {\uqphiPbm}_{op} \otimes \uqphiQbp \rightarrow \kq \, , & \quad \pi_+^\varphi \colon {\uqphiQbm}_{op} \otimes \uqphiPbp \rightarrow \kq \cr
\overline{\pi_-^\varphi} \colon \uqphiPbp_{op} \otimes \uqphiQbm \kq \, , & \quad \overline{\pi_+^\varphi} \colon \uqphiQbp_{op} \otimes \uqphiPbm \kq \cr } $$
which we select as given by
$$ \eqalign{
\pi_-^\varphi (L_\lambda,K_\alpha) = q^{-(r(\lambda)
\vert \alpha)} \, , \ \pi_-^\varphi (L_\lambda,E_j) = 0 \,
, & \ \pi_-^\varphi (F_i,K_\alpha) = 0 \, , \ \pi_-
^\varphi (F_i,E_j) = \delta_{ij} {q^{-(r(\tau_i) \vert
\tau_i)} \over (q_i^{-1} - q_i)} \cr
\pi_+^\varphi (K_\alpha,L_\lambda) = q^{-(r(\alpha) \vert
\lambda)} \, , \ \pi_+^\varphi (K_\alpha,E_j) = 0 \, ,
& \ \pi_+^\varphi (F_i,L_\lambda) = 0 \, , \
\pi_+^\varphi (F_i,E_j) = \delta_{ij} {q^{-(r(\tau_i) \vert
\tau_i)} \over (q_i^{-1} - q_i)} \cr
\overline{\pi_-^\varphi} (L_\lambda,K_\alpha) =
q^{(r(\lambda) \vert \alpha)} \, , \ \overline{\pi_-
^\varphi} (E_i,K_\alpha) = 0 \, , & \ \overline{\pi_-^\varphi} (L_\lambda,F_j) = 0 \, , \ \overline{\pi_-^\varphi} (E_i,F_j) = \delta_{ij} {q^{+(r(\tau_i) \vert \tau_i)} \over (q_i - q_i^{-1})} \cr
\overline{\pi_+^\varphi} (K_\alpha,L_\lambda) =
q^{(r(\alpha) \vert \lambda)} \, , \ \overline{\pi_+^\varphi} (E_i,K_\alpha) = 0 \, , & \ \overline{\pi_+^\varphi} (L_\lambda,F_j) = 0 \, , \ \overline{\pi_+^\varphi} (E_i,F_j) = \delta_{ij} {q^{+(r(\tau_i) \vert \tau_i)} \over (q_i - q_i^{-1})} \cr
\forall \, \; i, j = \unon \, , & \; \alpha \in Q \, , \ \lambda \in P \; . \cr } $$
\indent {\it Remark :\/} Hopf pairings like those
above were introduced by Drinfeld, Rosso, Tanisaki, and others, whence we shall call them {\it DRT pairings\/}; here again some differences occur with respect to [CV-1], [CV-2] (or even [DL] for the simplest case), because of different definition of the Hopf structure: cf.~also [Tn] and [DD].
\vskip7pt
In the sequel if $ \pi $ is any DRT pairing we will
also set $ \, {\langle x, y \rangle}_\pi \, $ for
$ \, \pi(x,y) \, $.
\vskip7pt
{\bf 2.4 PBW bases.} \ It is known that both
$ \uqphiQbm $ and $ \uqphiPbp $ have bases of
Poincar\'e-Birkhoff-Witt type (in short "PBW bases"): we fix
(once and for all) any reduced expression of $ w_0 $
(the longest element in the Weyl group $ W $), namely $
w_0 = s_{i_1} s_{i_2} \cdots s_{i_N} $ (with $ N:= \#(R^+)
= $ number of positive roots); thus we have a total convex
ordering (cf.~[Pa] and [DP], \S 8.2) $ \, \alpha^1,
\alpha^2 := s_{i_1} (\alpha_{i_2}), \dots, \alpha^N :=
s_{i_1} s_{i_2} \cdots s_{i_{N-1}} (\alpha_{i_N}) \, $,
hence --- following Lusztig and others: cf.~e.~g.~[Lu-2] --- we can construct root vectors $ E_{\alpha^r} $,
$ r=1, \dots, N \, $ as in [DP] or [CV-1] and get PBW bases of {\it increasing\/} ordered monomials \ $ \left\{\, L_\mu \cdot \prod_{r=1}^N F_{\alpha^r}^{f_r} \, \big\vert \, \mu \in M ; f_1, \dots, f_N \in \N \,\right\} $ \ for $ \, U_{q,\varphi}^M (\gerb_-) \, $ and \ $ \big\{\, L_\mu \cdot \prod_{r=1}^N E_{\alpha^r}^{e_r} \, \big\vert \, \mu \in M; e_1, \dots, e_N \in \N \,\big\} $ \ for $ \,
U_{q,\varphi}^M (\gerb_+) \, $ or similar PBW bases of {\it decreasing\/} ordered monomials.
\par
Now, for every monomial $ {\Cal E} $ in the $ E_i $'s,
define $ \, s({\Cal E}) := {1 \over 2} \varphi(wt({\Cal E})) $, $ r({\Cal E}) := {1 \over 2} r(\varphi(wt({\Cal E}))) $, $ \, \rbar({\Cal E}) := {1 \over 2} \rbar(\varphi(wt({\Cal E}))) \, $, where $ \, wt({\Cal E}) \, $ denotes the weight of $ \, {\Cal E} \, $ ($ E_i $ having weight $ \alpha_i \, $), and similarly for every monomial $ {\Cal F} $ in the $ F_i $'s, ($ F_i $ having weight $ -\alpha_i \, $). Then the values of DRT pairings on PBW monomials are given by
$$ \eqalign {
\pi_-^\varphi \Bigg(\prod_{r=N}^1 F_{\alpha^r}^{f_r} \cdot L_\lambda , \prod_{r=N}^1 E_{\alpha^r}^{e_r} \cdot K_\alpha \Bigg) = \hskip1truecm \qquad {\ } & \cr
= q^{-\left( r(\lambda) - r \left( \prod_{r=N}^1 F_{\alpha^r}^{f_r} \right) \Big\vert \alpha - s \left( \prod_{r=N}^1 E_{\alpha^r}^{e_r} \right) \right)} &
\prod_{r=1}^N \delta_{e_r,f_r} {{[e_r]}_{q_{\alpha^r}}\! ! \, \, q_{\alpha^r}^{+{e_r \choose 2}} \over {(q_{\alpha^r}^{-1} - q_{\alpha^r})}^{e_r}} \cr
\pi_+^\varphi \Bigg(\prod_{r=N}^1 F_{\alpha^r}^{f_r} \cdot K_\alpha , \prod_{r=N}^1 E_{\alpha^r}^{e_r} \cdot L_\lambda \Bigg) = \hskip1truecm \qquad {\ } & \cr
= q^{-\left( r(\alpha) - r \left( \prod_{r=N}^1 F_{\alpha^r}^{f_r} \right) \Big\vert \lambda - s \left( \prod_{r=N}^1 E_{\alpha^r}^{e_r} \right) \right)} & \prod_{r=1}^N \delta_{e_r,f_r} {{[e_r]}_{q_{\alpha^r}}\! ! \, \, q_{\alpha^r}^{+{e_r \choose 2}} \over {(q_{\alpha^r}^{-1} - q_{\alpha^r})}^{e_r}} \cr }
\eqno (2.2-a) $$
$$ \eqalign{
\overline{\pi_-^\varphi} \Bigg(\prod_{r=N}^1 L_\lambda
\cdot E_{\alpha^r}^{e_r} , K_\alpha \cdot \prod_{r=N}^1
F_{\alpha^r}^{f_r} \Bigg) = \hskip1truecm \qquad {\ } & \cr
= q^{\left(\rbar(\alpha) - \rbar \left( \prod_{r=N}^1
F_{\alpha^r}^{f_r} \right) \Big\vert \lambda - s \left( \prod_{r=N}^1 E_{\alpha^r}^{e_r} \right) \right)} & \prod_{r=1}^N \delta_{e_r,f_r} {{[e_r]}_{q_{\alpha^r}}\! ! \, \, q_{\alpha^r}^{-{e_r \choose 2}} \over {(q_{\alpha^r} - q_{\alpha^r}^{-1})}^{e_r}} \cr
\overline{\pi_+^\varphi} \Bigg( \prod_{r=N}^1 K_\alpha \cdot
E_{\alpha^r}^{e_r} , L_\lambda \cdot \prod_{r=N}^1
F_{\alpha^r}^{f_r} \Bigg) = \hskip1truecm \qquad {\ } & \cr
= q^{\left( r(\lambda) - r \left( \prod_{r=N}^1 F_{\alpha^r}^{f_r} \right) \Big\vert \alpha - s \left( \prod_{r=N}^1 E_{\alpha^r}^{e_r} \right) \right)} & \prod_{r=1}^N \delta_{e_r,f_r} {{[e_r]}_{q_{\alpha^r}}\! ! \, \, q_{\alpha^r}^{-{e_r \choose 2}} \over {(q_{\alpha^r} -
q_{\alpha^r}^{-1})}^{e_r}} \cr } \eqno (2.2-b) $$
(cf.~[CV-2] \S 1, taking care of our different normalizations; see also [DD] and [DL]).
\vskip7pt
{\bf 2.5 Integer forms and duality.} \ Let
$ \uqphiQbmhat $ be the \hbox{$ \kqqm $-subalge}bra of
$ \uqphiQbm $ generated by
$$ \left\{\, F_i^{(m)}, \left( K_i; c \atop t \right),
K_i^{-1} \, \bigg\vert \, m,c,t \in \N; i=\unon
\,\right\} $$
where $ \, F_i^{(m)} := F_i^m \big/ {{[m]}_{q_i}\! !} \, $
and $ \, \left(K_i; c \atop t \right) := \prod_{s=1}^t
{{K_i q_i^{c-s+1} - 1} \over {q_i^s - 1}} \, $ are the so-called "divided powers"; it is known (cf.~[Lu-2], [DL], [CV-2]) that $ \uqphiQbmhat $ is a Hopf subalgebra of $ \uqphiQbm $, having a PBW basis (as a \hbox{$\kqqm$-modu}le) of increasing ordered monomials
$$ \left\{\, \prod_{i=1}^n \left( K_i; 0 \atop t_i
\right) K_i^{-Ent({t_i/2})} \cdot \prod_{r=1}^N
F_{\alpha^r}^{(n_r)} \, \bigg\vert \, t_1, \dots, t_n, n_1,
\dots, n_N \in \N \,\right\} $$
and a similar PBW basis of decreasing ordered monomials; in particular $ \uqphiQbmhat $ is a \hbox{$ \kqqm $-form} of $ \uqphiQbm $. A similar definition gives us the \hbox{$ \kqqm $-subalge}bra $ \widehat{U}_{q,\varphi}^P (\gerb_+) $ of $ \uqphiPbp $ generated by divided powers in the $ E_i $'s and $ L_i $'s, which is an integer form of $ \uqphiPbp $ with a PBW basis of decreasing ordered monomials and a PBW basis of increasing ordered monomials. With the same procedure one defines Hopf algebras
$ \uqphiQbphat $ and $ \widehat{U}_{q,\varphi}^P (\gerb_-) \, $ (over $ \kqqm \, $) and locates PBW bases for them.
\par
Now let $ \, \uqphiPbptilde \, $ be the \hbox{$ \kqqm
$-subalge}bra of $ \uqphiPbp $ generated by
$$ \{ \ebar_{\alpha^1}, \dots, \ebar_{\alpha^N} \} \cup
\{\, L_1^\pm, \dots , L_n^\pm \,\} $$
where $ \, \ebar_{\alpha^r} := (q_{\alpha^r} -
q_{\alpha^r}^{-1}) E_{\alpha^r} \, $, $ \forall \, r=1,
\dots, N \, $; then (cf.~[DKP], [DP]) $ \uqphiPbptilde $ is a Hopf subalgebra of $ \uqphiPbp $, having a PBW basis (as a \hbox{$\kqqm$-modu}le)
$$ \left\{\, \prod_{i=1}^n L_i^{t_i} \cdot \prod_{r=1}^N
\ebar_{\alpha^r}^{n_r} \, \bigg\vert \, t_1,
\dots, t_n \in \Z ; n_1, \dots, n_N \in \N \,\right\} $$
of increasing ordered monomials and a similar PBW basis of decreasing ordered monomials; in particular
$ \uqphiPbptilde $ is a \hbox{$ \kqqm $-form} of
$ \uqphiPbp \, $. Similarly we define $ \, \uqphiQbmtilde \, $ to be the \hbox{$ \kqqm $-subalge}bra of
$ \uqphiQbm $ generated by $ \, \{ \fbar_{\alpha^1}, \dots, \fbar_{\alpha^N} \} \cup \{\, K_1^\pm, \dots , K_n^\pm \,\} \, $ (with $ \, \fbar_{\alpha^r} := (q_{\alpha^r} - q_{\alpha^r}^{-1}) F_{\alpha^r} \, $,
$ \forall \, r=1, \dots, N \, $), which
is an integer form of $ \uqphiQbm $, having
PBW $\kqqm$-bases of decreasing or increasing ordered monomials. With the same procedure one defines Hopf algebras $ \widetilde{U}_{q,\varphi}^Q (\gerb_+) $ and
$ \uqphiPbmtilde $ (over $ \kqqm \, $) and locates PBW bases for them.
\par
Similar constructions (and corresponding notations) and results also hold for the algebras $ U_{q,\varphi}^M (\gert) $, $ U_{q,\varphi}(\gern_+) $, and
$ U_{q,\varphi} (\gern_-) $.
\par
Finally, from the very definitions and from (2.2) one
immediately gets (as in [DL] \S 3)
$$ \eqalign {
\uqphiQthat & = \{\, y \in \uqphiQt \mid
\pi_-^\varphi (\uqphiPttilde,y) \leq \kqqm \,\} = \cr
& = \{\, x \in \uqphiQt \mid
\pi_+^\varphi (x,\uqphiPttilde) \leq \kqqm \,\} \cr
\uqphiPttilde & = \{\, y \in \uqphiPt \mid
\pi_+^\varphi (\uqphiQthat,y) \leq \kqqm \,\} = \cr
& = \{\, x \in \uqphiPt \mid
\pi_-^\varphi (x,\uqphiQthat) \leq \kqqm \,\} \cr
{\uqphiQbmhat}_{op} & = \{\, x \in
{\uqphiQbm}_{\text{op}} \mid \pi_+^\varphi (x,\uqphiPbptilde) \leq \kqqm \,\} \cr
{\uqphiPbmtilde}_{op} & = \{\, x \in
{\uqphiPbm}_{op} \mid \pi_-^\varphi (x,\uqphiQbphat) \leq
\kqqm \,\} \cr
{\uqphiQbphat}_{\text{op}} & = \{\, x \in
{\uqphiQbp}_{op} \mid \overline{\pi_+^\varphi}
(x,\uqphiPbmtilde) \leq \kqqm \,\} \cr
{\uqphiPbptilde}_{op} & = \{\, x \in
{\uqphiPbp}_{op} \mid \overline{\pi_-^\varphi}
(x,\uqphiQbmhat) \leq \kqqm \,\} \cr
\uqphiPbptilde & = \{\, y \in \uqphiPbp \mid
\pi_+^\varphi ({\uqphiQbmhat}_{op},y) \leq \kqqm \,\} \cr
\uqphiQbphat & = \{\, y \in \uqphiQbp \mid
\pi_-^\varphi ({\uqphiPbmtilde}_{op},y) \leq \kqqm \,\} \cr
\uqphiPbmtilde & = \{\, y \in \uqphiPbm \mid
\overline{\pi_+^\varphi} ({\uqphiQbphat}_{op},y) \leq \kqqm
\,\} \cr
\uqphiQbmhat & = \{\, y \in \uqphiQbm \mid
\overline{\pi_-^\varphi} ({\uqphiPbptilde}_{op},y) \leq
\kqqm \,\}. \cr } \eqno (2.3) $$
\vskip2truecm
\centerline{ \bf \S~3 Quantization of $ \ugtau $ }
\vskip10pt
{\bf 3.1 Drinfeld's double.} \ Let $ H_- $, $ H_+ $
be two arbitrary Hopf algebras on the ground field (or ring)
$ F $, and let $ \, \pi \colon {(H_-)}_{op} \otimes H_+ \rightarrow F \, $ be any arbitrary Hopf pairing. The {\it Drinfeld's double \/} $ \, D = D(H_-,H_+,\pi) \, $ is the algebra $ \, T(H_- \oplus H_+) \big/ {\Cal R} \, $, where $ {\Cal R} $ is the ideal of relations
$$ \eqalign{
1_{H_-} = 1 = 1_{H_+} \quad , & \qquad x \otimes y
= x y \qquad \qquad \quad \forall \, x, y \in H_+ \; \hbox{ or } \; x, y \in H_- \cr
\sum_{(x),(y)} \pi(y_{(2)},x_{(2)}) \, x_{(1)} \otimes
y_{(1)} = \! & \sum_{(x),(y)} \! \! \pi(y_{(1)},x_{(1)})
\, y_{(2)} \otimes x_{(2)} \qquad \forall \, x
\in H_+ , \, y \in H_- \, . \cr } $$
\indent Then (cf.~[DL], Theorem 3.6) $ D $ has a
canonical structure of Hopf algebra such that $ \, H_- $,
$ H_+ \, $ are Hopf subalgebras of it and multiplication
yields isomorphisms of coalgebras
$$ H_+ \otimes H_- \hookrightarrow D \otimes D @>{m}>> D \, , \quad H_- \otimes H_+ \hookrightarrow D \otimes D @>{m}>> D \, . \eqno (3.1) $$
\indent We apply this to get the Drinfeld's double
$ \, D_{q,\varphi}^M(\gerg) := D(\uqphiQbm, U_{q,\varphi}^M (\gerb_+), \pi_+^\varphi) \, $ (for any lattice $ M $, with $ Q \leq M \leq P $) which we call {\it quantum double\/}; from the very definition, $ D_{q,\varphi}^M (\gerg) $ is generated by $ \, K_\alpha $, $ L_\mu $,
$ F_i $, $ E_i $ --- identified with $ 1 \otimes K_\alpha $, $ L_\mu \otimes 1 $, $ 1 \otimes F_i $,
$ E_i \otimes 1 $ when thinking at $ \, D_{q,\varphi}^M (\gerg) \cong U_{q,\varphi}^M (\gerb_+) \otimes \uqphiQbm
\, $ --- ($ \alpha \in Q $, $ \mu \in M $, $ i =
\unon $), while relations defining the ideal $ {\Cal R} $ clearly reduce to commutation relations between generators, namely (for all $ \, i, j=\unon \, $)
$$ \eqalign{
K_\alpha L_\mu = L_\mu K_\alpha \; , \qquad K_\alpha E_j = & q_i^{(\alpha \vert \alpha_j)} E_j K_\alpha \; , \qquad L_\mu F_j = q_i^{-(\mu \vert \alpha_j)} F_j L_\mu \; , \cr
E_i F_j - F_j E_i = & \delta_{ij} {{L_{\alpha_i} -
K_{-\alpha_i}} \over {q_i - q_i^{-1}}} \ . \cr } \eqno (3.2) $$
\indent For later use we also record the following (deduced from (3.2))
$$ \eqalign {
E_i^r F_i^s = \sum_{t \geq 0}^{t \leq r,s} {\left[ r
\atop t \right]}_{q_i} {\left[ s \atop t \right]}_{q_i}
{{[t]}_{q_i} \! !}^2 \cdot F_i^{s-t} \cdot \left[
{K_{\alpha_i}}^\otimes_\otimes ; 2t - r - s \atop t \right]
\cdot E_i^{r-t} \cr
F_i^s E_i^r = \sum_{t \geq 0}^{t \leq r,s} {\left[ r \atop
t \right]}_{q_i} {\left[ s \atop t \right]}_{q_i}
{{[t]}_{q_i} \! !}^2 \cdot \cdot E_i^{r-t} \cdot \left[
{K_{-\alpha_i}}^\otimes_\otimes ; 2t - r - s \atop t \right]
\cdot F_i^{s-t} \cr } \eqno (3.3) $$
($ \forall \, i=\unon \, $) where $ \, \left[
{K_{\alpha_i}}^\otimes_\otimes ; c \atop t \right] := \prod_{p=1}^t {q_i^{c-p+1} \cdot
L_{\alpha_i} - K_{-\alpha_i} \cdot q_i^{-c+p-1} \over q_i^p - q_i^{-p}} \, $ and
$ \, \left[ {K_{-\alpha_i}}^\otimes_\otimes ; c \atop t \right] := $\break
$ := \prod_{p=1}^t {q_i^{c-p+1} \cdot K_{-\alpha_i} - L_{\alpha_i} \cdot q_i^{-c+p-1}
\over q_i^p - q_i^{-p}} \, $ for all $ \, c \in \Z $, $ \, t \in \N $.
\par
Finally, PBW bases of quantum Borel algebras clearly
provide (tensor) PBW bases of $ D_{q,\varphi}^M(\gerg) $
(identified with $ \, U_{q,\varphi}^M(\gerb_+) \otimes \uqphiQbm \, $, as we shall
always do in the sequel).
\vskip7pt
{\bf 3.3 The quantum algebra $ \, U_{q,\varphi}^M
(\gerg) \, $.} \ Let $ \gerK^P_\varphi $ be the ideal of
$ D_{q,\varphi}^P(\gerg) $ generated by the elements $ \, K \otimes 1 - 1 \otimes K \, $, $ \, K \in \uqQt \, $;
$ \gerK^P_\varphi $ is in fact a Hopf ideal, whence $ \,
D_{q,\varphi}^P(\gerg) \big/ \gerK^P_\varphi \, $ is a Hopf
algebra; then the above presentation of
$ D_{q,\varphi}^P(\gerg) $ yields the following one of
$ \, \uqphiPg := D_{q,\varphi}^P(\gerg) \Big/ \gerK^P_\varphi \, $: it is the associative
\hbox{$\kq$-alge}bra with 1 given by generators
$$ F_i \, , \ L_\lambda \ \, \text{(or } K_\lambda
\text{)} \, , \ E_i \quad (\lambda \in P; i=\unon) $$
and relations
$$ \eqalign {
L_0 = 1 \, , \; L_\lambda L_\mu = L_{\lambda+\mu} = L_\mu L_\lambda \, ,
\; L_\lambda F_i & = q^{-(\alpha_j | \lambda)} F_i L_\lambda \, , \; L_\lambda E_i =
q^{(\alpha_j | \lambda)} E_i L_\lambda \cr
E_i F_h - F_h E_i & = \delta_{ih} {{L_{\alpha_i} -
L_{-\alpha_i}} \over {q_i - q_i^{-1}}} \cr
\sum_{k = 0}^{1-a_{ij}} \! (-1)^k \! \left[ \! { 1- a_{ij} \atop k } \! \right]_{q_i} \!\!
E_i^{1-\aij-k} E_j E_i^k = 0 , & \, \sum_{k = 0}^{1-a_{ij}} \! (-1)^k \! \left[ \! { 1-a_{ij}
\atop k } \! \right]_{q_i} \!\! F_i^{1-\aij-k} F_j F_i^k = 0 \cr } \eqno (3.4) $$
(for all $ \, \lambda \in P $, $ i, j, h = \unon \, $ with $ \, i \neq j \, $) with the Hopf
structure given by
$$ \eqalign{
\Delta_\varphi (F_i) = F_i \otimes L_{-\alpha_i -
\tau_i} + L_{\tau_i} \otimes F_i \, , & \quad
\varepsilon_\varphi(F_i) = 0 \, , \quad S_\varphi (F_i) =
- F_i L_{\alpha_i} \cr
\Delta_\varphi (L_\lambda) = L_\lambda \otimes
L_\lambda \, , & \quad \varepsilon_\varphi (L_\lambda) = 1
\, , \quad S_\varphi (L_\lambda) = L_{-\lambda} \cr
\Delta_\varphi (E_i) = E_i \otimes L_{\tau_i} +
L_{\alpha_i - \tau_i} \otimes F_i \, , & \quad
\varepsilon_\varphi (E_i) = 0 \, , \quad S_\varphi (F_i) =
- L_{-\alpha_i} E_i \cr } \eqno (3.5) $$
for all $ \, i=\unon $, $ \lambda \in P $. Similarly, with any lattice $ M $ ($ \, Q \leq M \leq P \, $) we define $ \, \uqphiMg := D_{q,\varphi}^M(\gerg) \Big/
\gerK^M_\varphi \, $ which has a presentation by generators and relations as in (10.4--5), with $ M $ instead of
$ P $; in particular we shall
study $ \uqphiQg $.\footnote{Notice that for $ \, \varphi = 0 \, $ one recovers the usual 1-parameter quantum enveloping algebras: thus $ U_{q,0}^M (\gerg) $ is nothing but $ U_{q,M}(\gerg) $ of [DP], \S 9, hence in particular for $ \, M=Q \, $ we recover the "quantum group of adjoint type" of Jimbo, Lustzig, and others.}
\par
Finally we shall denote with $ \, pr^M_\varphi \colon
D_{q,\varphi}^M(\gerg) \twoheadrightarrow D_{q,\varphi}^M
(\gerg) \Big/\gerK^M_\varphi = \uqphiMg \, $ the
canonical Hopf algebra epimorphism, and we shall use
notation $ \, K_\alpha := L_\alpha \; \forall \, \alpha \in
Q \, $.
\vskip7pt
{\bf 3.4 Integer forms of $ U_{q,\varphi}^M (\gerg) $.}
\ Let $ \uqphiQghat $ be the \hbox{$\kqqm $-subalge}bra
of $ \uqphiQg $ generated by
$$ \left\{\, F_i^{(\ell)}, \left( K_i; c \atop t \right),
K_i^{-1}, E_i^{(m)} \, \bigg\vert \, \ell,c,t,m \in \N;
i=\unon \,\right\} \; ; $$
then (cf.~[DL], \S 3) this is a Hopf subalgebra
of $ \uqphiQg $, with PBW basis (over $ \kqqm $)
$$ \left\{\, \prod_{r=N}^1 \! E_{\alpha^r}^{(n_r)} \!
\cdot \! \prod_{i=1}^n \! \left( K_i; 0 \atop t_i \right) \!
K_i^{-Ent({t_i/2})} \! \cdot \! \prod_{r=1}^N \!
F_{\alpha^r}^{(m_r)} \, \bigg\vert \, n_1, \dots, n_N, t_1,
\dots, t_n, m_1, \dots, m_N \in \N \,\right\} ; $$
this is also a $\kq$-basis of $ \uqphiQg $, hence
clearly $ \uqQghat $ is a \hbox{$\kqqm$-form} of
$ \uqphiQg \, $.
\par
Let $ \, \uqphiPgtilde \, $ be the $ \kqqm $-subalgebra of $ \uqphiPg $ generated
by
$$ \{ \fbar_{\alpha^1}, \dots, \fbar_{\alpha^N} \} \cup
\{\, L_1^\pm, \dots, L_n^\pm \,\} \cup \{ \fbar_{\alpha^1},
\dots, \ebar_{\alpha^N} \} $$
(cf.~[DP], \S 12; see also [DK] and [DKP]); then this is
a Hopf subalgebra of $ \uqphiPg $, having a PBW basis (as
a \hbox{$\kqqm$-modu}le)
$$ \left\{\, \prod_{r=N}^1 \ebar_{\alpha^r}^{n_r} \cdot
\prod_{i=1}^n L_i^{t_i} \cdot \prod_{r=1}^N
\fbar_{\alpha^r}^{m_r} \, \bigg\vert \, t_1, \dots, t_n \in
\Z ; n_1, \dots, n_N, m_1, \dots, m_N \in \N
\,\right\} ; $$
this is also a $\kq$-basis of $ \uqphiPg $, hence
clearly $ \uqphiPgtilde $ is a \hbox{$\kqqm$-form} of
$ \uqphiPg \, $.
\vskip7pt
{\bf 3.5 Specialization at roots of 1 and quantum
Frobenius morphisms.} \ Since the integer forms we
defined are modules over $ \kqqm $, we can specialize them
at special values of $ q $. To this end, we assume for
simplicity that our ground field $ k $ contains all roots
of unity (so that $ \, \kqqm \Big/ (q - \varepsilon) = k
\, $ for any root of unity $ \varepsilon $; otherwise,
specialization will extend the ground field to $ \, \kqqm
\Big/ p_\ell (q) = k(\varepsilon) \, $, where
$ p_\ell (q) $ is the $ \ell $-th cyclotomic polynomial and $ \varepsilon $ denotes a primitive $ \ell $-th
root of 1); notice that this is the case for $ \, k = \C
\, $.
\par
Let us begin with $ \uqphiQghat $. For the case $ \,
\varepsilon = 1 \, $, set
$$ \uunophiQghat := \, \uqphiQghat \Big/ (q-1) \uqQghat \, \cong \, \uqphiQghat \otimes_{\kqqm} k $$
(where $ k $ is meant as a $ \kqqm $-algebra via $ \, k
\cong \kqqm/(q-1) \, $); let $ \, p_1 \colon \uqphiQghat
\twoheadrightarrow \uunophiQghat \, $ be the canonical
projection and set $ \, f_i := p_1 \left( F_i^{(1)}
\right) $, $ h_i := p_1 \left( \left( K_i ; 0 \atop 1 \right) \right) $, $ e_i := p_1 \left( E_i^{(1)} \right) $, for all $ i=\unon $. Furthermore, observe that the algebras $ \uqphiQghat $ for various $ \varphi $ are all canonically isomorphic; in particular they have the same presentation ({\it as algebras\/}) by generators and relations as $ \, \uqQghat = \widehat{U}_{q,0}^Q (\gerg)
\, $, which is given in [DL], \S 3.4; then one immediately finds that $ \, \uunophiQghat \, $ is a Poisson Hopf coalgebra, with a presentation by generators ($ f_i, h_i, e_i $) and relations which is exactly the same of
$ \ugtau $ (with $ \tau $ given by (2.1)) given in \S 1, hence we have a Poisson Hopf coalgebra isomorphism
$$ \uunophiQghat \cong \ugtau \; ; \eqno (3.6) $$
in a word, $ \uqphiQghat $ specializes to
$ \ugtau $ for $ \, q \rightarrow 1 \, $: in symbols,
$ \, \uqphiQghat @>{q \rightarrow 1}>> \ugtau \, $.
\par
Let now $ \varepsilon $ be a primitive $ \ell $-th
root of 1, for $ \ell $ {\it odd\/}, $ \, \ell > d:=
\max_i {\{d_i\}}_i \, $, and set
$$ \uepsilonphiQghat := \, \uqphiQghat \Big/ (q -
\varepsilon) \uqphiQghat \, \cong \, \uqphiQghat \otimes_{\kqqm} k $$
(where $ k $ is meant as a $ \kqqm $-algebra via $ \, k
\cong \kqqm \Big/ (q-\varepsilon) \, $); from [CV-2], \S 3.2 (cf.~also [Lu-2], Theorem 8.10, and [DL], Theorem 6.3, for the one-parameter case) we record the existence of a special Hopf algebra epimorphism
$$ \frhatgergtau \colon \uepsilonphiQghat
\twoheadrightarrow \uunophiQghat \cong \ugtau
\eqno (3.7) $$
\indent When $ \, \varphi = 0 \, $ --- whence $ \,
\tau = 0 \, $ --- and $ \ell = p $ is prime, it is shown
in [Lu-2], \S 8.15, that
$ {\widehat{{\Cal F}r}}_{\gerg^0} $ can be regarded as a {\it lifting of the Frobenius morphism $ \, G_{\Z_p} \rightarrow G_{\Z_p} \, $ to characteristic zero\/}; for this reason, we shall refer to $ \frhatgergtau $ as a {\it quantum Frobenius morphism\/}.
\par
Now for $ \uqphiPgtilde $. Set
$$ \uunophiPgtilde := \, \uqphiPgtilde \Big/ (q-1)
\uqphiPgtilde \, \cong \, \uqphiPgtilde \otimes_{\kqqm} k
\; ; $$
it is known (cf.~[DP], Theorem 12.1, and [DKP], Theorem 7.6 and Remark 7.7 {\it (c)\/}\,) that
$$ \uunophiPgtilde \cong F[H^\tau] \eqno (3.8) $$
as Poisson Hopf algebras over $ k $ (here $ H^\tau $ is
the Poisson group defined in \S 1.5); in a word,
$ \uqphiPgtilde $ specializes to $ F[H^\tau] $ as $ \, q
\rightarrow 1 \, $, or $ \, \uqphiPgtilde @>{q \rightarrow 1}>> F[H^\tau] \, $.
\par
For $ \varepsilon $ a primitive $ \ell $-th root of 1
($ \ell $ {\it odd\/}, $ \, \ell > d:= \max_i \{d_i\}
\, $), we set
$$ \uepsilonphiPgtilde := \, \uqphiPgtilde \Big/ (q -
\varepsilon) \uqphiPgtilde \, \cong \, \uqphiPgtilde
\otimes_{\kqqm} k \; ; $$
from [DKP], \S\S 7.6--7 (cf.~also [DP] for the one-parameter case) we record the existence of a special Hopf algebra monomorphism
$$ \frtildegergtau \colon \fhtau \cong \uunophiPgtilde
\hookrightarrow \uepsilonphiPgtilde \eqno (3.9) $$
\indent Just as for $ \frhatgergtau $, we shall refer
to $ \frtildegergtau $ as a {\it quantum Frobenius
morphism\/}; in fact in case $ \, \varphi = 0 \, $ and
$ \ell = p $ prime it can be regarded as a {\it lifting of
the Frobenius morphism $ \, H_{\Z_p} \rightarrow H_{\Z_p}
\, $ to characteristic zero\footnote{ Here
$ H_{\Z_p} $ denotes the Chevalley-type group-scheme over
$ \Z_p $ that one can clearly build up from the "Cartan
datum" associated to $ H $ (mimicking the usual procedure
one follows for $ G_{\Z_p} \, $). }\/}.
\vskip2truecm
\centerline{ \bf \S~4 Quantum function algebras }
\vskip10pt
{\bf 4.1 The quantum function algebras $ F_{q,\varphi}^M[B_\pm] $
(cf.~[DL] \S 4.1, [CV-2] \S\S 2-3).} \ Let
$ F_{q,\varphi}^M [B_\pm] $ be the {\it quantum
function algebra\/} relative to $ U_{q,\varphi}^{M'}
(\gerb_\pm) $, defined as the algebra of matrix
coefficients of
{\it positive\/}\footnote{i.~e.~those for which there
exists a basis in which the operators $ K_\mu $
($ \mu \in M' $) act diagonally with
eigenvalues powers of $ q $.}
finite dimensional representations of $ U_{q,\varphi}^{M'}
(\gerb_\pm) $ (cf.~[DL], \S 4). Then $ F_{q,\varphi}^M [B_\pm] $
is a Hopf algebra, which we call dual of
$ U_{q,\varphi}^{M'} (\gerb_\pm) $ because there
exists a natural perfect Hopf pairing (evaluation) among
them; it is also possible to realize $ F_{q,\varphi}^M [B_\pm] $ as a Hopf
subalgebra of $ {U_{q,\varphi}^{M'} (\gerb_\pm)}^\circ $ (hereafter for any Hopf
algebra $ H $ we will denote with $ H^\circ $ ($ \subseteq H^* \,
$) its dual Hopf algebra, in the standard sense of [SW],
ch.~VI). Notice that, since multiparameter quantum enveloping algebras (with the
same weight lattice $ M $) are all isomorphic {\it as algebras\/}, then
multiparameter quantum function algebras are all isomorphic {\it as coalgebras\/}.
DRT pairings provide Hopf algebra isomorphisms
$$ \fqphiPbp \cong {\uqphiPbm}_{op} \, , \quad \; \fqphiQbp \cong {\uqphiQbm}_{op} \, , \quad \; \fqphiPbm \cong {\uqphiPbp}_{op} \eqno (4.1) $$
induced by the pairing $ \, \pi_-^\varphi \, $, resp.~$ \,
\pi_+^\varphi \, $, resp.~$ \, \overline{\pi_-^\varphi}
\, $ (cf.~[DL], Proposition 4.2, and [CV-2], \S 2.3); by
means of these, DRT pairings can be seen as natural
evaluation pairings.
\vskip7pt
{\bf 4.2 Integer forms of $ \, F_{q,\varphi}^M [B_\pm]
\, $.} \ In \S 2.5 we introduced integer forms of
$ U_{q,\varphi}^{M'} (\gerb_\pm) \, $: here we define
the corresponding function algebras, which will result to be
integer forms of $ F_{q,\varphi}^M [B_\pm] $. We set
$$ \eqalign{
\widehat{F}_{q,\varphi}^P [B_\pm] := \left\{\, f \in
F_{q,\varphi}^P [B_\pm] \,\Big\vert\, \left\langle f,
\widehat{U}_{q,\varphi}^Q (\gerb_\pm) \right\rangle \subseteq \kqqm \,\right\} \cr
\widetilde{F}_{q,\varphi}^Q [B_\pm] := \left\{\, f \in
F_{q,\varphi}^Q [B_\pm] \,\Big\vert\, \left\langle f,
\widetilde{U}_{q,\varphi}^P (\gerb_\pm) \right\rangle \subseteq \kqqm \,\right\} \cr } \eqno (4.2) $$
where \ $ \langle \ , \ \rangle \colon F_{q,\varphi}^M
[B_\pm] \otimes U_{q,\varphi}^{M'} (\gerb_\pm) \rightarrow
\kq $ \ is the natural evaluation pairing; then from
(2.2) and (4.1) one gets
$$ \widehat{F}_{q,\varphi}^P [B_\pm] \cong
{\widetilde{U}_{q,\varphi}^P (\gerb_\mp)}_{op} \, , \quad
\widetilde{F}_{q,\varphi}^Q [B_\pm] \cong
{\widehat{U}_{q,\varphi}^Q (\gerb_\mp)}_{op} \, ;
\eqno (4.3) $$
in particular $ \widehat{F}_{q,\varphi}^P [B_\pm] $ and
$ \widetilde{F}_{q,\varphi}^Q [B_\pm] $ are Hopf
\hbox{$\kqqm$-subal}gebras and integer forms (over
$ \kqqm $) of $ F_{q,\varphi}^P [B_\pm] $ and $ F_{q,\varphi}^P [B_\pm] $
respectively; we also have
$$ \widehat{F}_{q,\varphi}^P [B_\pm] \cong
{\widetilde{U}_{q,\varphi}^P (\gerb_\mp)}^{op} \, , \quad
\widetilde{F}_{q,\varphi}^Q [B_\pm] \cong
{\widehat{U}_{q,\varphi}^Q (\gerb_\mp)}^{op} \, ,
\eqno (4.4) $$
different isomorphisms arising from different kinds of
DRT pairing (cf.~also (2.3)).
\vskip7pt
{\bf 4.3 The quantum function algebra $ F_{q,\varphi}^M
[G] $ and its integer forms.} \ Like in \S 4.1, we let
$ F_{q,\varphi}^M [G] $ be the {\it quantum function algebra\/} relative to $ U_{q,\varphi}^{M'} (\gerg) $, defined as the algebra of matrix coefficients of {\it positive\/} finite dimensional representations of
$ U_{q,\varphi}^{M'} (\gerg) $ (cf.~[DL], \S 4, and [CV-2], \S 2.1); this is a Hopf algebra, which we call dual of
$ U_{q,\varphi}^{M'} (\gerg) $ because there exists a natural perfect Hopf pairing (evaluation) among them; in fact $ F_{q,\varphi}^M [G] $ is a Hopf subalgebra of
$ {U_{q,\varphi}^{M'} (\gerg)}^\circ $.
\par
As for integer forms, let
$$ \eqalign{
\fqphiPghat := \left\{\, f \in \fqphiPg \,\Big\vert\, \left\langle f, \uqphiQghat \right\rangle \subseteq \kqqm \,\right\} \cr
\fqphiQgtilde := \left\{\, f \in \fqphiQg \,\Big\vert\, \left\langle f, \uqphiPgtilde \right\rangle \subseteq \kqqm \,\right\} \cr } \eqno (4.5) $$
where \ $ \langle \ , \ \rangle \colon F_{q,\varphi}^M [G] \otimes U_{q,\varphi}^{M'} (\gerg) \rightarrow \kq $ \ is the natural evaluation pairing; these are Hopf subalgebras (over $ \kqqm $) of $ \fqphiPg $ and
$ \fqphiQg $.
\vskip7pt
{\bf 4.4 Specialization at roots of 1.} \ It is
shown in [CV-2] (generalizing [DL]) that
$$ \fqphiPghat \, @>{q \rightarrow 1}>> \, F[G^\tau]
\eqno (4.6) $$
i.~e.~$ \; \funophiPghat := \fqphiPghat \Big/ (q-1) \fqphiPghat \, \cong \, F[G^\tau] \; $ as Poisson Hopf \hbox{$k$-alge}bras; in fact this result arises as dual of $ \; \uqphiQghat \, @>{q \rightarrow 1}>> \, \ug \; $. For $ \varepsilon $ a primitive $ \ell $-th root of 1
($ \ell $ {\it odd\/}, $ \, \ell > d:= \max_i {\{d_i\}}_i \, $) in $ k $ (we make the assumptions of the very
beginning of \S 3.5), we set
$$ \fepsilonphiPghat := \, \fqphiPghat \Big/ (q - \varepsilon) \fqphiPghat \, \cong \, \fqphiPghat \otimes_{\kqqm} k $$
(where $ k $ is meant as a $ \kqqm $-algebra via $ \, k
\cong \kqqm \Big/ (q-\varepsilon) \, $); finally, from [CV-2], \S 3.3 (and also [DL], Proposition 6.4, for the one-parameter case) we record the existence of another {\it quantum Frobenius morphism\/}, namely a special Hopf algebra monomorphism
$$ \frhatGtau \colon \fgtau \cong \funophiPghat
\hookrightarrow \fepsilonphiPghat \eqno (4.7) $$
(dual of $ \, \frhatgergtau \colon \uepsilonphiQghat
\twoheadrightarrow \uunophiQghat \cong \ugtau \, $).
\vskip7pt
{\bf 4.5 Completions of $ \, F_{q,\varphi}^M [G] \, $.}
\ For later use, we also introduce some completions of quantum function algebras
$ F_{q,\varphi}^M [G] $. Let $ \, \gerE_\varphi := Ker \left( \varepsilon \colon
F_{q,\varphi}^M [G] \rightarrow \kq \right) \, $: we define $ F_{q,\varphi}^{M,\infty}
[G] $ to be the $ \gerE_\varphi $-adic completion of $ F_{q,\varphi}^M [G] $, with
its induced structure of topological Hopf algebra. The same construction work for integer forms: we define
$ \fqphiPinfghat $, resp.~$ \fqphiQinfgtilde $, to be the
$ \gerE_\varphi $-adic completion of $ \fqphiPghat $,
resp.~$ \fqphiQgtilde $, where $ \, \gerE_\varphi := Ker \left( \varepsilon \colon
\fqphiPghat \rightarrow \kqqm \right) \, $, resp.~$ \, \gerE_\varphi := Ker \left(
\varepsilon \colon \fqphiQgtilde \rightarrow \kqqm \right) \, $; both of these are
topological Hopf algebras over $ \kqqm $; finally, we define $ \fqphiQgtildeqinf \, $
to be the $ (q-1) $-adic completion of $ \fqphiQgtilde $ (which is naturally embedded into $ \fqphiQinfgtilde $), which also is a topological Hopf algebra on its own. We shall prove later that all these are integer forms --- in topological sense --- of the corresponding Hopf algebras: for this reason we call all these new objects --- the
$ F_{q,\varphi}^{M,\infty} [G] $ and their integer forms --- {\bf (multiparameter) quantum formal groups\/}.
\vskip2truecm
\centerline{ \bf \S~5 Quantum formal groups }
\vskip10pt
{\bf 5.1 Formal Hopf algebras and quantum formal groups.}
\ In this subsection we introduce the notion of {\it
quantum formal group\/}. Recall (cf.~[Di], ch.~I) that
formal groups can be defined in a category of a special type of commutative topological algebras, whose underlying vector space (or module) is linearly compact; following Drinfeld's philosophy,we define quantum formal groups by simply dropping out any commutativity assumption of the classical notion of formal group; thus we quickly outline how
to modify the latter in order to define our new quantum objects, following the axiomatic construction of [Di], ch.~I.
\par
To begin with, let $ E $ be any vector space over
a field $ K $ (one can then generalize more or less
wathever follows to the case of free modules over a ring),
and let $ E^* $ be its (linear) dual; we write $ \, \langle x^*, x \rangle \, $ for $ x^*(x) $ when $
\, x \in E \, $, $ \, x^* \in E^* \, $. We define on
$ E^* $ the topology $ \sigma(E^*,E) $ as the coarsest
one such that for each $ \, x \in E \, $ the linear map
$ \, x^* \mapsto \langle x^*, x \rangle \, $ of $ E^* $
into $ K $ is continuous, when $ K $ is given the
discrete topology. A fundamental system of neighborhoods of
$ 0 $ in $ E^* $ therefore consists of finite
intersections of hyperplanes defined by equations $ \,
\langle x^*, x_j \rangle = 0 \, $ for any finite sequence
$ {\{x_j\}}_j $ in $ E $; we can describe this topology by choosing a basis $ {\{e_i\}}_{i \in I} $ of $ E \, $: to each $ i \in I $ we associate the linear (coordinate) form $ e_i^* $ on $ E $ such that $ \, \langle e_i^*, e_j \rangle = \delta_{i j} \, $, and we say that the family $ {\{e^*_i\}}_{i \in I} $ is the {\it pseudobasis\/} of $ E^* $ dual to $ {\{e_i\}}_{i \in
I} $; then the subspace $ E' $ of $ E $ which is (algebraically) generated by the $ e_i^* $ is dense in
$ E^* $, and $ E^* $ is nothing but the {\it completion\/} of $ E' $, when $ E' $ is given the topology for which a fundamental system of neighborhoods of $ 0 $ consists of the vector subspaces containing almost all the $ e_i^* $; thus elements of $ E^* $ can be described by series in the $ e_i^* $'s which in the given topology are in fact convergent. Finally, the topological vector spaces $ E^* $ are characterized by the property of linear compactness.
\par
When dualizing twice, any linear form on $ E^* $ which
is continuous for the topology $ \sigma(E^*,E) $ is
necessarily of type $ \, x^* \mapsto \langle x^*, x \rangle
\, $ for a unique $ x \in E \, $, hence one gets back
$ E $ as topological dual of $ E^* $.
\par
If now $ E $, $ F $ are any two vector spaces over
$ K $, let $ E^* $, $ F^* $ be their dual spaces, with the
topologies $ \sigma(E^*,E) $, $ \sigma(F^*,F) $; for any
linear map $ \, u \colon E \rightarrow F \, $, its
transposed map $ \, u^* \colon F^* \rightarrow E^* \, $
is continuous, and conversely for any linear map $ \,
v \colon F^* \rightarrow E^* \, $ which is continuous there
exists a unique linear map $ \, u \colon E \rightarrow
F \, $ such that $ \, v = u^* \, $.
\par
When considering tensor products, $ E^* \otimes F^* \, $
is naturally identified to a subspace of $ {(E \otimes
F)}^* $ by the formula $ \, \langle x^* \otimes y^*, x
\otimes y \rangle = \langle x^*, x \rangle \cdot \langle
y^*, y \rangle \, $; this also shows that if
$ {\{e_i\}}_{i \in I} $ and $ {\{f_j\}}_{j \in J} $
are bases of $ E $ and $ F $, and $ {\{e^*_i\}}_{i \in I} $ and $ {\{f^*_j\}}_{j \in J} $
their dual pseudobases in $ E^* $ and $ F^* $, then $ {\{e^*_i \otimes f^*_j\}}_{i \in
I, j \in J} $ is the dual pseudobasis of $ {\{e_i \otimes f_j\}}_{i \in I, j \in J} $ in $ {(E
\otimes F)}^* $. This proves that $ {(E \otimes F)}^* $ is the completion of $ E^*
\otimes F^* $ for the tensor product topology, i.~e.~the topology of $ E^* \otimes
F^* $ for which a fundamental system of neighborhoods of $ 0 $ consists of the sets
$ \, E^* \otimes V + W \otimes F^* \, $ where $ V $, resp. $ W $, ranges in a
fundamental system of neighborhoods of $ 0 $ consisting in vector subspaces of
$ F^* $, resp. $ E^* $; we denote this completion by $ E^* \otimeshat F^* $, and we
call it the {\it completed\/} (or {\it topological\/}) {\it tensor product\/} of $ E^* $
and $ F^* $; the natural embedding of $ E^* \otimes F^* $ into $ {(E \otimes F)}^* =
E^* \otimeshat F^* $ is then obviously continuous. Finally, when $ \, u \colon E_1
\rightarrow E_2 \, $, $ \, v \colon F_1 \rightarrow F_2 \, $ are linear maps, the
transposed map $ \, {(u \otimes v)}^* \colon {(E_2 \otimes F_2)}^* = {E_2}^*
\otimeshat {F_2}^* \longrightarrow {(E_1 \otimes F_1)}^* = {E_1}^* \otimeshat
{F_1}^* \, $ coincides with the continuous extension to $ {E_2}^* \otimeshat {F_2}^*
$ of the continuous map $ \, u^* \otimes v^* \colon {E_2}^* \otimes {F_2}^*
\rightarrow {E_1}^* \otimes {F_1}^* \, $; thus it is also denoted by $ u^* \otimeshat
v^* \, $.
\par
We define a {\bf linearly compact algebra} to be a
topological algebra whose underlying vector space (or free
module) is linearly compact (this is exactly the same definition of [Di], ch.~I, \S 2.7,
but for the fact that we do not require commutativity): then linearly compact
algebras form a full subcategory of the category of topological algebras; morever, for
any two objects $ A_1 $ and $ A_2 $ in this category, their topological tensor
product $ A_1 \otimeshat A_2 $ is defined.
\par
Dually, within the category of linearly compact vector
spaces we define {\bf linearly compact coalgebras}
as triplets $ (C, \Delta, \varepsilon) $ with $ \, \Delta
\colon C \rightarrow C \otimeshat C \, $ and $ \,
\varepsilon \colon C \rightarrow K \, $ satisfying the
usual coalgebra axioms. Then one checks (with the same
arguments as [Di], which never need neither commutativity
nor cocommutativity) that $ \, {(\ )}^* \colon (A, m, 1)
\mapsto (A^*, m^*, 1^*) \, $ defines a contravariant
functor from algebras to linearly compact
coalgebras, while\break
$ {(\ )}^* \colon (C, \Delta, \varepsilon) \mapsto (C^*,
\Delta^*, \varepsilon^*) \, $ defines a contravariant
functor from coalgebras to linearly compact algebras.
\par
Finally, we define a {\bf formal Hopf algebra} as a
datum $ (H, m, 1, \Delta, \varepsilon, S) $ such that
$ (H, m, 1) $ is a linearly compact algebra,
$ (H, \Delta, \varepsilon) $ is a linearly compact
coalgebra, and the usual compatibility axioms of Hopf
algebras are satisfied. In the following, we shall also
consider "usual" Hopf algebras as particular formal Hopf
algebras.
\par
According to Drinfeld's dictionary, we define {\bf
quantum formal group} the {\it spectrum\/} of a formal
Hopf algebra (whereas {\it classical\/} formal groups are
spectra of {\it commutative\/} formal Hopf algebra:
cf.~[Di], ch.~I).
\vskip7pt
We will now use the language of quantum formal groups to
construct a suitable dual of Drinfeld's double:
\vskip7pt
\proclaim{Proposition 5.2} Let $ H_- $, $ H_+ $ be Hopf
$F$-algebras, let \ $ \, \pi \colon {(H_-)}_{op} \otimes H_+ \rightarrow F \, $ \ be an arbitrary Hopf pairing, and let $ \, D:= D(H_-,H_+,\pi) \, $ be the corresponding Drinfeld's double. Then there exist $F$-algebra isomorphisms
$$ D^* \cong {H_+}^* \otimeshat {H_-}^* \; , \quad D^*
\cong {H_-}^* \otimeshat {H_+}^* $$
dual of the $F$-coalgebra isomorphisms \ $ D \cong H_+
\otimes H_- \, , \; D \cong H_- \otimes H_+ $ \ provided
by multiplication (cf.~\S 3.1).
\endproclaim
\demo{Proof} It follows from the previous discussion that
dualizing (3.1) yields vector space isomorphisms
$$ D^* \cong {H_+}^* \otimeshat {H_-}^* \; , \quad D^*
\cong {H_-}^* \otimeshat {H_+}^* \; ; $$
then a straightforward computation shows that this is an
isomorphism of algebras, the right-hand-side being thought
of as a topological tensor product {\it of algebras\/}.
$ \square $
\enddemo
\vskip7pt
We apply all this to quantum groups. To begin with,
consider the Drinfeld's double $ D_{q,\varphi}^M(\gerg) $,
identified as a coalgebra with $ U_{q,\varphi}^M(\gerb_+) \otimes \uqphiQbm $; this is a Hopf algebra, whose dual is the formal Hopf algebra $ {D_{q,\varphi}^Q(\gerg)}^* $
isomorphic to $ {U_{q,\varphi}^M(\gerb_+)}^* \otimeshat {\uqphiQbm}^* $.
\par
From DRT pairings one gets natural Hopf algebra
embeddings $ U_{q,\varphi}^{P \atop Q} (\gerb_\mp)
\!\hookrightarrow\! U_{q,\varphi}^{Q \atop P} {(\gerb_\pm)}^* $; from these we get algebra embeddings
$ \, U_{q,\varphi}^{P \atop Q} (\gerb_-) \otimeshat
\uqphiPbp \hookrightarrow U_{q,\varphi}^{Q \atop P}
{(\gerb_+)}^* \otimeshat {\uqphiQbm}^* $, hence we will identify $ U_{q,\varphi}^{P \atop Q} (\gerb_-) \otimeshat \uqphiPbp $ with its image in $ U_{q,\varphi}^{Q \atop P} {(\gerb_+)}^* \otimeshat {\uqphiQbm}^* $.
\vskip7pt
\proclaim{Lemma 5.3} The algebra $ \, U_{q,\varphi}^{P
\atop Q} (\gerb_-) \otimeshat \uqPbp \, $ is a formal
Hopf subalgebra of\break
$ D_{q,\varphi}^{Q \atop P}{(\gerg)}^* = U_{q,\varphi}^{Q
\atop P} {(\gerb_+)}^* \otimeshat {\uqphiQbm}^* $.
\endproclaim
\demo{Proof} For the sake of simplicity we stick to $ \, \varphi = 0 \, $, the general case being entirely similar.
\par
It is known (cf.~[Lu-1], [APW], [Tk]) that the dual Hopf algebra of the quantized enveloping algebra $ \, U_{q,0}^Q (\gerg) = \uqQg \, $ is
$$ {\uqQg}^\circ \cong \fqPg \rtimes \kq[{\Z_2}^n] $$
where the symbol $ \, \rtimes \, $ denotes a semidirect
product of Hopf algebras and $ \, \kq[{\Z_2}^n] \, $ is the (Hopf) group algebra of the
multiplicative group $ \, {\Z_2}^n \, $ (with $ \, \Z_2 = \{-1, +1\} \, $), acting on
$ \, \uqQg \, $ by \ $ {(\nu_i)}_i (E_j) := 0 \, , \; {(\nu_i)}_i (K_j) := \nu_j \, ,
\; {(\nu_i)}_i (F_j) := 0 $ \ for all $ \, {(\nu_i)}_i \in {\Z_2}^n \, $. Similarly, letting
$ \, k^\star := k \setminus \{0\} \, $, one has
$$ \eqalign{
{\uqQt}^\circ & \cong \uqPt \rtimes \kq[{(k^\star)}^n] \cr
{\uqQbm}^\circ & \cong \uqPbp \rtimes \kq[{(k^\star)}^n] \cr
{\uqQbp}^\circ & \cong \uqPbm \rtimes \kq[{(k^\star)}^n] \cr } $$
(the multiplicative group $ \, {(k^\star)}^n \, $, acting on $ \, \uqQg \, $ by \
$ {(\nu_i)}_i (E_j) := 0 \, , \; {(\nu_i)}_i (K_j) := \nu_j \, , \; {(\nu_i)}_i (F_j) := 0 $ \ for
all $ \, {(\nu_i)}_i \in {(k^\star)}^n \, $), which in short hold because
\hfill\break
\centerline { $ \uqQt^\circ = {\kq[Q]}^\circ \cong \kq[X(Q)] = \kq[P \rtimes {(k^\star)}^n] = \uqPt \rtimes
\kq[{(k^\star)}^n] $ }
(where $ X(Q) $ is the group of characters $ \, Q
\rightarrow \kq^\star \, $, clearly $ \, X(Q) \cong
{(\kq^\star)}^n \cong P \rtimes {(k^\star)}^n \, $ with the
action of $ \, \nu = (\nu_1, \dots, \nu_n) \in {(k^\star)}^n \, $ given by $ \, \langle
\nu, K_i \rangle = \nu_i \, $, and we use the well-known fact $ \, {K[\Gamma]}^\circ
\cong K[X(\Gamma)] \, $ for all fields $ K $ and all commutative groups
$ \Gamma \, $), and quantum Borel algebras admit triangular decomposition. It is
immediate to check --- looking at PBW bases and using [Sw], Theorem 6.1.3, namely:
If $ H $ is a commutative Hopf algebra which is finitely generated (as an algebra),
then $ H^\circ $ is dense in $ H^* $ --- that $ \, {U_q^Q (\gerb_\pm)}^\circ \cong
U_q^P (\gerb_\mp) \rtimes \kq[{(k^\star)}^n] \, $ is dense in $ {U_q^Q
(\gerb_\pm)}^* $, so that \hfill\break
\centerline { $ {\uqQbp}^* \cong \uqPbm \widehat{\rtimes} \kq[{(k^\star)}^n] \, , \quad {\uqQbm}^* \cong \uqPbp \widehat{\rtimes} \kq[{(k^\star)}^n] $ }
(the symbols $ \widehat{\rtimes} $ denoting the obvious
formal series completion); it follows that
$$ {D_q^{Q \atop P}(\gerg)}^* \cong \left( \uqPbm
\widehat{\rtimes} \kq[{(k^\star)}^n] \right) \otimeshat \left( \uqPbp
\widehat{\rtimes} \kq[{(k^\star)}^n] \right) \, . $$
\indent Then looking at definitions one immediately
convinces himself that
$$ \displaylines{
\Delta \left( U_q^{P \atop Q} (\gerb_-) \otimeshat
\uqPbp \right) \subseteq \left( U_q^{P \atop Q} (\gerb_-) \otimeshat \uqPbp \right)
\otimeshat \left( U_q^{P \atop Q} (\gerb_-) \otimeshat \uqPbp \right) \cr
S \left( U_q^{P \atop Q} (\gerb_-) \otimeshat \uqPbp \right) = U_q^{P \atop Q}
(\gerb_-) \otimeshat \uqPbp \cr } $$
whence the thesis. $ \square $
\enddemo
\vskip7pt
In light of the previous lemma, the non-degenerate
canonical pairing\break
$ {D_{q,\varphi}^{Q \atop P}(\gerg)}^* \otimes D_{q,\varphi}^{Q \atop P} (\gerg) \rightarrow \kq \, $ given by evaluation restricts to a non-degenerate pairing
$$ \left(\! U_{q,\varphi}^{P \atop Q} (\gerb_-) \otimeshat \uqphiPbp \!\right) \!\otimes\! D_{q,\varphi}^{Q \atop P}(\gerg) = \left(\! U_{q,\varphi}^{P \atop Q} (\gerb_-) \otimeshat \uqphiPbp \!\right) \!\otimes\! \left(\! U_{q,\varphi}^{Q \atop P} (\gerb_+) \!\otimes\! \uqphiQbm \!\right) \rightarrow \kq $$
of formal Hopf algebras, which is defined by
$$ \langle f \otimes g, x \otimes y \rangle = {\langle f,
x \rangle}_{\pi_\mp^\varphi} \cdot {\langle g, y
\rangle}_{\overline{\pi_-^\varphi}} = \pi_\mp^\varphi (f,x)
\cdot \overline{\pi_-^\varphi} (g,y) $$
for all $ \, f \in U_{q,\varphi}^{P \atop Q} (\gerb_-) $,
$ g \in \uqphiPbp $, $ x \in U_{q,\varphi}^{Q \atop P}
(\gerb_+) $, $ y \in \uqphiQbm \, $.
\par
Now consider root vectors $ F_\alpha $, $ E_\alpha $
($ \, \alpha \in R^+ \, $) defined in \S 2.4: we set
$$ \eqalign{
F^\varphi_\alpha := K_{\tau_\alpha} F_\alpha =
F_\alpha K_{\tau_\alpha} \, , & \quad E^\varphi_\alpha :=
K_{\tau_\alpha} E_\alpha = E_\alpha K_{\tau_\alpha} \cr
\fbar^\varphi_\alpha := K_{\tau_\alpha} \fbar_\alpha =
(q_\alpha - q_\alpha^{-1}) F^\varphi_\alpha \, , & \quad
\ebar^\varphi_\alpha := K_{\tau_\alpha} \ebar_\alpha =
(q_\alpha - q_\alpha^{-1}) E^\varphi_\alpha \cr } $$
and $ \, E^\varphi_i := E^\varphi_{\alpha_i} \, $, $ \, F^\varphi_i := F^\varphi_{\alpha_i} \, $, $ \, \ebar^\varphi_i := \ebar^\varphi_{\alpha_i} \, $, $ \, \fbar^\varphi_i := \fbar^\varphi_{\alpha_i} \, $ for all
$ i $ (cf.~[CV-2], \S 1.7). From now on, we let $ M $ be any lattice such that $ \, Q \leq M \leq P \, $, and
$ M' $ its dual lattice.
\vskip7pt
\proclaim{Definition 5.4} \ We call $ \aphiMinf $ the closed topological subalgebra of $ \uqphiMbm \otimeshat
\uqphiPbp $ generated by\hfill\break
\centerline{ $ \, \{ \, F^\varphi_\alpha \otimes 1,
L_{-(1+\varphi)(\lambda)} \otimes L_{(1-\varphi)(\lambda)},
1 \otimes E^\varphi_\alpha \mid \alpha \in R^+; \lambda \in
M \,\} \, $ }
that is, $ \aphiMinf $ is the closure in $ \uqphiMbm \otimeshat \uqphiPbp $ of the subalgebra $ A^{\varphi,M} $ generated by \hfill\break
\centerline{ $ \, \{ \, F^\varphi_\alpha \otimes 1,
L_{-(1+\varphi)(\lambda)} \otimes L_{(1-\varphi)(\lambda)},
1 \otimes E^\varphi_\alpha \mid \alpha \in R^+; \lambda \in
M \,\} \, $. $ \square $ }
\endproclaim
\vskip7pt
\proclaim{Proposition 5.5} \ The algebra $ \aphiMinf $ is the subset
$ {\Cal A}^{\varphi,M}_\infty $ in $ \, \uqphiMbm \otimeshat \uqphiPbp $\break
$ \left( \subset {D_{q,\varphi}^{M'} (\gerg)}^* \right) \, $ of functions on
$ D_{q,\varphi}^{M'} (\gerg) $ which vanish on $ \gerK^{M'}_\varphi \, $. In
particular, $ \aphiMinf $ is a formal Hopf subalgebra of $ \, {U_{q,\varphi}^{M'}
(\gerb_+)}^* \otimeshat {\uqphiQbm}^* $.
\endproclaim
\demo{Proof} (cf.~[DL], \S 4.2, and [CV-2], Lemma 2.5)
Again we give for simplicity the proof for $ \, \varphi = 0
\, $, the general case arising exactly in the same way
putting $ \varphi $ throughout.
\par
From the very definition of $ \gerK^{M'} :=
\gerK^{M'}_0 $ (cf.~\S 3.3) it is clear that $ {\Cal A}^M_\infty := {\Cal A}^{0,M}_\infty $
is nothing but the subset of $ \, \phi \in \uqMbm \otimeshat \uqPbp \, $ enjoying the
property
$$ \phi(E L \otimes N F) = \phi(E L N \otimes F) = \phi(E
\otimes L N F) \; , \eqno (5.1) $$
for all $ \, E \in U_q(\gern_+) $, $ L, N \in U_q^{M'}(\gert) $, $ F \in U_q(\gern_-) $; thus we have to show that $ A^M_\infty $ is exactly the subset of functionals in $ \, \uqPbm \otimeshat \uqPbp $ enjoying property (5.1).
\par
First of all $ A^M_\infty $ have this property, for
any element of $ A^M_\infty $ is a linear combination of
elements $ F K_{-\lambda} \otimes K_\lambda E $ where
$ F $, $ E $ are monomials in the $ F_i $'s, $ E_i $'s;
thus we have
$$ \langle F K_{-\lambda} \otimes K_\lambda E , E' L
\otimes N F' \rangle = \pi_-(F,E') \cdot \pi_-(K_{-
\lambda},L) \cdot \overline{\pi_-}(K_\lambda,N) \cdot
\overline{\pi_-}(E,F') \eqno (5.2) $$
for all $ \, E' \in U_q(\gern_+) $, $ L, N \in U_q^{M'}(\gert) $, $ F' \in U_q(\gern_-) $. If we look at $ K_\lambda $, $ \lambda \in M $, as characters of the algebra $ U_q^{M'}(\gert) $, letting $ \, L(\lambda) = \pi_-(K_{-\lambda},L) \; \forall \, L \in U_q^{M'}(\gert)
\, $, then (5.2) reads
$$ \langle F K_{-\lambda} \otimes K_\lambda E , E' L
\otimes N F' \rangle = \pi_-(F,E') \, \overline{\pi_-}(E,F')
\, L(\lambda) \, N(\lambda) \; ; \eqno (5.3) $$
similarly we find
$$ \eqalignno{
\langle F K_{-\lambda} \otimes K_\lambda E , E'
\otimes L N F' \rangle & = \pi_-(F,E') \, \overline{\pi_-
}(E,F') \, 1(\lambda) \, LN(\lambda) & (5.4) \cr
\langle F K_{-\lambda} \otimes K_\lambda E , E' L N
\otimes F' \rangle & = \pi_-(F,E') \, \overline{\pi_-}(E,F') \, LN(\lambda) \, 1(\lambda) & (5.5) \cr } $$
and right-hand-side of (5.3--4--5) are clearly equal.
\par
On the other hand, any $ \, \phi \in \uqMbm
\otimes \uqPbp \, $ is a (possibly infinite) linear
combination of elements $ F K_{-\sigma} \otimes K_\lambda
E $; as above computation gives
$$ \displaylines{
\langle F K_{-\sigma} \otimes K_\lambda E , E' L
\otimes N F' \rangle = \pi_-(F,E') \, \overline{\pi_-}(E,F')
\, L(\sigma) \, N(\lambda) \cr
\langle F K_{-\sigma} \otimes K_\lambda E , E' \otimes
L N F' \rangle = \pi_-(F,E') \, \overline{\pi_-}(E,F') \,
LN(\lambda) \cr
\langle F K_{-\sigma} \otimes K_\lambda E , E' L N
\otimes F' \rangle = \pi_-(F,E') \, \overline{\pi_-}(E,F')
\, LN(\sigma) \cr } $$
thus (5.1) holds if and only if $ \sigma = \lambda $.
\par
Now consider the Hopf structure. Since $ \gerK^{M'} $ is a Hopf ideal we have
$ \, S(\gerK^{M'}) = \gerK^{M'} \, $, whence $ \, \langle S(f), \gerK^{M'} \rangle =
\langle f, S(\gerK^{M'}) \rangle = 0 \, $ for all $ \, f \in A^M_\infty \, $, so that $ \,
S(A^M_\infty) = A^M_\infty \, $. On the other hand, let $ f \in A^M_\infty $ with $ \,
\Delta(f) = \sum_n g_n \otimes h_n \in \left( \uqMbm \otimeshat \uqPbp \right)
\otimeshat $\break
$ \otimeshat \left( \uqMbm \otimeshat \uqPbp \right) = \left( \uqMbm
\otimes \uqPbp \right) \otimeshat \left( \uqMbm \otimes \uqPbp \right) \, $ (with $ n
$ ranging in a possibly infinite set of indices); we can take a similar expression so
that $ \, g_n \neq g_{n'} \, $ for $ n \neq n' \, $, and for each fixed factor $
g_{\overline n} $ there exists an $ x_{\overline n} \in \uqQbp \otimes \uqQbm $ such
that $ \, \langle g_n, x_{\overline n} \rangle \neq 0 \, $ if and only if $ n= {\overline
n} $; then for any fixed $ g_{\overline n} $ we have \hfill\break
\centerline { $ 0 = \langle f, x_{\overline n} \cdot y \rangle = \langle \Delta(f), x_{\overline n} \otimes y \rangle = \sum_n \langle g_n, x_{\overline n} \rangle \cdot \langle h_n, y \rangle = \langle g_{\overline n}, x_{\overline n} \rangle \cdot \langle h_{\overline n} , y \rangle $ }
for all $ \, y \in \gerK^{M'} \, $, whence $ \, h_{\overline n} \in {\Cal A}^M_\infty = A^M_\infty \, $; we conclude that $ h_n \in A^M_\infty $ for all $ n $, and similarly for $ g_n $, hence $ \, \Delta(A^M_\infty) \leq A^M_\infty \otimes A^M_\infty \, $. The thesis follows.
$ \square $
\enddemo
\vskip7pt
{\bf 5.6} {\it Remark:\/} notice that $ \aphiMinf $ can also be defined ---~as an {\it algebra\/}~--- as the
$ \gerE' $-adic completion of the subalgebra
$ A^{\varphi,M} $, where $ \, \gerE' := \gerE \cap A^{\varphi,M} \, $ with $ \, \gerE := Ker \left( \varepsilon \colon \aphiMinf \rightarrow \kq \right) \, $; thus $ \aphiMinf $ is a formal Hopf algebra, complete in the $ \gerE $-adic topology.
\vskip7pt
We now present one of the crucial steps, namely the
realization of the quantum function algebra $ \fqphiMg $ as a formal Hopf subalgebra of $ \aphiMinf $, thus improving [CV-2], Lemma 2.5 (which extends [DL], Lemma 4.3, to the multiparameter case).
\vskip7pt
\proclaim{Theorem 5.7} \ Let $ \, pr^{M'}_\varphi \colon
D_{q,\varphi}^{M'} (\gerg) \twoheadrightarrow U_{q,\varphi}^{M'} (\gerg) \, $ be the canonical epimorphism of \S 3.3. Then dualization gives an embedding of formal Hopf algebras
$$ {(pr^{M'}_\varphi)}^* \colon {U_{q,\varphi}^{M'} (\gerg)}^* \hookrightarrow {D_{q,\varphi}^{M'} (\gerg)}^* \cong {U_{q,\varphi}^{M'} (\gerb_+)}^* \otimeshat
{U_{q,\varphi}^{M'} (\gerb_-)}^* \; ; $$
the restriction of $ {(pr^{M'}_\varphi)}^* $ gives two
embeddings of formal Hopf algebras, namely\break
$ \, \mu^M_\varphi \colon \fqphiMg \hookrightarrow
A^{\varphi,M}_\infty \, $, whose image is contained in
$ A^{\varphi,M} $, and $ \, \mu^{\varphi,M}_\infty \colon
\fqphiMinfg \hookrightarrow A^{\varphi,M}_\infty \, $, the
second being continuous extension of the first.
\endproclaim
\demo{Proof} The statements about restrictions are the
unique non-trivial facts to prove. Recall (cf.~\S 3.1) that the identification
$ \, D_{q,\varphi}^{M'} (\gerg) = U_{q,\varphi}^{M'} (\gerb_+) \otimes
U_{q,\varphi}^{M'} (\gerb_-) \, $ is given by the composition
$$ U_{q,\varphi}^{M'} (\gerb_+) \otimes U_{q,\varphi}^{M'} (\gerb_-) {\buildrel j_+
\otimes j_- \over \hookrightarrow} D_{q,\varphi}^{M'} (\gerg) \otimes
D_{q,\varphi}^{M'} (\gerg) @>{m_D}>> D_{q,\varphi}^{M'} (\gerg) $$
where $ \, j_\pm \colon U_{q,\varphi}^{M'} (\gerb_\pm) \hookrightarrow
D_{q,\varphi}^{M'} (\gerg) \, $ are the natural Hopf embeddings, $ m_D \colon
D_{q,\varphi}^{M'} \otimes D_{q,\varphi}^{M'} (\gerg) @>>> D_{q,\varphi}^{M'} (\gerg)
\, $ is the multiplication of $ D_{q,\varphi}^{M'} (\gerg) \, $ and we look at this
composition as a Hopf algebra isomorphism (thus pulling back the Hopf structure on
$ U_{q,\varphi}^{M'} (\gerb_+) \otimes U_{q,\varphi}^{M'} (\gerb_-) $); by duality, the
identification $ {D_{q,\varphi}^{M'} (\gerg)}^* = {U_{q,\varphi}^{M'} (\gerb_+)}^*
\otimeshat {U_{q,\varphi}^{M'} (\gerb_-)}^* $ is given by $ {(m_D \circ (j_+ \otimes
j_-))}^* = \left( j_+^* \otimeshat j_-^* \right) \circ {m_D}^{\! *} \, $. If $ m_U $ is
the multiplication of $ \uqphiQg $, we have $ \, m_U \circ pr^{M'} \otimes pr^{M'} =
pr^{M'} \circ m_D \, $, hence dualizing \hfill\break
\centerline { $ U_{q,\varphi}^{M'} (\gerb_+) \otimes U_{q,\varphi}^{M'} (\gerb_-)
{\buildrel j_+ \otimes j_- \over \hookrightarrow} D_{q,\varphi}^{M'} (\gerg) \otimes
D_{q,\varphi}^{M'} (\gerg) {\buildrel m_D \over \rightarrow} D_{q,\varphi}^{M'} (\gerg)
{\buildrel pr^{M'} \over \twoheadrightarrow} U_{q,\varphi}^{M'} (\gerg) $ }
yields $ \, {\big( pr^{M'} \circ m_D \circ (j_+ \otimes j_-) \big)}^* = {\big( pr^{M'} \circ
j_+ \big)}^* \otimeshat {\left( pr^{M'} \circ j_- \right)}^* \circ m_U^* \, $; but $ \,
pr^{M'} \circ j_\pm \, $ coincides with the natural embedding $ i_\pm \colon
U_{q,\varphi}^{M'} (\gerb_\pm) \hookrightarrow U_{q,\varphi}^{M'} (\gerg) $, thus
$ {(pr^{M'} \circ m_U \circ (j_+ \!\otimes\! j_-))}^* \! = $
$ = \left( i_+^* \otimeshat i_-^* \right) \circ m_U^* \, $. Now $ m_U^* $ is nothing but
the comultiplication of $ {U_{q,\varphi}^{M'} (\gerg)}^* $, which restricts to the
comultiplication $ \Delta $ of $ \fqphiMg $, while $ \, \rho_\pm := i_\pm^* \colon
{U_{q,\varphi}^{M'} (\gerg)}^* \rightarrow {U_{q,\varphi}^{M'} (\gerb_\pm)}^* \, $ is
the "restriction" map, which maps $ \fqphiMg $ to $ F_{q,\varphi}^M [B_\pm] $;
therefore
\par
$ \, {\left( pr^{M'} \circ m_U \circ (j_+ \otimes j_-)
\right)}^* \left( \fqphiMg \right) = \left( i_+^* \otimeshat
i_-^* \right) \left( \Delta \left( \fqphiPg \right) \right)
\subseteq $
\par
$ {\ } \qquad \hfill \subseteq \rho_+ (\fqphiPg) \otimes \rho_- (\fqphiPg) = \fqphiPbp \otimes \fqphiPbm \cong \uqphiPbm \otimes \uqphiPbp \; ; $
\break
we conclude that the embedding $ \, {U_{q,\varphi}^{M'}
(\gerg)}^* \hookrightarrow {D_{q,\varphi}^{M'} (\gerg)}^* \cong
{U_{q,\varphi}^{M'} (\gerb_p)}^* \otimeshat {U_{q,\varphi}^{M'} (\gerb_-)}^* \, $
maps $ \fqphiMg $ into $ \, \uqphiMbm \otimes \uqphiPbp $. From the very
definitions it follows that $ \mu^M (\fqphiMg) $ vanishes on $ \gerK^P $, hence
(Proposition 5.5) we conclude that $ \, \mu^M \left( \fqphiMg \right) \subseteq
A^{\varphi,M} \, $.
\par
Finally, from what we already proved it follows that the
$ \gerE_F $ (the kernel of\break
$ \, \varepsilon \colon \fqphiMg \rightarrow \kq \, $) is
exactly the preimage of $ \gerE_A $ (the kernel of $ \,
\varepsilon \colon \aphiMinf \rightarrow \kq \, $), whence
--- because of \S 4.5 and \S 5.6 --- the map $ \, \mu^M_\infty
\colon \fqphiMinfg \rightarrow \aphiMinf \, $ coincides
with the continuous extension of $ \, \mu^M \colon \fqphiMg
\rightarrow \aphiMinf \, $, as claimed. $ \square $
\enddemo
\vskip7pt
{\bf 5.8 Integer forms of $ A^{\varphi,M}_\infty \, $.}
\ In order to specialize our quantum objects, we need integer forms. Since
$ \, \aphiMinf \leq {\left( pr^{M'} \right)}^* \left( {U_{q,\varphi}^{M'}}^* \right) \, $
(Proposition 5.5), $ \aphiMinf $ can be thought of as a formal Hopf algebra of linear
functions on $ U_{q,\varphi}^{M'} $; therefore we define integer forms of
$ \aphiMinf $ as dual of integer forms of $ U_{q,\varphi}^{M'} $, that is
$$ \eqalign {
\aphiPinfhat := & \left\{\, f \in \aphiPinf \,\bigg\vert\, \left\langle f, \uqphiQghat
\right\rangle \subseteq \kqqm \,\right\} \, {} \cr
\aphiQinftilde := & \left\{\, f \in \aphiQinf \,\bigg\vert\, \left\langle f, \uqphiPgtilde
\right\rangle \subseteq \kqqm \,\right\} \, . \cr } $$
\indent In the sequel, for a {\it pseudobasis\/}
we will mean a complete system of linear generators which
are linearly independent (i.~e.~a basis in the sense of
topological free modules).
\vskip7pt
\proclaim{Proposition 5.9} \ The algebra $ \aphiPinfhat $ is topologically generated by
$$ \left\{\, \fbar^\varphi_{\alpha^h} \otimes 1, L_{-
(1+\varphi)(\lambda)} \otimes L_{(1-\varphi)(\lambda)}, 1
\otimes \ebar^\varphi_{\alpha^k} \,\Big\vert\, h, k =1,
\ldots, N; \lambda \in P_+ \,\right\} $$
and has the $\kqqm$-pseudobasis of PBW type
$$ \left\{\, \prod_{h=N}^1 {\left( \fbar^\varphi_{\alpha^h}
\right)}^{a_h} \cdot L_{-(1+\varphi)(\lambda)} \otimes
L_{(1-\varphi)(\lambda)} \cdot \prod_{k=1}^N {\left(
\ebar^\varphi_{\alpha^k} \right)}^{d_k} \,\bigg\vert\, a_h, d_k \in \N; \lambda \in P_+ \,\right\} \; ;
\eqno (5.6) $$
in particular, it is a $ \kqqm $-form of the formal Hopf
algebra $ \aphiPinf \, $.
\hfill\break
\indent (b) The algebra $ \aphiQinftilde $ is
topologically generated by
$$ \left\{ {\left( F^\varphi_{\alpha^h} \right)}^{(a)} \!
\otimes \! 1, \left( K_{-(1+\varphi)(\alpha_i)} \! \otimes \! K_{(1-\varphi)(\alpha_i)}; c
\atop t \right), 1 \! \otimes \! {\left( E^\varphi_{\alpha^k} \right)}^{(d)} \,\Big\vert\,
\forall \, h, k, i, \, a, t, d \in \N, \, c \in \Z \right\} $$ and has the $ \kqqm
$-pseudobasis of PBW type
$$ \left\{\prod_{h=N}^1 \! {\left(
F^\varphi_{\alpha^h} \right)}^{(a_h)} \! \cdot \! \prod_{i=1}^n \! \left(\!
K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)}; 0 \atop t_i \! \right) \!
\cdot \! \prod_{k=1}^N \! {\left( E^\varphi_{\alpha^k} \right)}^{(d_k)} \bigg\vert a_h,
t_i, d_k \in \N \right\} \! ; \eqno (5.7) $$ in particular, it is a $ \kqqm $-form of the
formal Hopf algebra $ \aphiQinf \, $. $ \square $
\endproclaim
\demo{Proof} {\it (a)\/} It is clear that (5.6) is a (PBW-
type) $\kq$-basis of $ \aphiPinf $; thus, given
$ f \in \aphiPinf $, write it as a (possibly infinite) sum
$$ f = \sum_\sigma a_\sigma \cdot \fbar^\varphi_\sigma
\cdot L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
\cdot \ebar^\varphi_\sigma $$
where $ \, a_\sigma \in \kq \, $ and the
$ \fbar^\varphi_\sigma $'s, resp.~the
$ L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma $'s,
resp.~the $ \ebar^\varphi_\sigma $'s, are ordered
monomials in the $ \fbar^\varphi_{\alpha^k} $'s, resp.~in the $ \, L_{-(1+\varphi)
(\lambda)} \otimes L_{(1-\varphi)(\lambda)} \, $'s, resp.~in the
$ \ebar^\varphi_{\alpha^k} $'s ($ k=1, \dots, N $,
$ \lambda \in P $). For any monomial of divided powers
$ \, \ehat \cdot (K) \cdot \fhat \, $ in the PBW basis of $ \uqphiQghat $ we have
$$ \eqalign{
\left\langle f, \ehat \cdot (K) \cdot \fhat \right\rangle = &
\sum_\nu a_\sigma \cdot \left\langle \fbar^\varphi_\sigma \cdot
L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma \cdot
\ebar^\varphi_\sigma, \ehat \cdot (K) \otimes 1 \cdot \fhat
\right\rangle = \cr
= & \sum_\sigma a_\sigma \cdot {\left\langle
\fbar^\varphi_\sigma, \ehat \right\rangle}_{\pi_-^\varphi} \cdot {\left\langle
L^{\varphi,-}_\sigma, (K) \right\rangle}_{\pi_-^\varphi} \cdot {\left\langle
L^{\varphi,+}_\sigma, 1 \right\rangle}_{\overline{\pi_-^\varphi}} \cdot {\left\langle
\ebar^\varphi_\sigma, \fhat \right\rangle}_{\overline{\pi_-^\varphi}} \cr } $$
\indent For any fixed $ \bar\sigma $, there exist divided powers monomials
$ \ehat_{\bar \sigma} $, $ \fhat_{\bar \sigma} $ such that $ \, {\left\langle
\fbar^\varphi_{\bar \sigma}, \ehat_{\bar \sigma} \right\rangle}_{\pi_-^\varphi} = $
$ = \pm q^r \, $, $ \, \quad {\left\langle \ebar^\varphi_{\bar \sigma}, \fhat_{\bar
\sigma} \right\rangle}_{\overline{\pi_-^\varphi}} = \pm q^s \, $ for some $ \, r, s \in
\Z $, and $ \, {\left\langle \fbar^\varphi, \ehat_{\bar \sigma}
\right\rangle}_{\pi_-^\varphi} = 0 \, $, $ \, {\left\langle \ebar^\varphi, \fhat_{\bar
\sigma} \right\rangle}_{\overline{\pi_-^\varphi}} = 0 \, $ for {\it all\/} PBW
monomials $ \fbar^\varphi \neq \fbar^\varphi_{\bar \sigma} $ and $ \ebar^\varphi
\neq \ebar^\varphi_{\bar \sigma} $ (cf.~(2.2)); then \hfill\break
\centerline { $ \left\langle f, \ehat_{\bar \sigma} \cdot (K) \cdot \fhat_{\bar \sigma}
\right\rangle = \sum_{\sigma \in {\Cal S}} a_\sigma \epsilon_\sigma q^{r+s} \cdot
{\left\langle L^{\varphi,-}_\sigma, (K) \right\rangle}_{\pi_-^\varphi} $ }
(where $ \, {\Cal S} := \left\{\, \sigma \mid \fbar^\varphi_\sigma =
\fbar^\varphi_{\bar \sigma}, \ebar^\varphi_\sigma = \ebar^\varphi_{\bar \sigma}
\,\right\} \, $ and $ \, \epsilon_\sigma = \pm 1 \, $) which can be rewritten
as \hfill\break
\centerline { $ \left\langle \fhat_{\bar \sigma} \triangleright f \triangleleft
\ehat_{\bar \sigma}, (K) \otimes 1 \right\rangle = \sum_{\sigma \in {\Cal S}} a_\sigma
\epsilon_\sigma q^{r+s} \cdot {\left\langle L^{\varphi,-}_\sigma, (K)
\right\rangle}_{\pi_-} $ } (where $ \triangleright $, resp.~$ \triangleleft $, denotes
the left, resp.~right, action of $ \uqphiQg $ onto $ {\uqphiQg}^* \, $); therefore
$$ \left( \fhat_{\bar \sigma} \triangleright f \triangleleft \ehat_{\bar \sigma} \right)
\! {\Big\vert}_{{\uqphiQthat}^{\otimes 2}} = \sum_{\sigma \in {\Cal S}} a_\sigma
\epsilon_\sigma q^{r+s} \cdot L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
\eqno (5.8) $$
(notice that $ \, \uqphiQthat = \uqQthat \, $ as Hopf algebras). Now if
$ \, f \in \aphiPinfhat $, we have that $ \, \left( \fhat_{\bar \sigma} \triangleright f
\triangleleft \ehat_{\bar \sigma} \right) {\Big\vert}_{{\uqphiQthat}^{\otimes 2}} \, $ is
an integer valued function on $ \, {\uqphiQthat}^{\otimes 2} = {\uqQthat}^{\otimes 2}
\, $, whence from (5.8) and (2.3) we get $ \, a_\sigma \epsilon_\sigma q^{r+s} \in
\kqqm \, $, hence $ \, a_\sigma \in \kqqm \, $, for all $ \sigma \in {\Cal S} $; in
particular $ \, a_{\bar \sigma} \in \kqqm \, $: since this can be done for all $ \sigma
$ such that $ a_\sigma \neq 0 $ we conclude that $ \, a_\sigma \in \kqqm \, $ for all
coefficients $ a_\sigma $. Conversely, if the latter holds we clearly have $ \, f \in
\aphiPinfhat \, $. It is then clear that $ \aphiPinfhat $ has the above PBW
pseudobasis over $ \kqqm $, has the claimed set of (topological) generators, and is a
$ \kqqm $-form of $ \aphiPinf $.
\par
Now look at the Hopf structure. Let $ \, f \in
\aphiPinfhat \, $; write $ \Delta(f) $ in terms of the
PBW basis,
$$ \Delta(f) = \sum_\sigma a_\sigma \cdot \left( \fbar^\varphi_\sigma L^{\varphi,-}_\sigma \otimes
L^{\varphi,+}_\sigma \ebar^\varphi_\sigma \right) \otimes
\left( {\left( \fbar^\varphi_\sigma \right)}' {\left(
L^{\varphi,-}_\sigma \right)}' \otimes {\left(
L^{\varphi,+}_\sigma \right)}' {\left( \ebar^\varphi_\sigma
\right)}' \right) $$
(for possibly infinite $ \sigma \, $); as above we have to prove that $ \, a_\sigma \in
\kqqm \, $ for all $ \sigma $, and we proceed as in the first part of the
proof. Since $ \, f \in \aphiPinfhat \, $, i.~e.~it is an
integer valued function on $ \uqphiQghat $, then
$ \Delta(f) $ is an integer valued function on $ \,
\uqphiQghat \otimes \uqphiQghat $. For any fixed $ \bar \sigma $ such that
$ \, a_{\bar \sigma} \neq 0 \, $ we locate PBW monomials $ \ehat_{\bar \sigma} $,
$ \fhat_{\bar \sigma} $, $ \ehat'_{\bar \sigma} $,
$ \fhat'_{\bar \sigma} $ such that
$$ \displaylines{
\left( \left( \fhat^\varphi_{\bar \sigma} \otimes \fhat'_{\bar \sigma} \right)
\triangleright \Delta(f) \triangleleft \left( \ehat_{\bar \sigma} \otimes \ehat'_{\bar
\sigma} \right) \right) \! {\Big\vert}_{{\uqphiQthat}^{\otimes 2}} = \hfill \qquad {} \cr
{} \quad \hfill = \sum_{\sigma \in {\Cal S}} a_\sigma
\epsilon_\sigma q^{r+s+r'+s'} \! \cdot \left(
L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma \right) \otimes \left( {\left(
L^{\varphi,-}_\sigma \right)}' \otimes {\left( L^{\varphi,+}_\sigma \right)}' \right)
\cr } $$
where $ \, {\Cal S} := \left\{\, \sigma \,\Big\vert\, \fbar^\varphi_\sigma = \fbar^\varphi_{\bar \sigma}, \ebar^\varphi_\sigma = \ebar^\varphi_{\bar \sigma},
{\left( \fbar^\varphi_\sigma \right)}' = {\left( \fbar^\varphi_{\bar \sigma} \right)}',
{\left( \ebar^\varphi_\sigma \right)}' = {\left( \ebar^\varphi_{\bar \sigma} \right)}' \,\right\} \, $ and $ \epsilon_\sigma = \pm 1 $; but $ \, \left( \left( \fhat_{\bar \sigma} \otimes \fhat'_{\bar \sigma} \right) \triangleright \Delta(f) \triangleleft
\left( \ehat_{\bar \sigma} \otimes \ehat_{\bar \sigma} \right) \right) {\Big\vert}_{{\uqQthat}^{\otimes 2}} \, $ is integer valued, hence like above we get
$ \, a_\sigma \epsilon_\sigma q^{r+s+r'+s'} \in \kqqm \, $ and $ \, a_\sigma \in \kqqm \, $ for all $ \sigma \in {\Cal S} $, whence we conclude the same for all
$ a_\sigma $: thus $ \, \Delta(f) \in \aphiPinfhat \otimeshat \aphiPinfhat \, $, i.~e.~$ \aphiPinfhat $
is a (formal) subcoalgebra of $ \aphiPinf $. Furthermore, $ 1 \in \aphiPinfhat $ by definition, $ \, \varepsilon \left( \aphiPinfhat \right) \leq \kqqm \, $, because $ \, 1 \in \uqphiQghat \, $ and $ S \left( \aphiPinfhat \right) = \aphiPinfhat $ because $ S \left( \uqphiQghat \right) = \uqphiQghat $. Thus $ \aphiPinfhat $ is a (formal) Hopf
subalgebra of $ \aphiPinf $ over $ \kqqm $, q.e.d. The last part of claim {\it (a)\/} is trivial: just remark that (for all $ \, \lambda \in P_+ \, $) $ \, L_{(1+\varphi)(\lambda)} \otimes L_{-(1-\varphi)(\lambda)} = {\left( L_{-(1+\varphi)(\lambda)} \otimes
L_{(1-\varphi)(\lambda)} \right)}^{-1} = $
$ = {\left( 1 -\left( 1 - L_{-(1+\varphi)(\lambda)} \otimes L_{(1-\varphi)(\lambda)} \right) \right)}^{-1} = \sum_{n=0}^\infty {\left( 1 - L_{-(1+\varphi)(\lambda)} \otimes L_{(1-\varphi)(\lambda)} \right)}^n \, $.
\par
{\it (b)\/} mimick step by step the proof of {\it
(a)\/}. $ \square $
\enddemo
\vskip7pt
{\bf 5.10} {\it Remarks: \ (a)} \ Notice that, letting
\hfill\break
\centerline { $ \aphiPhat := \left\{\, f \in A^{\varphi,P} \,\Big\vert\, \left\langle f, \uqphiQghat \right\rangle \subseteq \kqqm \,\right\} $ }
\centerline { $ \aphiQtilde := \left\{\, f \in A^{\varphi,Q} \,\Big\vert\, \left\langle f,
\uqphiPgtilde \right\rangle \subseteq \kqqm \,\right\} $ }
one can define the {\it algebra\/} $ \aphiPinfhat $, resp.~$ \aphiQinftilde \, $, as the $ \gerE' $-adic completion of the subalgebra $ \aphiPhat $,
resp.~$ \aphiQtilde \, $, where $ \, \gerE' := \gerE \cap \aphiPhat \, $ with $ \, \gerE := Ker \left( \varepsilon \colon \aphiPinf \rightarrow \kq \right) \, $,
resp.~$ \, \gerE' := \gerE \cap $
$ \cap \aphiQtilde \, $ with $ \, \gerE := Ker \left( \varepsilon \colon \aphiQinf \rightarrow \kq \right) \, $); therefore $ \aphiPinfhat $, resp.~$ \aphiQinftilde \, $,
is a formal Hopf\break
algebra over $ \kqqm $, complete in the $ \gerE'' $-adic
topology, with $ \, \gerE'' := $\break
$ := Ker \left( \varepsilon \colon \aphiPinfhat \rightarrow
\kqqm \right) \, $, resp.~$ \, \gerE'' := Ker \left( \varepsilon \colon \aphiQinftilde \rightarrow \kqqm
\right) \, $.
\par
{\it (b)} \ The analysis in the proof of Proposition 5.9 also shows that
\hfill\break
\centerline { $ \uqphiQghat := \left\{\, x \in \uqphiQg \,\Big\vert\, \left\langle \aphiPhat , x \right\rangle \subseteq \kqqm \,\right\}
\, $\phantom{.} }
\centerline { $ \uqphiPgtilde := \left\{\, x \in \uqphiPg \,\Big\vert\, \left\langle \aphiQtilde, x \right\rangle \subseteq \kqqm \,\right\} \, $. }
\vskip7pt
The following proposition --- whose proof is trivial from the previous results --- describes the behavior of
integer forms under the natural embeddings
$ \mu^{\varphi,M} $, $ \mu^{\varphi,M}_\infty $; in
particular, it affords another proof of the fact that
$ \fqphiPghat $, resp.~$ \fqphiQgtilde $, is a
Hopf subalgebra (over $ \kqqm $) of $ \fqphiPg $,
resp.~$ \fqphiQg $.
\vskip7pt
\proclaim{Proposition 5.11} The natural embeddings
preserve integer forms, namely
$$ \eqalign{
\fqphiPghat = {\left( \mu^{\varphi,P} \right)}^{-1} \left( \aphiPinfhat \right) \, , & \quad \fqphiQgtilde =
{\left( \mu^{\varphi,Q} \right)}^{-1} \left( \aphiQinftilde
\right) \cr
\fqphiPinfghat = {\left( \mu^{\varphi,P}_\infty \right)}^{-1} \left( \aphiPinfhat \right) \, , & \quad
\fqphiQinfgtilde = {\left( \mu^{\varphi,Q}_\infty \right)}^{-1} \left( \aphiQinftilde \right) \cr } $$
so that restriction gives embeddings of formal Hopf algebras over $ \kqqm $
$$ \eqalign{
\mu^{\varphi,P} \colon \fqphiPghat \hookrightarrow
\aphiPinfhat \, , & \quad \mu^{\varphi,Q}
\colon \fqphiQgtilde \hookrightarrow \aphiQinftilde \cr
\mu^{\varphi,P}_\infty \colon \fqphiPinfghat
\hookrightarrow \aphiPinfhat \, , &
\quad \mu^{\varphi,Q}_\infty \fqphiQinfgtilde
\hookrightarrow \aphiQinftilde \, . \; \square \cr } $$
\endproclaim
\vskip7pt
{\bf 5.12.} \ The result above can in fact be refined,
extending embeddings to isomorphisms. Let $ \, \mu \in M \cap P_+ \, $, and let $ V_{-\mu} $ be an irreducible \hbox{$ U_{q,\varphi}^{M'} (\gerg) $-modu}le of lowest weight $ -\mu $ (recall that $ \, U_{q,\varphi}^{M'} (\gerg) \cong U_q^{M'}\!(\gerg) \, $ as algebras, hence their representation theory is the same). Let $ v_{-\mu} \neq 0 $ be a lowest weight vector of $ V_{-\mu} $, and let $ \, \phi_{-\mu} \in {V_{-\mu}}^{\! *} \, $ be the linear functional on $ V_{-\mu} $ defined by \ {\it (a)\/}~$ \phi_{-\mu} (v_{-\mu}) = 1 $ \ and \ {\it (b)\/}~$ \phi_{-\mu} $ vanishes on the unique
\hbox{$ U_{q,\varphi}^{M'} (\gert) $-inva}riant complement
of $ \kq.v_{-\mu} $ in $ V_{-\mu} \, $; let $ \,
\psi_{-\mu} := c_{\phi_{-\mu},v_{-\mu}} \, $ be the
corresponding matrix coefficient, i.~e. $ \, \psi_{-\mu}
\colon x \mapsto \phi_{-\mu} (x.v_{-\mu}) \, $ for all
$ \, x \in U_{q,\varphi}^{M'} (\gerg) \, $; finally, let $ \, \rho := \sum_{i=1}^n \omega_i \, $, $ \, \delta := \sum_{i=1}^n \alpha_i \, $. Then we can refine Proposition 5.11 with the following result, which improves [DL], Theorem 4.6, and [CV-2], Lemma 2.5:
\vskip7pt
\proclaim{Theorem 5.13}
\hfill\break
\indent (a1) The embedding $ \, \mu^{\varphi,P}
\colon \fqphiPghat \hookrightarrow \aphiPhat \, $ extends
to a $\kqqm$-algebra isomorphism
$$ \mu^{\varphi,P} \colon \fqphiPghat \left[ \psi_{-\rho}^{-1} \right] @>{\cong}>> \aphiPhat \, ; $$
it follows that $ \fqphiPghat $ is a $ \kqqm $-form of
$ \fqphiPg $.
\hfill\break
\indent (a2) The algebra $ \fqphiPghat\left[\psi_{-
\rho}^{-1}\right] $ is naturally embedded into
$ \fqphiPinfghat $; it follows that the isomorphism in (a1)
coincides with the restriction of
$ \mu^{\varphi,P}_\infty $, and the latter is in fact an
isomorphism $ \, \mu^{\varphi,P}_\infty \colon \fqphiPinfghat @>{\cong}>> \aphiPinfhat \, $ of formal Hopf algebras over $ \kqqm $.
\hfill\break
\indent (b1) The embedding $ \, \mu^{\varphi,Q} \colon \fqphiQgtilde \hookrightarrow \aphiQtilde \, $ extends to a $\kqqm$-algebra isomorphism
$$ \mu^{\varphi,Q} \colon \fqphiQgtilde\left[\psi_{-
\delta}^{-1}\right] @>{\cong}>> \aphiQtilde \, ; $$
it follows that $ \fqphiQgtilde $ is a $ \kqqm $-form of
$ \fqphiQg $.
\hfill\break
\indent (b2) The algebra $ \fqphiQgtilde\left[\psi_{-
\delta}^{-1}\right] $ is naturally embedded into
$ \fqphiQinfgtilde $; it follows that the isomorphism in
(b1) coincides with the restriction of
$ \mu^{\varphi,Q}_\infty $, and the latter is in fact an
isomorphism $ \, \mu^{\varphi,Q}_\infty \colon \fqphiQinfgtilde @>{\cong}>> \aphiQinftilde \, $ of formal Hopf algebras over $ \kqqm $.
\endproclaim
\demo{Proof} Once again we stick for simplicity to the case
$ \, \varphi = 0 \, $.
\par
It is shown in [DL], \S\S 4.4--6, that $ \, \mu^P \colon
\fqPghat \hookrightarrow \aPhat \, $ extends to a $\kqqm$-algebra isomorphism $ \, \mu^P \colon \fqPghat \left[
\psi_{-\rho}^{-1}\right] \cong \aPhat \, $; then scalar extension gives a $\kq$-algebra isomorphism $ \, \mu^P \colon \fqPg \left[ \psi_{-\rho}^{-1} \right] \cong A^P
\, $; then $ \aPhat $ be a $ \kqqm $-form of $ A^P $ implies $ \fqphiPghat $ be a $ \kqqm $-form of
$ \fqphiPg $; thus {\it (a1)\/} holds. The same argument
(disregarding integrality) shows that $ \, \mu^Q \colon \fqQg \hookrightarrow A^Q \, $ extends to a $\kq$-algebra isomorphism $ \, \mu^Q \colon \fqQg \left[ \psi_{-\delta}^{-1} \right] \cong A^Q \, $. Furthermore, computations in [DL] give \ $ \mu^Q (\psi_{-\mu}) = K_{-\mu} \otimes K_\mu $ \ for all weights $ \, \mu \, $; thus in particular for $ \mu = \alpha_i $ and $ \mu = \delta $ we get
$$ \mu^Q \left( \psi_{-\alpha_i} \right) = K_i^{-1} \otimes K_i \, , \; \mu^Q \left( \psi_{-\delta} \right) = K_{-\delta} \otimes K_\delta \, , \; \mu^Q \left( \psi_{-\delta}^{-1} \right) = K_\delta \otimes K_{-\delta} \eqno (5.9) $$
whence $ \, K_i^{-1} \otimes K_i \in Im \left( \mu^Q \right) \, $ for all $ i=\unon \, $. Even more, from the proof of Theorem 4.6 of [DL] we also get that $ \, F_i \otimes 1, 1 \otimes E_i \in \mu^Q \left( \fqQg \right)
\, $, hence also $ \, F_i^{(f)} \otimes 1, 1 \otimes E_i^{(e)} \in \mu^Q \left( \fqQg \right) \, $ (for all
$ i=\unon $, $ f, e \in \N $); then from Proposition 5.11 we get $ \, F_i^{(f)} \otimes 1, 1 \otimes E_i^{(e)} \in \mu^Q \left( \fqQgtilde \right) \, $, for all $ i=\unon $ and $ f, e \in \N $; similarly, we have $ \, K_i^{-1} \otimes K_i = \mu^Q (z_i) \in \mu^Q \left( \fqQgtilde \right) \, $, with $ \, z_i := \psi_{-\alpha_i} \in
\fqQgtilde \, $, because of Proposition 5.11. Thus (from (5.9))
$$ K_i \otimes K_i^{-1} = \left( \prod_{j=1 \atop j \neq i}^n K_j^{-1} \otimes K_j \right) \cdot (K_\delta \otimes K_{-\delta}) \in \mu^Q \left( \fqQgtilde\left[\psi_{-\delta}^{-1}\right] \right) $$
and on the other hand
$$ \displaylines{
\left( K_i^{-1} \otimes K_i; c \atop t \right) :=
\prod_{s=1}^t {{q_i^{c-s+1} \cdot K_i^{-1} \otimes K_i - 1}
\over {q_i^s - 1}} = \hfill \qquad {\ } \cr
{\ } \qquad \hfill = \prod_{s=1}^t {{\mu^Q(z_i) q_i^{c-
s+1} - 1} \over {q_i^s - 1}} = \left( \mu^Q(z_i); c \atop t
\right) = \mu^Q \left( \left( z_i; c \atop t \right) \right)
\cr } $$
and again by Proposition 5.11 we get $ \left( z_i; c \atop
t \right) \in \fqQgtilde $, thus $ \, \left( K_i^{-1}
\otimes K_i; c \atop t \right) \, $ (for all $ i $,
$ c $, $ t $). All this and Proposition 5.9 give $ \, \mu^Q \left( \fqQgtilde \left[ \psi_{-\delta}^{-1} \right] \right) = \aQtilde \, $; then $ \aQtilde $ be a
$ \kqqm $-form of $ A^Q $ implies $ \fqphiQgtilde $ be a $ \kqqm $-form of $ \fqphiQg $, thus {\it (b1)\/} holds.
\par
Now consider {\it (a1)\/}: notice that $ \, K_{-\delta} \otimes K_\delta = 1 - (1 - K_{-\delta} \otimes K_\delta) = 1 - y \, $, where $ \, y := 1 - K_{-\delta} \otimes K_\delta \in \gerE := Ker (\varepsilon \colon \aPinfhat \rightarrow \kqqm) \, $; thus on the other side we have $ \, \psi_{-\delta} = 1 - (1 - \psi_{-\delta}) = 1 - x \, $, where $ \, x := 1 - \psi_{-\delta} \in \gerE := Ker \left( \varepsilon \colon \fqPghat \rightarrow \kqqm \right) \, $ (for $ \, \mu^P \colon \fqPghat \hookrightarrow \aPinfhat \, $ is a morphism of formal
Hopf algebras); therefore $ \psi_{-\delta} $ is invertible in $ \fqPinfghat $, its inverse being $ \, \sum_{n=0}^\infty x^n = \sum_{n=0}^\infty {(1 - \psi_{-\delta})}^n \, $ (cf.~\S 4.5), hence $ \fqPghat \left[ \psi_{-\delta}^{-1} \right] $ embeds into $ \fqPinfghat $ and {\it (a2)\/} follows. Similarly for {\it (b2)\/}.
\par
The same argument works for general $ \varphi $ too, starting from [CV-2], Proposition 2.7. $ \square $
\enddemo
\vskip7pt
{\bf 5.14 Gradings.} \ Recall that $ U_{q,\varphi}^M
(\gerb_\pm) $ has a $ Q_\pm $-grading $ U_{q,\varphi}^M
\!(\gerb_\pm) \!=\! \oplus_{\alpha \in Q_\pm} \! {(U_{q,\varphi}^M \!
(\gerb_\pm))}_\alpha $ given by the direct sum decomposition into weight spaces for
the adjoint action of $ U_{q,\varphi}^M (\gert) $. These are gradings of Hopf algebras
(in the obvious usual sense), which are inherited by integer forms $ \uqphiQbphat $,
$ \uqphiPbmtilde $, etc.~etc., and are compatible with (or respected by) DRT pairings,
viz.~for instance $ \, \pi_-^\varphi \left({ \left( \uqphiQbm \right)}_\beta, {\left(
\uqphiPbp \right)}_\gamma \right) = 0 \, $ for all $ \, \beta \in Q_- $, $ \gamma \in
Q_+ \, $ such that $ \, \beta + \gamma \neq 0 \, . $
\par
The gradings of Borel subalgebras provide a $ Q $-grading of the Hopf algebra
$ \, D_{q,\varphi}^M (\gerg) = $\break
$ = U_{q,\varphi}^M (\gerb_+) \otimes \uqphiQbm \, $ (inherited by the quotient Hopf
algebra $ \uqphiMg $), the subspace\break
$ \, {\left( U_{q,\varphi}^M (\gerb_+) \right)}_\beta \otimes {\left( \uqphiQbm
\right)}_\gamma \, $ having degree $ \beta - \gamma $, and a $ Q $-grading of the
formal Hopf algebra\break
$ \, U_{q,\varphi}^M (\gerb_+) \otimeshat \uqphiQbm \, $;
it is clear that the restriction of the natural evaluation pairing\break
$ \, {D_{q,\varphi}^M (\gerg)}^* \otimes D_{q,\varphi}^M (\gerg) \rightarrow \kq \, $
respects these gradings.
\vskip7pt
{\bf 5.15 Computational results.} \ We now look for concrete information about the
formal Hopf algebra structure of $ \aphiPinf $ and $ \aphiQinf \, $ and their integer
forms.
\par
First of all, the counity $ \, \varepsilon \colon
{D_{q,\varphi}^Q (\gerg)}^* \rightarrow \kq \, $ is defined
as $ \, \varepsilon := 1^* \, $, hence $ \, \varepsilon
(x^*) := \langle x^*, 1 \rangle \, $, for all $ \, x^* \in
{D_{q,\varphi}^Q (\gerg)}^* \, $; thus in particular the
restriction $ \, \varepsilon \colon \aphiPinfhat
\rightarrow \kqqm \, $ is defined by
$$ \varepsilon \left( \fbar^\varphi_{\alpha^k} \otimes 1
\right) = 0 \, , \quad \varepsilon \left( L_{-(1+\varphi)
(\lambda)} \otimes L_{(1-\varphi)(\lambda)} \right) = 1 \, ,
\quad \varepsilon \left( 1 \otimes \ebar^\varphi_{\alpha^k}
\right) = 0 $$
and by scalar extension these same formulas define also $ \varepsilon \colon
\aphiPinf \rightarrow \kq $. Similarly,\break
$ \, \varepsilon \colon \aphiQinftilde \rightarrow \kqqm \, $ comes out to be defined
by
$$ \displaylines{
\varepsilon \left( {(F^\varphi_i)}^{(f)} \otimes 1
\right) = 0 \, , \; \qquad \varepsilon \left( K_{-(1+\varphi) (\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} \right) = 1 \cr
\varepsilon \left( \left( K_{-(1+\varphi) (\alpha_i)}
\otimes K_{(1-\varphi)(\alpha_i)} ; c \atop t \right)
\right) = \left( c \atop t \right)_{q_i} \, , \cr
\varepsilon \left( K_i \otimes K_i^{-1} \right) = 1 \, ,
\; \qquad \varepsilon \left( 1 \otimes {(E^\varphi_i)}^{(e)}
\right) = 0 \cr } $$
and by scalar extension these same formulas define also
$ \, \varepsilon \colon \aphiQinf \rightarrow \kq \, $.
\par
Now for the antipode $ \, S \colon {D_{q,\varphi}^Q
(\gerg)}^* \rightarrow {D_{q,\varphi}^P (\gerg)}^* \, $: it
is by definition the transpose of the antipode of
$ D_{q,\varphi}^Q (\gerg) $, hence it is characterized by
$ \, \langle S(x^*), x \rangle = \langle x^*, S(x) \rangle
\, $, for all $ \, x^* \in {D_{q,\varphi}^Q (\gerg)}^* $,
$ x \in D_{q,\varphi}^Q (\gerg) $. Now consider $ \, F^\varphi_i \otimes 1 \in \uqphiPbm \otimeshat \uqphiPbp \leq {D_{q,\varphi}^Q (\gerg)}^* \, $: it is homogeneous of degree $ \, -\alpha_i \, $, whence $ S \left( F^\varphi_i \otimes 1 \right) $ has the same degree
(for $ S $ is degree preserving, the $ Q $-grading being
compatible with the (formal) Hopf structure: cf.~\S 5.14). Thus writing $ S \left( F^\varphi_i \otimes 1 \right) $ as a series \hfill\break
\centerline { $ S \left( F^\varphi_i \otimes 1 \right) = \sum_\sigma c_\sigma \cdot F^\varphi_\sigma L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma E^\varphi_\sigma $ }
(where $ \, F^\varphi_\sigma L^{\varphi,-}_\sigma \otimes
L^{\varphi,+}_\sigma E^\varphi_\sigma \, $ denotes a PBW
monomial) we have $ \, c_\sigma \neq 0 \, $
only if\break
$ \, deg(F^\varphi_\sigma L^{\varphi,-}_\sigma \otimes
L^{\varphi,+}_\sigma E^\varphi_\sigma) :=
deg(F^\varphi_\sigma) + deg(E^\varphi_\sigma) = -\alpha_i
\, $.
\par
The integer form $ \aphiPinfhat $ of $ \aphiPinf $
inherits a $ Q $-grading, and the same procedure gives us
$$ S \left( \fbar^\varphi_{\alpha^h} \otimes 1 \right) =
\sum_\sigma c_\sigma^- \cdot \fbar^\varphi_\sigma
L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
\ebar^\varphi_\sigma \eqno (5.10) $$
(for $ \, h= 1, \ldots, N \, $) where $ \, c_\sigma^- \in
\kqqm \, $ and the $ \fbar^\varphi_\sigma
L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
\ebar^\varphi_\sigma $'s are PBW monomials of degree $ \,
deg(\fbar^\varphi_\sigma) + deg(\ebar^\varphi_\sigma) = -
\alpha^h \, $. Similarly for $ \, {(F^\varphi_i)}^{(f)}
\otimes 1 \in \aphiQinftilde \, $ one has
$$ S \left( {(F^\varphi_i)}^{(f)} \otimes 1 \right) =
\sum_\sigma a_\sigma^- \cdot \fhat^\varphi_\sigma
(K^{\varphi,-}_\sigma) \otimes (K^{\varphi,+}_\sigma)
\ehat^\varphi_\sigma \eqno (5.11) $$
(for $ \, f \in \N, i= \unon \, $) where $ \, a_\sigma^-
\in \kqqm \, $ and the $ \fhat^\varphi_\sigma
(K^{\varphi,-}_\sigma) \otimes (K^{\varphi,+}_\sigma)
\ehat^\varphi_\sigma $'s are PBW monomials of degree $ \,
deg(\fhat^\varphi_\sigma) + deg(\ehat^\varphi_\sigma) = - f
\alpha_i \, $. But we can compare $ \aphiPinfhat $ and
$ \aphiQinftilde $ by means of the obvious embedding $ \,
\aphiQinf \hookrightarrow \aphiPinf \, $; since $ \, {(\fbar^\varphi_{\alpha^h})}^f \otimes 1 = {(q_{\alpha^h} - q_{\alpha^h}^{-1})}^f \cdot
{[f]}_{q_{\alpha^h}} \cdot {(F^\varphi_{\alpha^h})}^{(f)} \otimes 1 \, $ and $ \, 1 \otimes {(\ebar^\varphi_{\alpha^k})}^e = {(q_{\alpha^k} -
q_{\alpha^k}^{-1})}^e \cdot {[e]}_{q_{\alpha^k}} \cdot 1 \otimes {(E^\varphi_{\alpha^k})}^{(e)} \, $ ($ \, h, k= 1, \ldots, N; f, e \in \N \, $), direct
comparison between (5.10) and (5.11) gives $ \, a_\sigma^- = {(q - q^{-1})}^{\text{e}_\sigma} \cdot b_\sigma \cdot c_\sigma^- \, $, where $ \, b_\sigma \in \kqqm \, $ and
$ \, \text{e}_\sigma := \sum_{s=1}^N (f_s + e_s) \, $, with
$ \, \fhat^\varphi_\sigma = \prod_{h=N}^1 {(F^\varphi_{\alpha^h})}^{(f_h)} \, $, $ \, \ehat^\varphi_\sigma = \prod_{k=1}^N {(E^\varphi_{\alpha^k})}^{(e_k)} \, $. In particular, we
can write $ S \left( {(F^\varphi_i)}^{(f)} \otimes 1
\right) $ as a series
$$ S \left( {(F^\varphi_i)}^{(f)} \otimes 1 \right) =
\sum_{\kappa=0}^\infty {(q - q^{-1})}^\kappa \cdot d_\kappa
\cdot \fhat^\varphi_\kappa (K^{\varphi,-}_\kappa) \otimes
(K^{\varphi,+}_\kappa) \ehat^\varphi_\kappa
\eqno (5.12) $$
(with $ d_\kappa \in \kqqm $ for all $ \kappa \in \N $)
which is convergent in the $ (q-1) $-adic topology of
$ \aphiQinftilde $.
\par
Similarly one finds that $ S \left( K_{-(1+\varphi)
(\alpha_i)} \otimes K_{(1-\varphi) (\alpha_i)} \right) $,
$ S \left( K_{-(1+\varphi) (-\alpha_i)} \otimes K_{(1-
\varphi) (-\alpha_i)} \right) $, $ S \left( \left( K_{-
(1+\varphi) (\alpha_i)} \otimes K_{(1-\varphi) (\alpha_i)} ;
c \atop t \right) \right) $ and $ S \left( 1 \otimes
{(E^\varphi_i)}^{(e)} \right) $ are expressed by series
which are convergent in the $ (q-1) $-adic topology
of $ \aphiQinftilde $. In particular, all such series are
in fact {\it finite\/} sums modulo $ (q-1) $: we will now compute the first terms
of these series.
\par
For $ S \left( F^\varphi_i \otimes 1 \right) $ the first term --- call it
$ {\Cal F}_1 $ --- attached to $ \kappa=0 $ in the series (5.12) corresponds to
the summands $ \, a_\sigma^- \cdot \fhat^\varphi_\sigma (K^{\varphi,-}_\sigma)
\otimes (K^{\varphi,+}_\sigma) \ehat^\varphi_\sigma \, $ in (5.11) such that $ \,
\partial \left( \fhat^\varphi_\sigma \right) + $
$ + \partial \left( \ehat^\varphi_\sigma
\right) = 1 \, $, with $ \, \partial \left( \prod_{h=1}^N {(F^\varphi_{\alpha^h})}^{(f_h)}
\right) := \sum_{h=N}^1 f_h \, $, $ \, \partial \left( \prod_{k=N}^1
{(E^\varphi_{\alpha^k})}^{(e_k)} \right) := \sum_{k=1}^N e_k \, $; thus we find a
condition $ \, \sum_{s=1}^N (f_s + e_s) = 1 \, $; on the other hand, these terms must
have degree $ \, deg \left( \fhat^\varphi_\sigma \right) + deg \left(
\ehat^\varphi_\sigma \right) = - \alpha_i \, $: we conclude that $ \,
\fhat^\varphi_\sigma = F^\varphi_i \, $ and $ \, \ehat^\varphi_\sigma = 1 \, $. Now
remark that $ {\Cal F}_1 $ takes non-zero values only on PBW monomials (of $
\uqphiPgtilde $) like $ \, \ebar^\varphi_i \cdot L_\lambda \, $, $ \lambda \in P \, $;
let $ V_{1,i} $ be the $ \kqqm $-span of such monomials: this is a free $ \uqphiPt
$-module of rank 1; then direct computation shows that $ \, {\Cal F}_1 + q_i^{-2}
\cdot F^\varphi_i K_{-(1+\varphi) (-\alpha_i)} \otimes K_{(1-\varphi) (-\alpha_i)} \, $
is zero in $ {(V_{1,i})}^* \, $: we conclude that $ \, {\Cal F}_1 = - q_i^{-2} \cdot
F^\varphi_i K_{-(1+\varphi) (-\alpha_i)} \otimes K_{(1- \varphi) (-\alpha_i)} \, $,
hence we can write (for all $ \, i= \unon \, $)
$$ S \left( F^\varphi_i \otimes 1 \right) \equiv - q^{(\alpha_i \vert \alpha_i + \tau_i)} \cdot F^\varphi_i K_{-(1+\varphi) (-\alpha_i)} \otimes K_{(1-\varphi)(-\alpha_i)} \eqno \mod \,
{(q-1)}^2 \; . $$
\indent With similar arguments we find that (for all
$ \, i= \unon $, $ c \in \Z $, $ t \in \N \, $)
$$ \eqalignno{
S & \left( K_{-(1+\varphi)(\pm\alpha_i)} \otimes K_{(1-
\varphi)(\pm\alpha_i)} \right) \equiv K_{-(1+\varphi)
(\mp\alpha_i)} \otimes K_{(1-\varphi)(\mp\alpha_i)} &
\mod {(q-1)}^2 \cr
S & \left( \left( K_{-(1+\varphi)(\alpha_i)} \otimes
K_{(1-\varphi)(\alpha_i)} ; c \atop t \right) \right) \equiv
- K_{-(1+\varphi)(-\alpha_i)} \otimes K_{(1-\varphi)(-
\alpha_i)} \, \cdot & {} \cr
& \quad \cdot \left( K_{-(1+\varphi)(\alpha_i)} \otimes
K_{(1-\varphi)(\alpha_i)}) ; c \atop t \right) & \mod {(q-1)}^2 \cr
S & (1 \otimes E^\varphi_i) \equiv - q^{(\alpha_i \vert
\alpha_i - \tau_i)} \cdot K_{-(1+\varphi)(-\alpha_i)}
\otimes K_{(1-\varphi)(-\alpha_i)} E^\varphi_i & \mod {(q-1)}^2 \cr } $$
\indent The comultiplication $ \, \Delta \colon {D_{q,\varphi}^Q (\gerg)}^*
\rightarrow {D_{q,\varphi}^Q (\gerg)}^* \otimeshat {D_{q,\varphi}^Q (\gerg)}^* \, $ is
by definition the transpose of the multiplication of $ D_{q,\varphi}^Q (\gerg) $, hence
it is characterized by $ \, \langle \Delta(x^*), y \otimes z \rangle = \langle x^*, y
\cdot z \rangle \, $, for all $ \, x^* \in {D_{q,\varphi}^Q (\gerg)}^* $, $ y, z \in
D_{q,\varphi}^Q (\gerg) $. Now consider $ \, F^\varphi_i \otimes 1 \in \uqphiPbm
\otimeshat \uqphiPbp \leq {D_{q,\varphi}^Q (\gerg)}^* \, $: it is homogeneous of
degree $ \, -\alpha_i \, $, whence $ \Delta \left( F^\varphi_i \otimes 1 \right) $ has
the same degree. Thus writing $ \Delta \left( F^\varphi_i \otimes 1 \right) $
as a series \hfill\break
\centerline { $ \Delta \left( F^\varphi_i \otimes 1 \right) = \sum_\sigma s_\sigma
\cdot \left( F^\varphi_\sigma L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
E^\varphi_\sigma \right) \otimes \left( {F^\varphi_\sigma}' {L^{\varphi,-}_\sigma}'
\otimes {L^{\varphi,+}_\sigma}' {E^\varphi_\sigma}' \right) $ }
(where $ \, F^\varphi_\sigma L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
E^\varphi_\sigma \, $ and $ \, {F^\varphi_\sigma}' {L^{\varphi,-}_\sigma}' \otimes
{L^{\varphi,+}_\sigma}' {E^\varphi_\sigma}' \, $ are PBW monomials) we have $ \,
c_\sigma \neq $\break
$ \neq 0 \, $ only if $ \, deg((F^\varphi_\sigma L^{\varphi,-}_\sigma
\otimes L^{\varphi,+}_\sigma E^\varphi_\sigma) \otimes
({F^\varphi_\sigma}' {L^{\varphi,-}_\sigma}' \otimes
{L^{\varphi,+}_\sigma}' {E^\varphi_\sigma}')) :=
deg(F^\varphi_\sigma) + deg(E^\varphi_\sigma) + $\break
$ + deg({F^\varphi_\sigma}') + deg({E^\varphi_\sigma}') = -
\alpha_i \, $. Applying this argument to $ \aphiPinfhat $ and proceeding like for the
antipode, we get
$$ \Delta \left( \fbar^\varphi_{\alpha_i} \otimes 1 \right) = \sum_\sigma c_\sigma^-
\cdot \left( \fbar^\varphi_\sigma L^{\varphi,-}_\sigma \otimes L^{\varphi,+}_\sigma
\ebar^\varphi_\sigma \right) \otimes \left( {\left( \fbar^\varphi_\sigma \right)}'
{\left( L^{\varphi,-}_\sigma \right)}' \otimes {\left( L^{\varphi,+}_\sigma \right)}'
{\left( \ebar^\varphi_\sigma \right)}' \right) \eqno (5.13) $$
where $ \, c_\sigma^- \in \kqqm \, $ and the occurring PBW monomials have degree
$ \, -\alpha_i \, $. Similarly
$$ \Delta \! \left(\! {(F^\varphi_i)}^{\!(f)} \!\otimes\! 1 \!\right) = \sum_\sigma
a_\sigma^- \!\cdot\! \left( \fhat^\varphi_\sigma (K^{\varphi,-}_\sigma) \!\otimes\!
(K^{\varphi,+}_\sigma) \ehat^\varphi_\sigma \right) \!\otimes\! \left( \hbox{$
\fhat^\varphi_\sigma $}' ({K^{\varphi,-}_\sigma}') \!\otimes\!
({K^{\varphi,+}_\sigma}') \hbox{$ \ehat^\varphi_\sigma $}' \right) \eqno (5.14) $$
where $ \, a_\sigma^- \in \kqqm \, $ and the occurring PBW monomials have degree
$ \, -f \alpha_i \, $. Now comparison between (5.13) and (5.14) gives again a relation
$ \, a_\sigma^- = {(q - q^{-1})}^{\text{e}_\sigma} \cdot b_\sigma \cdot c_\sigma^-
\, $ where $ \, b_\sigma \in \kqqm \, $ and $ \, \text{e}_\sigma := \sum_{s=1}^N (f_s
+ e_s + f'_s + e'_s) \, $, with $ \, \fhat^\varphi_\sigma = \prod_{h=N}^1
{(F^\varphi_{\alpha^h})}^{(f_h)} \, $, $ \, \ehat^\varphi_\sigma = \prod_{k=1}^N
{(E^\varphi_{\alpha^k})}^{(e_k)} \, $ and $ \, {(\fhat^\varphi_\sigma)}' =
\prod_{h=N}^1 {(F^\varphi_{\alpha^h})}^{(f'_h)} \, $, $ \, {(\ehat^\varphi_\sigma)}' =
\prod_{k=1}^N {(E^\varphi_{\alpha^k})}^{(e'_k)} \, $. Therefore
$$ \Delta \! \left(\! {(F^\varphi_i)}^{(f)} \!\otimes\! 1 \!\right) \! = \!
\sum_{\kappa=0}^\infty {(q - q^{-1})}^\kappa d_\kappa \!\cdot\! \left(\!
\fhat^\varphi_\kappa (K^{\varphi,-}_\kappa) \!\otimes\! (K^{\varphi,+}_\kappa)
\ehat^\varphi_\kappa \!\right) \!\otimes\! \left(\! \hbox{$ \fhat^\varphi_\kappa $}'
\! ({K^{\varphi,-}_\kappa}') \!\otimes\! ({K^{\varphi,+}_\kappa}') \hbox{$
\ehat^\varphi_\kappa $}' \!\right) $$
($ d_\kappa \in \kqqm , \; \forall \, \kappa \!\in\! \N $), a convergent series in the
$ (q-1) $-adic topology of $ \aphiQinftilde \otimeshat \aphiQinftilde $. Similarly, we
find that $ \Delta \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)}
\right) $, $ \Delta \left( K_{-(1+\varphi)(-\alpha_i)} \otimes
K_{(1-\varphi)(-\alpha_i)} \right) $,\break
$ \Delta \left( \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} ; c
\atop t \right) \right) $ and $ \Delta \left( 1 \otimes {(E^\varphi_i)}^{(e)} \right) $
are expressed by convergent series in the $ (q-1) $-adic topology of $ \aphiQinftilde
\otimeshat \aphiQinftilde $. In particular, all such series are in fact {\it finite\/}
sums modulo $ (q-1) $, whose first terms are given (proceeding like for $ S $) by
$$ \eqalignno{
\!\!\!\!\! \Delta & \left(F^\varphi_i \otimes 1 \right) \equiv \left( F^\varphi_i \otimes
1 \right) \otimes \left( 1 \otimes 1 \right) + & {} \cr
& + \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} \right) \otimes
\left( F^\varphi_i \otimes 1 \right) & \mod {(q-1)}^2 \cr
\!\!\!\!\! \Delta & \big( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} \big)
\equiv & {} \cr
& \equiv \big( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} \big) \otimes \big(
K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} \big) & \mod {(q-1)}^2 \cr
\!\!\!\!\! \Delta & \big( K_{-(1+\varphi)(-\alpha_i)} \otimes K_{(1-\varphi)(-\alpha_i)} \big)
\equiv & {} \cr
& \equiv \big( K_{-(1+\varphi) (-\alpha_i)} \!\otimes\! K_{(1-\varphi)(-\alpha_i)} \big) \! \otimes \!
\big( K_{-(1+\varphi)(-\alpha_i)} \!\otimes\! K_{(1-\varphi)(-\alpha_i)} \big) & \mod {(q-1)}^2
\cr
\!\!\!\!\! \Delta & \bigg( \bigg( {K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} ; 0 \atop 1} \bigg) \bigg) \equiv & {} \cr
& \equiv \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} ; 0 \atop 1 \right)
\otimes (K_{-(1+\varphi)(\alpha_i)} \! \otimes \! K_{(1- \varphi)(\alpha_i)}) + & {} \cr
& {\ } \quad + (1 \otimes 1) \otimes \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} ; 0 \atop 1 \right) & \mod
{(q-1)}^{\phantom{2}} \cr } $$
$$ \eqalignno{
\!\!\!\!\! \Delta & \left( 1 \otimes E^\varphi_i \right) \equiv (1 \otimes 1) \otimes \left( 1
\otimes E^\varphi_i \right) + & {} \cr
& + \left( 1 \otimes E^\varphi_i \right) \otimes \left(
K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)}
\right) & \mod {(q-1)}^2 \cr } $$
(for all $ \, i= \unon $); moreover, for later use we also
record that the same analysis gives
$$ \displaylines{
\Delta \left(\! K_{-(1+\varphi)(\alpha_i)} \!\otimes\! K_{(1-\varphi)(\alpha_i)} \!\right)
\!\equiv\! \left(\! K_{-(1+\varphi) (\alpha_i)} \!\otimes\! K_{(1-\varphi)(\alpha_i)} \!\right)
\!\otimes\! \left(\! K_{-(1+\varphi)(\alpha_i)} \!\otimes\! K_{(1-\varphi)(\alpha_i)} \!\right) +
\hfill \cr
{\ } \ + \sum_{s=1}^N {(q_{\alpha^s} -
q_{\alpha^s}^{-1})}^2 \! \cdot [\langle \alpha_i,
\alpha^s \rangle]_{q_{\alpha^s}} \! \cdot \left( K_{-
(1+\varphi)(\alpha^s)} \otimes K_{(1-\varphi)(\alpha^s)}
E^\varphi_{\alpha^s} \right) \otimes \cr
{\ } \quad \hfill \otimes \left( F^\varphi_{\alpha^s} K_{-(1+\varphi)(\alpha^s)} \otimes
K_{(1-\varphi)(\alpha^s)} \right) \; \qquad \hfill \mod {(q-1)}^4 \cr } $$
from which we immediately get
$$ \eqalign{
\Delta \bigg( & \bigg( {K_{-(1+\varphi)(\alpha_i)}
\otimes K_{(1-\varphi)(\alpha_i)} ; 0 \atop 1} \bigg) \bigg) = \cr
& = \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-
\varphi)(\alpha_i)} ; 0 \atop 1 \right) \otimes \left( K_{-(1+\varphi)(\alpha_i)} \otimes
K_{(1-\varphi)(\alpha_i)} \right) + \cr
& + (1 \otimes 1) \otimes \left( K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1-\varphi)(\alpha_i)} ; 0
\atop 1 \right) + \cr
& + \sum_{s=1}^N (d_{\alpha^s})_q \, (q_{\alpha^s} - 1)
{(1 + q_{\alpha^s}^{-1})}^2 \cdot [\langle \alpha_i, \alpha^s \rangle]_{q_{\alpha^s}} \cdot
\big( K_{-(1+\varphi)(\alpha^s)} \otimes \cr
& \otimes K_{(1-\varphi)(\alpha^s)} E^\varphi_{\alpha^s} \big) \otimes
\big( F^\varphi_{\alpha^s} K_{-(1+\varphi)(\alpha^s)} \otimes K_{(1-\varphi)(\alpha^s)} \big) \;
\qquad \hfill \mod {(q-1)}^3 \, . \cr } $$
\vskip7pt
{\it Remark:\/} in the Appendix we shall explicitly write down the formulas for $ S $ and
$ \Delta $ in the case of $ SL(2,k) $. In any case the technique we outlined here can be
succesfully followed in order to compute the higher order terms of the series above up to any
fixed degree.
\vskip7pt
The computations above have an important consequence; first a definition:
\vskip7pt
\proclaim{Definition 5.16} We call $ \aphiQqinftilde $ the $ (q-1) $-adic completion (which is
naturally embedded into $ \aphiQinftilde $) of the $ \kqqm $-algebra $ \aphiQtilde $ defined in
\S 11.9. $ \square $
\endproclaim
\vskip7pt
The previous results about the (formal) Hopf structure of $ \aphiQinftilde $ then imply
$$ \aphiQqinftilde \hbox{ \it \ is a formal Hopf subalgebra of \ } \aphiQinftilde \hbox{ \it \ over \ } \kqqm \, . \eqno (5.15) $$
\indent Moreover, we can improve Theorem 5.13 {\it (b2)\/} by the following
\vskip7pt
\proclaim{Theorem 5.17} The algebra
$ \fqphiQgtilde\left[\psi_{-\delta}^{-1}\right] $ is naturally embedded into $ \fqphiQgtildeqinf $; it follows that the $ \kqqm $-algebra isomorphism $ \, \mu^{\varphi,Q} \colon \fqphiQgtilde \left[\psi_{-\delta}^{-1}\right] @>{\cong}>> \aphiQtilde \, $ (cf.~Theorem 5.13 (b1)) continuosly extends to an isomorphism of topological Hopf algebras over $ \kqqm $
$$ \mu^{\varphi,Q}_{(q-1),\infty} \colon
\fqphiQgtildeqinf @>{\cong}>> \aphiQqinftilde \; . $$
\endproclaim
\demo{Proof} Let consider the case $ \, \varphi = 0 \, $.
From the proof of Theorem 5.13 we recall that \hfill\break
\centerline{ $ \mu^Q (\psi_{-\delta}) = K_{-\delta}
\otimes K_\delta = \prod_{i=1}^n K_i^{-1} \otimes K_i =
\prod_{i=1}^n \mu^Q (\psi_{-\alpha_i}) $ }
(cf.~(5.9)); since definitions give \hfill\break
\centerline{ $ K_i^{-1} \otimes K_i = 1 + (q_i - 1) \left( K_i^{-1} \otimes K_i ; 0 \atop 1 \right) = 1 - (-1) (q_i - 1) \left( K_i^{-1} \otimes K_i ; 0 \atop 1
\right) $ }
we have \hfill\break
$$ \eqalign{
{(K_i^{-1} \otimes K_i)}^{-1} = \left( 1 - (-1) (q_i - 1) \left( K_i^{-1} \otimes K_i ; 0 \atop 1 \right) \right)^{-1} = & \cr
= \sum_{n=0}^\infty {(-1)}^n {(q_i - 1)}^n \cdot \left(
K_i^{-1} \otimes K_i ; 0 \atop 1 \right)^n & \in \aQqinftilde \cr } \eqno (5.16) $$
but extending $ \mu^Q $ to $ \fqQgtildeqinf $ one
gets\hfill\break
\centerline{ $ \mu^Q_{(q-1),\infty} \left( \sum_{n=0}^\infty {(-1)}^n {(q_i - 1)}^n \cdot \left( z_i ; 0 \atop 1 \right)^n \right) = \sum_{n=0}^\infty {(-1)}^n {(q_i - 1)}^n \cdot \left( K_i^{-1} \otimes K_i ; 0 \atop 1 \right)^n $ }
this along with (5.16) and injectivity of $ \mu^Q_\infty \colon \fqQinfgtilde @>{\cong}>> \aQinftilde \, $ (cf.~Theorem 5.13) gives
$$ \psi_{-\delta}^{-1} = \sum_{n=0}^\infty {(-1)}^n {(q_i
- 1)}^n \cdot \left( z_i ; 0 \atop 1 \right)^n \in
\fqQgtildeqinf \; . $$
\indent The same argument also works in the general
case, replacing $ \, K_i^{-1} \otimes K_i \, $ with\break
$ \, K_{-(1+\varphi)(\alpha_i)} \otimes K_{(1- \varphi)(\alpha_i)} \, $. The thesis follows. $ \square $
\enddemo
\vskip2truecm
\centerline{ \bf \S~6 Quantization of $ \uhtau $ }
\vskip10pt
{\bf 6.1 The quantum algebra $ \, \uqphiinfh \, $.} \
The results we discussed in \S 5 can acquire more brightness
by means of an axiomatic approach: to this end, we introduce now a new object $ \uqphiMinfh $, by means of a presentation by generators and relations, which is with respect to $ \uhtau $ what $ U_{q,\varphi}^M (\gerg) $ is
with respect to $ \ugtau $.
\par
Fix a lattice $ M $ ($ \, Q \leq M \leq P \, $); let
$ {U_{q,\varphi}^M(\gerh)}_0 $ be the associative
\hbox{$\kq$-alge}bra with 1 with generators
$$ F^\varphi_i \, , \ L^\varphi_\mu \, , \
E^\varphi_i \quad (\mu \in M; i=\unon) $$
and relations
$$ \eqalign {
L^\varphi_0 = 1 \, , & \qquad L^\varphi_\mu
L^\varphi_\nu = L^\varphi_{\mu + \nu} \cr
L^\varphi_\mu F^\varphi_j = q^{(\alpha_j |
(1+\varphi)(\mu))} F^\varphi_j L^\varphi_\mu \, , &
\qquad L^\varphi_\mu E^\varphi_j = q^{(\alpha_j | (1-
\varphi)(\mu))} E^\varphi_j L^\varphi_\mu \cr
E^\varphi_i F^\varphi_h & - F^\varphi_h E^\varphi_i
=0 \cr
\sum_{k = 0}^{1-a_{ij}} (-1)^k q^{+c^k_{ij}} \left[ {1-a_{ij} \atop k } \right]_{q_i} & {(E^\varphi_i)}^{1-\aij-k} E^\varphi_j {(E^\varphi_i)}^k = 0 \cr
\sum_{k = 0}^{1-a_{ij}} (-1)^k q^{-c^k_{ij}} \left[
{1-a_{ij} \atop k } \right]_{q_i} & {(F^\varphi_i)}^{1-
\aij-k} F^\varphi_j {(F^\varphi_i)}^k = 0 \cr } \eqno
(6.1) $$
where $ \, c^k_{ij}:= -(k \alpha_i | \tau_j + (1-\aij-k) \tau_i) - (\alpha_j | (1-\aij-k) \tau_i) \, $
and $ \, i, j, h = \unon \, $, $ i \neq j $, $ \mu, \nu \in M $. Then let $ \, \varepsilon' \colon
{U_{q,\varphi}^M (\gerh)}_0 \rightarrow \kq \, $ be the $ \kq $-algebra morphism defined by
$$ \varepsilon' \left( F^\varphi_i \right) := 0 \, , \;
\varepsilon' \left( L^\varphi_\mu \right) := 1 \, , \;
\varepsilon' \left( E^\varphi_i \right) := 0 \, , \quad
\forall \, i=\unon, \, \mu \in M $$
and set $ \, \gerE := Ker(\varepsilon') \, $. Finally, we
define $ \uqphiMinfh $ to be the $ \gerE $-adic
completion of $ {U_{q,\varphi}^M (\gerh)}_0 $, with its
natural structure of topological $ \kq $-algebra. Roughly
speaking, this is an algebra of (non-commutative) formal series in the given generators with the
(6.1) as commutation rules; in short, $ \uqphiMinfh $ is the topological algebra over $ \kq
$ generated by $ \, F^\varphi_i $, $ L^\varphi_\mu $, $ E^\varphi_i $ $ (\mu \in M; i=\unon) \,
$ with relations (6.1). When $ \, M = Q \, $ we shall also set $ \, K^\varphi_\alpha :=
L^\varphi_\alpha \, $ for all $ \alpha \in Q \, $ (and more in general $ L^\varphi_i :=
L^\varphi_{\omega_i} $, $ K^\varphi_i := K^\varphi_{\alpha_i} $, for all $ i=\unon $).
\par
The key point now is that results in \S 5 ---
especially Proposition 5.4 --- provides a concrete
realization of the algebras $ \uqphiMinfh $, while endowing them with a structure of formal Hopf algebra too:
\vskip7pt
\proclaim{Theorem 6.2} \ The rule
$$ \nu_M^\varphi \colon \; \quad F^\varphi_i \mapsto F^\varphi_i \otimes 1 \, , \; \quad
L^\varphi_\lambda \mapsto L^\varphi_{-(1+\varphi)(\lambda)} \otimes
L^\varphi_{(1-\varphi)(\lambda)} \, , \; \quad E^\varphi_i \mapsto 1 \otimes E^\varphi_i $$
($ \, \forall \, i= \unon, \, \lambda \in M $) defines a $ \kq $-algebra isomorphism $ \,
\nu^\varphi_M \colon \uqphiMinfh @>{\cong}>> \aphiMinf \, $; then the pull-back of the Hopf
structure of $ \aphiMinf $ defines a formal Hopf algebra structure on $ \uqphiMinfh $, so that
$ \nu^\varphi_M $ is an isomorphism of formal Hopf algebras over $ \kq $. \endproclaim
\demo{Proof} Trivial from the very definitions.
$ \square $
\enddemo
\vskip7pt
{\bf 6.3 Integer forms of $ \uqphiMinfh $.} \ In this
subsection we introduce integer forms of $ \uqphiQinfh $
and $ \uqphiPinfh $; in this formal approach, it
can be done like for $ U_{q,\varphi}^M (\gerg) $ in \S 3.4.
\par
Let $ \uqphiQinfhhat $ be the complete $ \kqqm $-subalgebra of $ \uqphiQinfh $ generated
by
$$ \left\{\, {\left( F^\varphi_i \right)}^{(\ell)}, \left( K^\varphi_i; c \atop t \right),
{\left( E^\varphi_i \right)}^{(m)} \, \bigg\vert \, \ell,c,t,m \in \N; i=\unon \,\right\}
\eqno (6.2) $$
where $ \, {(F^\varphi_i)}^{(\ell)} $, $ \, \left( K^\varphi_i; c \atop t \right) \, $, and $ \,
{(E^\varphi_i)}^{(m)} \, $ are usual divided powers (we do not need the $ {(K^\varphi_i)}^{-1} $'s
among generators, because $ \, {(K^\varphi_i)}^{-1} = {\left( 1 - ( 1 - K^\varphi_i) \right)}^{-1} = $\break
$ = \sum_{n=0}^\infty {(1 - K^\varphi_i)}^n \; $); one can easily check --- for instance looking at quantum Borel subalgebras, which are naturally embedded inside
$ \uqphiQinfhhat $ as well as inside $ \uqphiQghat $, and exploiting the analogous results for $ \uqphiQghat $ --- that $ \uqphiQinfhhat $ has a PBW pseudobasis (over
$ \kqqm $)
$$ \left\{\, \prod_{r=N}^1 {(F^\varphi_{\alpha^r})}^{(m_r)} \cdot \prod_{i=1}^n \left( K^\varphi_i; 0 \atop t_i \right) \cdot \prod_{r=1}^N {(E^\varphi_{\alpha^r})}^{(n_r)} \, \bigg\vert \, m_r, t_i, n_r \in \N \; \, \forall \, r , \forall \, i \,\right\}
\; . \eqno (6.3) $$
\indent Then $ \uqphiQinfhhat $ is an integer form (over $ \kqqm $) of $\uqphiQinfh $ {\it in topological sense\/}, that is $ \, \kq \otimes_{\kqqm} \uqphiQinfhhat \, $ is dense in $ \uqphiQinfh $ (with respect to the
$ \gerE $-adic topology).
\par
Now let $ \, \uqphiPinfhtilde \, $ be the complete \hbox{$ \kqqm $-subalge}bra of $ \uqphiPinfh $ generated by
$$ \left\{ \fbar^\varphi_{\alpha^1}, \dots,
\fbar^\varphi_{\alpha^N} \right\} \cup \left\{\,
{(L^\varphi_1)}^{\pm 1}, \dots, {(L^\varphi_n)}^{\pm 1} \,\right\} \cup \left\{ \fbar^\varphi_{\alpha^1}, \dots,
\ebar^\varphi_{\alpha^N} \right\} $$
where $ \, \fbar^\varphi_{\alpha^r} := (q_{\alpha^r} -
q_{\alpha^r}^{-1}) F^\varphi_{\alpha^r} \, $, $ \,
\ebar^\varphi_{\alpha^r} := (q_{\alpha^r} - q_{\alpha^r}^{-
1}) E^\varphi_{\alpha^r} \, $, $ \forall \, r=1, \dots, N
\, $, as usual; one checks --- again via Borel subalgebras --- that $ \uqphiPinfhtilde $ has a
PBW pseudobasis (over $ \kqqm $)
$$ \left\{\, \prod_{r=N}^1 {\left( \fbar^\varphi_{\alpha^r} \right)}^{m_r} \cdot \prod_{i=1}^n
{(L^\varphi_i)}^{t_i} \cdot \prod_{s=1}^N {\left( \ebar^\varphi_{\alpha^s} \right)}^{n_s} \, \bigg\vert \, m_r, t_i, n_s \in \N \; \forall \, r, i, s \,\right\} $$
therefore $ \uqphiPinfhtilde $ is a $ \kqqm $-integer form of $ \uqphiPinfh $ (in topological
sense).
\par
Finally, we define $ \uqphiQhhatqinf $ to be the $ (q-
1) $-adic completion of the $ \kqqm $-subalgebra of
$ \uqphiQinfh $ generated by the set in (6.2); then
$ \uqphiQhhatqinf $ is a $ \kqqm $-subalgebra of
$ \uqphiQinfhhat $, with (6.3) as a PBW pseudobasis (as
a topological $\kqqm$-modu-\break
le); notice also that $ \, {(K^\varphi_i)}^{-1} \in \uqphiQinfhhat \, $, for $ \, {(K^\varphi_i)}^{-1}
= {\left( 1 - (1 - K^\varphi_i) \right)}^{-1} = $\break
$ = \sum_{n=0}^\infty {(1 - K^\varphi_i)}^n = \sum_{n=0}^\infty {(-1) (q_i - 1) \left( K^\varphi_i ;
0 \atop 1 \right)}^{\! n} \in \uqphiQhhatqinf \, $. In particular\break
$ \uqphiQhhatqinf $ is an integer form of $ \uqphiQinfh $ (in topological sense). In the sequel we
shall also use the notation $ \, \uqphiQhhat := \uqphiQhhatqinf \, $.
\par
Now from the very definitions, Proposition 5.9, Definition 5.16, and (5.15), we easily get the
following (notice the reversal of accents "$ \, \, \widehat{\ } \, \, $" and "$ \, \, \widetilde{\ }
\, \, $"):
\vskip7pt
\proclaim{Proposition 6.4} \hfill\break
\indent The formal Hopf algebra isomorphism
$ \, \nu^\varphi_Q \colon \uqphiQinfh {\buildrel \cong \over
\rightarrow} \aphiQinf \, $, resp.~$ \, \nu^\varphi_P \colon \uqphiPinfh {\buildrel \cong \over
\rightarrow} \aphiPinf \, $, resp.~$ \, \nu^\varphi_Q \colon \uqphiQinfh {\buildrel \cong \over
\rightarrow} \aphiQinf \, $, maps $ \uqphiQinfhhat $, resp.~$ \uqphiPinfhtilde $,
resp.~$ \uqphiQhhatqinf $, isomorphically onto $ \aphiQinftilde $, resp.~$ \aphiPinfhat $,
resp.~$ \aphiQqinftilde $; in particular, $ \uqphiQinfhhat $, resp.~$ \uqphiPinfhtilde $,
resp.~$ \uqphiQhhatqinf $, is a formal Hopf subalgebra, hence a (topological) $ \kqqm $-form of the formal Hopf algebra $ \uqphiQinfh $, resp.~$ \uqphiPinfh $,
resp.~$ \uqphiQinfh $. $ \square $
\endproclaim
\vskip7pt
{\bf 6.5. Presentation of $ \uqphiQhhat $.} \ Exploiting the analogous result for
$ \uqQghat $ (cf.~[DL], \S 3.4), we can present $ \, \uqphiQhhat := \uqphiQhhatqinf \, $ as the
$ (q-1) $-adic completion of the associative $\kqqm$-algebra with 1 generated by $ \,
K^\varphi_i $, $ {(K^\varphi_i)}^{-1} $, $ \left(\! K^\varphi_i; c \atop t \!\right) $,
$ {(E^\varphi_i)}^{(r)} $, $ {(F^\varphi_i)}^{(s)} \, $ ($ c \!\in\! \Z $, $ t, r, s \!\in\! \N $,
$ i=\unon $) with relations
$$ K^\varphi_i {(K^\varphi_i)}^{-1} = 1 =
{(K^\varphi_i)}^{-1} K^\varphi_i \, , \;
{(K^\varphi_i)}^{\pm 1} {(K^\varphi_j)}^{\pm 1} =
{(K^\varphi_j)}^{\pm 1} {(K^\varphi_i)}^{\pm 1} \, , \quad \quad \forall \; i, j $$
$$ {(K^\varphi_i)}^{\pm 1} \left( K^\varphi_j; c \atop t
\right) = \left( K^\varphi_j; c \atop t \right)
{(K^\varphi_i)}^{\pm 1} \, , \qquad \qquad \forall \; i, j $$
$$ \left( K^\varphi_i; c \atop 0 \right) = 0 \, , \;
(q_i -1) \left( K^\varphi_i; 0 \atop 1 \right) = K^\varphi_i
- 1 \, , \quad \qquad \forall \; i $$
$$ \left( K^\varphi_i; c \atop t \right) \left(
K^\varphi_i; c-t \atop s \right) = \left( t+s \atop t
\right)_{\! q_i} \left( K^\varphi_i; c \atop t+s \right)
\, , \qquad \qquad \forall ;, i, t, s $$
$$ \left( K^\varphi_i; c+1 \atop t \right) -q_i^t \left(
K^\varphi_i; c \atop t \right) = \left( K^\varphi_i; c \atop
t-1 \right) \, , \quad \qquad \qquad \forall \; t \geq
1 $$
$$ \left( K^\varphi_i; c \atop t \right) = \sum_{p \geq
0}^{p \leq c, t} q_i^{(c-p)(t-p)} \left( c \atop p
\right)_{\! q_i} \left( K^\varphi_i; 0 \atop t-1 \right)
\, , \qquad \qquad \forall \; c \geq 0 $$
$$ \left( K^\varphi_i; -c \atop t \right) = \sum_{p=0}^t
{(-1)}^p q_i^{-t(c+p) + p(p+1)/2} \left( p+c-1 \atop p
\right)_{\! q_i} \left( K^\varphi_i; 0 \atop t-p \right)
\, , \qquad \forall \, c \geq 1 $$
$$ \left( K^\varphi_i; c+1 \atop t \right) - \left(
K^\varphi_i; c \atop t \right) = q_i^{c-t+1} K^\varphi_i
\left( K^\varphi_i; c \atop t-1 \right) \, , \quad \qquad \forall \; t \geq 1 $$
$$ K^\varphi_i {(E^\varphi_j)}^{(p)} = q^{p (\alpha_j |
\alpha_i - 2 \tau_i)} {(E^\varphi_j)}^{(p)} \, , \quad \qquad \qquad \forall \; i, j $$
$$ K^\varphi_i {(F^\varphi_j)}^{(p)} = q^{p (\alpha_j |
\alpha_i + 2 \tau_i)} {(F^\varphi_j)}^{(p)} \, , \quad \qquad \qquad \forall \; i, j $$
$$ \left( K^\varphi_i; c \atop t \right)
{(E^\varphi_j)}^{(p)} = {(E^\varphi_j)}^{(p)} \left(
K^\varphi_i; c + p d_i^{-1} (\alpha_j | \alpha_i - 2 \tau_i)
\atop t \right) \, , \quad \qquad \forall \; i, j $$
$$ \left( K^\varphi_i; c \atop t \right)
{(F^\varphi_j)}^{(p)} = {(F^\varphi_j)}^{(p)} \left(
K^\varphi_i; c + p d_i^{-1} (\alpha_j | \alpha_i + 2 \tau_i)
\atop t \right) \, , \quad \qquad \forall \; i, j $$
$$ {(E^\varphi_i)}^{(r)} {(E^\varphi_i)}^{(s)} = {\left[
r+s \atop r \right]}_{\! q_i} {(E^\varphi_i)}^{(r+s)} \, , \quad {(E^\varphi_i)}^{(0)} = 1 \quad \qquad \forall
\; i $$
$$ {(F^\varphi_i)}^{(r)} {(F^\varphi_i)}^{(s)} = {\left[
r+s \atop r \right]}_{\! q_i} {(F^\varphi_i)}^{(r+s)} \, , \quad {(F^\varphi_i)}^{(0)} = 1 \quad \qquad \forall
\; i $$
$$ \sum_{r+s=1-\aij} {(-1)}^s q^{-(k \alpha_i | \tau_j +
(1-\aij-k) \tau_i) - (\alpha_j | (1-\aij-k) \tau_i)}
{(E^\varphi_i)}^{(r)} E^\varphi_j {(E^\varphi_i)}^{(s)} = 0
\, , \quad \forall \; i \neq j $$
$$ \sum_{r+s=1-\aij} {(-1)}^s q^{(k \alpha_i | \tau_j +
(1-\aij-k) \tau_i) + (\alpha_j | (1-\aij-k) \tau_i)}
{(F^\varphi_i)}^{(r)} F^\varphi_j {(F^\varphi_i)}^{(s)} = 0
\, , \quad \forall \; i \neq j $$
$$ {(E^\varphi_i)}^{(r)} {(F^\varphi_j)}^{(s)} =
{(F^\varphi_j)}^{(s)} {(E^\varphi_i)}^{(r)} \, , \quad \qquad \qquad \qquad \forall \, i, j \, ; $$
in fact it is enough to use generators $ \, \left( K^\varphi_i ; 0 \atop t \right) $,
$ {(E^\varphi_i)}^{(r)} $, $ {(F^\varphi_i)}^{(s)} $. Then computations in \S 5.13 --- via the
isomorphism $ \nu^\varphi_Q \, $ --- yield the following (for all $ \, i= \unon \, $):
$$ \eqalignno{
\Delta \left( F^\varphi_i \right) & \equiv F^\varphi_i \otimes 1 + K^\varphi_i \otimes
F^\varphi_i & \quad \mod {(q-1)}^2 \cr
\Delta \left( K^\varphi_i \right) & \equiv K^\varphi_i \otimes K^\varphi_i & \quad
\mod {(q-1)}^2 \cr
\Delta \left( {\left( K^\varphi_i \right)}^{-1} \right) & \equiv {\left( K^\varphi_i \right)}^{-1}
\otimes {\left( K^\varphi_i \right)}^{-1} & \quad \mod {(q-1)}^2 \cr
\Delta \left( \left( K^\varphi_i ; 0 \atop 1 \right) \right) & \equiv \left( K^\varphi_i ; 0 \atop 1
\right) \otimes K^\varphi_i + 1 \otimes \left( K^\varphi_i ; 0 \atop 1 \right) + & {} \cr
+ \sum_{s=1}^N (d_{\alpha^s})_q (q_{\alpha^s} - 1) {(1 + q_{\alpha^s}^{-1})}^2 \! & \cdot \!
[\langle \alpha_i, \alpha^s \rangle]_{q_{\alpha^s}} \! \cdot \! K^\varphi_{\alpha^s}
E^\varphi_{\alpha^s} \! \otimes \! F^\varphi_{\alpha^s} K^\varphi_{\alpha^s} & \quad
\mod {(q-1)}^3 \cr
\Delta \left( E^\varphi_i \right) & \equiv E^\varphi_i \otimes K^\varphi_i + 1 \otimes
E^\varphi_i & \quad \mod {(q-1)}^2 \cr } $$
$$ \displaylines{
\varepsilon \left( F^\varphi_i \right) = 0 \, , \hskip2cm \varepsilon \left( E^\varphi_i \right)
= 0 \cr
\varepsilon \left( K^\varphi_i \right) = 1 \, , \qquad \varepsilon \left( \left( K^\varphi_i ; 0
\atop 1 \right) \right) = 0 \, , \qquad \varepsilon \left( {(K^\varphi_i)}^{-1} \right) = 1 \cr } $$
$$ \eqalignno{
S \left( F^\varphi_i \right) & \equiv - q_i^{-2} \cdot F^\varphi_i {\left( K^\varphi_i
\right)}^{-1} & \mod {(q-1)}^2 \cr
S \left( K^\varphi_i \right) & \equiv {(K^\varphi_i)}^{-1} & \mod {(q-1)}^2 \cr
S \left( {(K^\varphi_i)}^{-1} \right) & \equiv
K^\varphi_i & \mod {(q-1)}^2 \cr
S \left( E^\varphi_i \right) & \equiv - q_i^{+2} \cdot
{(K^\varphi_i)}^{-1} E^\varphi_i & \mod {(q-1)}^2 \cr } $$
\vskip2truecm
\centerline{ \bf \S~7 Quantum Poisson pairing }
\vskip10pt
{\bf 7.1 The natural embedding $ \, \xi_\varphi^M \colon F_{q,\varphi}^M [G] @>>> \uqphiinfh \, $.} \ In this section
we fit together the results of \S 5 and \S 6 to construct perfect Hopf pairings
\ $ \uqphiMinfh \otimes U_{q,\varphi}^{M'} (\gerg) \rightarrow
\kq $ \ which --- in a sense to be explained in \S 8 --- will provide a
quantization of three classical objects: the Hopf pairings $ \, \fgtau \otimes \ugtau \rightarrow
k \, $ (or $ \, \finfgtau \otimes \ugtau \rightarrow k \, $) and $ \, \uhtau \otimes \fhtau
\rightarrow k \, $ (or $ \, \uhtau \otimes \finfhtau \rightarrow k \, $) and the Lie bialgebra
pairing $ \, \gerh^\tau \otimes \gerg^\tau \rightarrow k \, $; therefore we shall
call them "quantum Poisson pairing". We define formal Hopf
algebra embeddings
$$ \xi_\varphi^M := {(\nu^\varphi_M)}^{-1} \circ
\mu_\varphi^{M'} \colon F_{q,\varphi}^{M'} [G]
\hookrightarrow \uqphiMinfh $$
and review results of \S 5 in terms of \S 6; in particular, we single out the following:
\vskip7pt
\proclaim{Theorem 7.2} \hfill\break
\indent (a) The embedding $ \, \xi_\varphi^P \colon
\fqphiPg \hookrightarrow \uqphiPinfh \, $ induces
embeddings of algebras \hfill\break
\centerline{ $ \xi_\varphi^P \colon \fqphiPg \hookrightarrow {U_{q,\varphi}^P (\gerh)}_0 \, , \quad \quad \xi_\varphi^P \colon \fqphiPghat \hookrightarrow \uqphiPhtilde_0 \; $ }
isomorphisms of algebras \hfill\break
\centerline{ $ \xi_\varphi^P \colon \fqphiPg \left[ \psi_{-\rho}^{-1} \right] @>{\cong}>> {U_{q,\varphi}^P (\gerh)}_0 \, , \quad \quad \xi_\varphi^P \colon \fqphiPghat \left[ \psi_{-\rho}^{-1} \right] @>{\cong}>>
\uqphiPhtilde_0 \; $ }
and isomorphisms of formal Hopf algebras \hfill\break
\centerline { $ \xi_\varphi^P \colon \fqphiPinfg @>{\cong}>> \uqphiPinfh \, , \quad \quad \xi_\varphi^P \colon \fqphiPinfghat @>{\cong}>> \uqphiPinfhtilde \; ; $ }
in particular $ \fqphiPinfghat $ is an integer form of
$ \fqphiPinfg $ (in topological sense).
\hfill\break
\indent (b) The embedding $ \, \xi_\varphi^Q \colon
\fqphiQg \hookrightarrow \uqphiQinfh \, $ induces
embeddings of algebras \hfill\break
\centerline{ $ \xi_\varphi^Q \colon \fqphiQg \hookrightarrow \uqphiQh_0 \, , \quad \quad \xi_\varphi^Q \colon \fqphiQgtilde \hookrightarrow \uqphiQhhat_0 \; $ }
isomorphisms of algebras \hfill\break
\centerline{ $ \xi_\varphi^Q \colon \fqphiQg \left[ \psi_{-\delta}^{-1} \right] @>{\cong}>> \uqphiQh_0 \, , \quad \quad \xi_\varphi^Q \colon \fqphiQgtilde [\psi_{-\delta}^{-1}] @>{\cong}>> \uqphiQhhat_0 \; $ }
and isomorphisms of formal Hopf algebras \hfill\break
\centerline { $ \xi_\varphi^Q \colon \fqphiQinfg @>{\cong}>> \uqphiQinfh \, , \quad \quad \xi_\varphi^Q \colon \fqphiQinfgtilde @>{\cong}>> \uqphiQinfhhat $ }
\centerline { $ \xi_\varphi^Q \colon \fqphiQgqinf @>{\cong}>> \uqphiQhqinf \, , \quad \quad \xi_\varphi^Q \colon \fqphiQgtildeqinf @>{\cong}>> \uqphiQhhatqinf \; $ }
where $ \, \fqphiQgqinf := \kq \otimes_{\kqqm} \fqphiQgtildeqinf \, $ and $ \, \uqphiQhqinf := \kq \otimes_{\kqqm} \uqphiQhhatqinf \, $; in particular
$ \fqphiQinfgtilde $ and $ \fqphiQgtildeqinf $ are integer forms of $ \fqphiQinfg $ (in topological sense). $ \square $
\endproclaim
\vskip7pt
{\bf 7.3 The quantum Poisson pairing.} \ Since
$ \aphiMinf $ is a formal Hopf subalgebra of
$ {U_{q,\varphi}^{M'} (\gerg)}^* $ (cf.~\S 5), restriction of the evaluation pairing yields perfect Hopf pairings
$$ \langle \ \, , \ \rangle \colon \aphiPinf \otimes
\uqphiQg @>>> \kq \, , \quad \quad \langle \ \, , \ \rangle \colon \aphiQinf \otimes \uqphiPg @>>> \kq \; . $$
\indent Exploiting the isomorphisms $ \nu^\varphi_M $
we then fix the following
\vskip7pt
\proclaim{Definition 7.4} We define {\bf quantum Poisson
pairing\/} the perfect pairing of formal Hopf algebras \
$ \pi^\varphi_q \colon \uqphiMinfh \otimes U_{q,\varphi}^{M'}(\gerg) \rightarrow \kq $ \ given by \ $ \pi^\varphi_q (h,g) := \langle {(\nu^\varphi_M)}^{-1} (h), g \rangle $ \ for all $ \, h \in \uqphiMinfh $, $ g \in U_{q,\varphi}^{M'}(\gerg) \, $. $ \square $
\endproclaim
\vskip7pt
In next paragraph we will show that this provides a
quantization of the "classical" Poisson pairing $ \pi_{\Cal
P}^\tau \colon \gerh^\tau \otimes \gerg^\tau \longrightarrow
k \, $, whence our choice of the name; for the time being,
we remark only that the previous analysis ensures that integer forms --- $ \uqphiPinfhtilde $,
$ \uqphiQinfhhat $, $ \uqphiQghat $ and $ \uqphiPgtilde $
--- of our quantized enveloping algebras are the orthogonal of each other with respect to $ \pi^\varphi_q $;
furthermore, $ \pi^\varphi_q $ restrict to perfect Hopf pairings
$$ \pi^\varphi_{q,H^\tau} \colon \uqphiQinfhhat \otimes
\uqphiPgtilde \rightarrow \kqqm \, , \qquad
\pi^\varphi_{q,G^\tau} \colon \uqphiPinfhtilde \otimes
\uqphiQghat \rightarrow \kqqm \, ; $$
with the same symbol $ \pi^\varphi_{q,G^\tau} $ we shall
also denote the Hopf pairings\break
$ \, \pi^\varphi_{q,G^\tau} \colon \fqphiPghat \otimes
\uqphiQghat \longrightarrow \kqqm \, $,
resp.~$ \pi^\varphi_{q,H^\tau} \colon \fqphiQgtilde \otimes
\uqphiPgtilde \longrightarrow \kqqm \, $, gotten from
restriction of $ \, \uqphiPinfhtilde \otimes \uqphiQghat
\rightarrow \kqqm \, $, resp.~$ \, \uqphiQinfhhat \otimes \uqphiPgtilde @>>> \kqqm \, $ (hereafter we identify
$ F_{q,\varphi}^M [G] $, with its image in $ \uqphiMinfh $ via $ \xi_\varphi^M $, and similarly for integer forms); we shall also use the same notations when $ \uqphiQhhat $ replaces $ \uqphiQinfhhat $.
\vskip2truecm
\centerline{ \bf \S~8 Specialization at roots of 1 }
\vskip10pt
{\bf 8.1 The case $ \, q \rightarrow 1 \, $: specialization of $ \uqphiQhhat $ to $ \uhtau $.} \
As in \S 3.5, we assume for simplicity that our ground field $ k $ contains all roots of unity (so
that $ \, \kqqm \Big/ (q - \varepsilon) = k \, $ for any root of unity $ \varepsilon $; otherwise,
specializing $ q $ to $ \varepsilon $ will extend the ground field to $ \, \kqqm \Big/ p_\ell (q) =
k(\varepsilon) \, $, where $ p_\ell (q) $ is the
$ \ell $-th cyclotomic polynomial); in particular
this is the case for $ \, k = \C \, $. Moreover, we fix
$ \, \tau = (\tau_1, \ldots, \tau_n) := {1 \over 2} (\varphi(\alpha_1), \ldots, \varphi(\alpha_n)) \, $.
\par
To begin with, we consider specialization of $ \, \uqphiQhhat := \uqphiQhhatqinf \, $ at
$ \, q=1 \, $: set
$$ \uunophiQhhat := \, \uqphiQhhat \Big/ (q-1) \uqphiQhhat \, \cong \, \uqphiQhhat \otimes_{\kqqm} k $$
($ k $ is a $ \kqqm $-algebra via $ \, k \cong \kqqm \Big/ (q-1) \, $); let $ \, p^\varphi_1 \colon
\uqphiQhhat \twoheadrightarrow \uunophiQhhat \, $ be the canonical map and set $ \, \text{f}^\tau_i \!:=\! p^\varphi_1 \! \left( {(F^\varphi_i)}^{(1)} \right) $,
$ \text{h}^\tau_i \!:=\! p^\varphi_1 \! \left( \! \left( K^\varphi_i ; 0 \atop 1 \right) \! \right) $,
$ \text{e}^\tau_i \!:=\! p^\varphi_1 \! \left(
{(E^\varphi_i)}^{(1)} \right) \, $ ($ i=\unon $).
Then one immediately finds that $ \, \uunophiQhhat \, $ is
cocommutative, hence inherits from $ \uqphiQhhat $ a
canonical co-Poisson Hopf structure; furthermore, it also
inherits from $ \uqphiQhhat $ a presentation by generators
($ \, \text{f}^\tau_i, \text{h}^\tau_i, \text{e}^\tau_i
\, $) and relations which is exactly the same as
$ \uhtau $ given in (1.3--5). To give an example, we
consider the case $ \, \varphi = 0 \, $ (hence $ \, \tau
= 0 \, $) and compute $ \, \delta (\text{h}_i) \, $;
by definition of quantization of co-Poisson structure,
$ \delta $ is given by $ \, \delta := {\Delta -
\Delta^{op} \over q-1} \, $ (where $ \,
\Delta^{op} \, $ denotes opposite comultiplication),
whence we have
$$ \delta(h_i) := {\Delta \left( \left( K_i ; 0 \atop 1
\right) \right) - \Delta^{op} \left( \left( K_i ; 0 \atop 1 \right) \right)
\over q-1}{\Bigg\vert}_{q=1} = \hfill \quad {\ } $$
$$ = \bigg( \sum_{s=1}^N (d_{\alpha^s})_q^2 \cdot {(1 +
q_{\alpha^s}^{-1})}^2 \cdot [\langle \alpha_i, \alpha^s
\rangle]_{q_{\alpha^s}} \cdot (K_{\alpha^s} E_{\alpha^s}
\otimes F_{\alpha^s} K_{\alpha^s} - F_{\alpha^s}
K_{\alpha^s} \otimes K_{\alpha^s} E_{\alpha^s})
\bigg){\Bigg\vert}_{q=1} = $$
$$ \displaylines{
= \sum_{s=1}^N d_{\alpha^s}^2 \cdot 2^2 \cdot \langle
\alpha_i, \alpha^s \rangle \cdot (\text{e}_{\alpha^s}
\otimes \text{f}_{\alpha^s} - \text{f}_{\alpha^s} \otimes
\text{e}_{\alpha^s}) = \cr
= 4 d_i^{-1} \cdot \sum_{s=1}^N d_{\alpha^s}^2 \cdot
(\alpha^s \vert \alpha_i) \cdot (\text{e}_{\alpha^s} \otimes
\text{f}_{\alpha^s} - \text{f}_{\alpha^s} \otimes
\text{e}_{\alpha^s}) \qquad \text{q.~e.~d.} \cr } $$
\indent Thus we have proved the following
\vskip7pt
\proclaim{Theorem 8.2} The formal topological Hopf algebra
$ \uqphiQhhat $ specializes for $ \, q \rightarrow 1 \, $ to the Poisson Hopf coalgebra
$ \uhtau \, $, i.~e.~there exists an isomorphism of Poisson Hopf coalgebras
$$ \uunophiQhhat \cong \uhtau \, . \; \square $$
\endproclaim
\vskip7pt
By the way, this also makes it possible to recover
without effort the dual result (3.8) $ \, \uqphiPgtilde @>{q \rightarrow 1}>> \fhtau \, $
(cf.~\S 3.5), whose original proof --- more precisely, the proof of the part about Poisson structures --- exhibited in [DKP] and [DP] is lenghty involved and complicated, requiring very hard computations; namely, we have the following corollary:
\vskip7pt
\proclaim{Corollary 8.3} The Hopf algebra
$ \uqphiPgtilde $ specializes to the Poisson Hopf algebra
$ \fhtau $ for $ \, q \rightarrow 1 \, $; in other words there exists an isomorphism of Poisson
Hopf algebras
$$ \uunophiPgtilde := \, \uqphiPgtilde \otimes_{\kqqm} k \, \cong \, \fhtau \, . $$
\endproclaim
\demo{Proof} Since $ \uunophiPgtilde $ is perfectly
paired with $ \, \uunophiQhhat \cong \uhtau \, $ and the
latter is cocommutative, the former is commutative; thus
$ \uunophiPgtilde $ is a commutative Hopf algebra,
finitely generated over $ k $, hence it is the function
algebra of an affine algebraic group, say
$ {\Cal H}^\tau $; moreover, from its deformation
$ \uqphiPgtilde $ the Hopf algebra $ \, \uunophiPgtilde =
F[{\Cal H}^\tau] \, $ inherits --- in the usual way ---
a Poisson structure, whence $ {\Cal H}^\tau $ is a Poisson
group. As in [DP] the presentation of $ \uqphiPgtilde $
makes it clear that $ \, (\, \uunophiPgtilde = \,) \, F[{\Cal H}^\tau] \cong F[H^\tau] \, $ as Hopf
algebras, hence $ \, {\Cal H}^\tau = H \, $ as algebraic groups (this is the trivial part, also in [DP],
[DKP]). Moreover, the Hopf pairing among $ \, \uunophiQhhat \cong \uhtau \, $ and $ \,
\uunophiPgtilde = F[{\Cal H}^\tau] = F[H^\tau] \, $ is compatible with Poisson and co-Poisson
structures, that is \ $ \langle h, \{f,g\} \rangle = \langle \delta(h), f \otimes g \rangle \, $ for
all $ \, h \in \uunophiQhhat $ \ and $ \, f, g \in \uunophiPgtilde \, $, where $ \delta $ is the
Poisson co-bracket of $ \uunoQhhat $ and $ \{\,\ ,\ \,\} $ can be either the Poisson bracket $
{\{\,\ ,\ \,\}}_\star $ of $ F[H^\tau] $ (giving the Poisson structure of $ H^\tau \, $) or the
Poisson bracket $ {\{\,\ ,\ \,\}}_\circ $ of $ \, F[{\Cal H}^\tau] = \uunophiPgtilde \, $ (arising
from its quantization $ \uqphiPgtilde $): since the pairing is perfect, we must have identity $ \,
{\{\,\ ,\ \,\}}_\star = {\{\,\ ,\ \,\}}_\circ \, $, whence the thesis. $ \square $ \enddemo
\vskip7pt
{\bf 8.4 The case $ \, q \rightarrow 1 \, $:
specialization of $ \fqphiQgtilde $ and
$ \fqphiQgtildeqinf $ to $ \uh $.} \ Consider the
isomorphism \ $ \xi_\varphi^Q \colon \fqphiQgtildeqinf @>{\cong}>> \uqphiQhhatqinf =:
\uqphiQhhat $ \ provided by Theorem 7.2 {\it (b)\/}; specializing $ q $ to 1 we get a Hopf
isomorphism (over $ k $)
$$ {(\xi_\varphi^Q)}_1 \colon {\fqphiQgtildeqinf \Big/
(q-1) \fqphiQgtildeqinf} @>{\cong}>> \uunophiQhhat \; ;
\eqno (8.1) $$
in particular, we remark that now both $ {\fqphiQgtildeqinf
\Big/ (q-1) \fqphiQgtildeqinf} $ and $ \uunophiQhhat $
are {\it non-formal\/} Hopf algebras; furthermore, the
very definitions imply that
$$ {\fqphiQgtildeqinf \Big/ (q-1) \fqphiQgtildeqinf} =
\funophiQgtilde $$
where $ \, \funophiQgtilde := {\fqphiQgtilde} \Big/ {(q-1)
\fqphiQgtilde} \, $, hence (8.1) coincides with the
specialization
$$ {(\xi_\varphi^Q)}_1 \colon \funophiQgtilde :=
{\fqphiQgtilde \Big/ (q-1) \fqphiQgtilde} \hookrightarrow
\uunophiQhhat \eqno (8.2) $$
of $ \, \xi_\varphi^Q \colon \fqphiQgtilde \hookrightarrow
\uqphiQhhat \, $, thus also (8.2) is an isomorphism; clearly, both (8.1) and (8.2) are isomorphisms
of Poisson Hopf coalgebras. Thus from $ \, \uunophiQhhat \cong \uhtau \, $ we get:
\vskip7pt
\proclaim{Theorem 8.5} For $ \, q \rightarrow 1 \, $, the Hopf algebra $ \fqphiQgtilde $ and the
formal Hopf algebra\break
$ \fqphiQgtildeqinf $ both specialize to the Poisson Hopf coalgebra $ \uhtau
$. $ \square $ \endproclaim
\vskip7pt
{\bf 8.6 The case $ \, q \rightarrow 1 \, $:
specialization of $ \fqphiPinfghat $ and
$ \uqphiPinfhtilde $ to $ \finfgtau $.} \ We shall now show that $ \fqphiPinfghat $ and
$ \uqphiPinfhtilde $ are (isomorphic) quantizations of $ \finfgtau $; in particular
$ \; \uqphiPinfhtilde @>{q \rightarrow 1}>> \finfgtau \; $, can be seen as the dual counterpart
(in the sense of Poisson duality) of $ \; \uqphiPgtilde \, @>{q \rightarrow 1}>> \, \fhtau
\; $ (cf.~(3.8)).
\vskip7pt
\proclaim{Theorem 8.7} For $ \, q \rightarrow 1 \, $, the formal Hopf algebras $ \fqphiPinfghat $
and $ \uqphiPinfhtilde $ both specialize to the formal Poisson Hopf algebra $ \finfgtau $.
\endproclaim
\demo{Proof} We stick for simplicity to the case $ \,
\varphi = 0 \, $, generalization being straightforward.
\par
We provide two proofs, the first one more geometric (a consequence of (4.6)), the second one
more algebraic (a consequence of Theorem 8.2).
\par
{\it Geometric proof.\/} Recall that $ \finfg $ can be
defined as the $ m_e $-adic completion of $ {\Cal O}_e $
(where $ \, ({\Cal O}, m_e) \, $ denotes the local ring
of $ \, e \in G \, $); it is also clear that this
coincides with the $ \germ_e $-adic completion of
$ \fg $, where $ \germ_e $ is the maximal ideal of $ \,
e \in G \, $. On the other hand, $ \fqPinfghat $ is the $ \gerE $-adic completion of $ \fqPghat $,
with $ \, \gerE := Ker \left( \varepsilon \colon \fqPghat \rightarrow \kqqm \right) \, $ (cf.~\S 4.6); since
$ \, \fqPghat \, @>{q \rightarrow 1}>> \, F[G] \, $ (cf.~\S 4.5), $ \fqPinfghat $ specializes to the
$ \gerE_1 $-adic completion of $ \fg $, with $ \, \gerE_1 := Ker (\varepsilon \colon \fg
\rightarrow k) \, $; but $ \gerE_1 = \germ_e \, $, whence the thesis.
\par
{\it Algebraic proof.\/} Recall that $ \finfg $ can be
defined as the {\it full\/} linear dual of $ \ug $, that
is $ \, \finfg := {\ug}^* \, $. From the proof of Lemma 5.3 we recall that
$$ {D_q^Q(\gerg)}^* \cong \left( \uqPbm \widehat{\rtimes}
\kq[{(k^\star)}^n] \right) \otimeshat \left( \uqPbp
\widehat{\rtimes} \kq[{(k^\star)}^n] \right) \, ; $$
a quick review of the proof of Proposition 5.5 will then
convince the reader that
$$ {\uqQg}^* \cong {\left( pr^Q \right)}^* \left({\uqQg}^*\right) = U_q(\gern_-) \otimeshat \uqPt \otimeshat \kq[{\Z_2}^n] \otimeshat U_q(\gern_+)
\eqno (8.3) $$
as $ \kq $-vector spaces (with notations of Theorem 5.7). But when looking at integer forms we have \
$ {\uqQthat}^\circ = \uqPttilde \, $, \ in short because non-scalar elements of $ \kq[{(\Z_2)}^n] $ take values outside $ \kqqm $ when evaluated on $ \uqQthat $: in fact $ \, \left\langle \nu, \left( K_i ; 0 \atop 1 \right) \right\rangle = {\nu_i - 1 \over q_i-1} \notin \kqqm $ unless $ \, \nu_i = 1 \, $ (for all $ \nu \in {\Z_2}^n $, $ i=\unon \, $). Similarly we have \ $ {\uqQbphat}^\circ = \fqPbphat \cong \uqPbmtilde_{op} \, $, $ \, {\uqQbmhat}^\circ = \fqPbmhat \cong \uqPbptilde_{op} \, $, $ \, {\uqQghat}^\circ = \fqPghat \, $, \ so that (8.3) yields
$$ {\uqQghat}^* \cong {\left( pr^Q \right)}^* \left({\uqQghat}^*\right) = {{\widetilde{U}}_q (\gern_-)} \otimeshat {{\widetilde{U}}_q^P (\gert)} \otimeshat {{\widetilde{U}}_q (\gern_+)} \; ; \eqno (8.4) $$
on the other hand, we record that $ \, \uqPinfhtilde \cong \left( {\left( pr^Q \right)}^* \circ
{(\xi_q^P)}^{-1} \right) \left( \uqPinfhtilde \right) \cong $\break
$ \cong {\left( pr^Q \right)}^* \left( \fqPinfghat \right) = \aPinfhat = {{\widetilde{U}}_q
(\gern_-)} \otimeshat {{\widetilde{U}}_q^P (\gert)} \otimeshat {{\widetilde{U}}_q (\gern_+)} \, $
hence comparing with (8.4) we get
$$ {\uqQghat}^* \cong \uqPinfhtilde \; .
\eqno (8.5) $$
\indent Now specialize $ q $ to $ 1 $: then
$ \uqQghat $ specializes to $ \ug $; dualizing, specialization of $ q $ at $ 1 $ gives, in force
of (8.5), \ $ \uunoPinfhtilde := \uqPinfhtilde \otimes_{\kqqm} k \cong {\uqQghat}^*
\otimes_{\kqqm} k = $\break
$ = {\left( \uqQghat \otimes_{\kqqm} k \right)}^* = {\uunoQghat}^* \cong {\ug}^* \, $, \ i.~e.
$$ \uunoPinfhtilde \cong {\ug}^* = \finfg $$
that is $ \uqPinfhtilde $ specializes to $ \, {\ug}^* = \finfg \, $ for $ \, q \rightarrow \, $, q.~e.~d.
$ \square $
\enddemo
\vskip7pt
{\bf 8.8 The case $ \, q \rightarrow \varepsilon
\, $: quantum Frobenius morphisms.} \ Let
$ \varepsilon $ be a primitive $ \ell $-th root of 1 in
$ k $, for $ \ell $ {\it odd\/}, $ \, \ell > d:= \max_i
\{d_i\} \, $, and set
$$ \uepsilonphiQhhat := \, \uqphiQhhat \Big/ (q -
\varepsilon) \uqphiQhhat \, \cong \, \uqphiQhhat
\otimes_{\kqqm} k $$
($ k $ being a $ \kqqm $-algebra via $ \, k \cong \kqqm \Big/ (q - \varepsilon) \, $); we have the following result, which is the analog for $ \uqphiQinfh $ of a similar result about $ \uqphiQg $, which can be found in [CV-2], \S 3.2 (cf.~also [Lu-2], Theorem 8.10, and [DL], Theorem 6.3):
\vskip7pt
\proclaim{Theorem 8.9} There exists a (formal) Hopf algebra
epimorphism
$$ \frhatgerhtau \colon \uepsilonphiQhhat
\twoheadrightarrow \uunophiQhhat \cong \uhtau $$
defined by
$$ \eqalignno{
\frhatgerhtau \left( {(F^\varphi_i)}^{(s)}
\Big\vert_{q=\varepsilon} \right) & := \, {(F^\varphi_i)}^{(s
/ \ell)} \Big\vert_{q=1} & \hbox{ \ if \ } \ell \Big\vert s \, , \; \, 0 \hbox{ \, otherwise \, } \cr
\frhatgerhtau \left(\left( K^\varphi_i; 0 \atop s \right)
\bigg\vert_{q=\varepsilon} \right) & := \, \left( K^\varphi_i; 0 \atop s / \ell \right)
\bigg\vert_{q=1} & \hbox{ \ if \ } \ell \Big\vert s \, , \; \, 0 \hbox{ \, otherwise \, } \cr
\frhatgerhtau \left( {(K^\varphi_i)}^{-1} \Big\vert_{q=\varepsilon} \right) & := \, 1 & {}
\cr
\frhatgerhtau \left( {(E^\varphi_i)}^{(s)}
\Big\vert_{q=\varepsilon} \right) & := \, {(E^\varphi_i)}^{(s
/ \ell)} \Big\vert_{q=1} & \hbox{ \ if \ } \ell \Big\vert s \, , \; \, 0 \hbox{ \, otherwise \, }
\cr } $$
for all $ \, i=\unon \, $, $ s \in \N \, $. Moreover,
$ \frhatgerhtau $ is adjoint of $ \frtildegergtau $ (cf.~\S 3.5) with respect to the (specialized)
quantum Poisson pairings, viz.~(for all $ \, h \in \uepsilonphiQhhat $, $ g \in \uunophiPgtilde \, $)
$$ \pi^\varphi_{1,H^\tau} \left( \frhatgerhtau(h), g \right) = \pi^\varphi_{\varepsilon,H^\tau} \left( h, \frtildegergtau(g) \right) \, . $$
\endproclaim
\demo{Proof} Consider the Hopf algebra embedding $ \, \frtildegergtau \colon \fhtau \cong \uunophiPgtilde \hookrightarrow \uepsilonphiPgtilde \, $; its (linear) dual is an epimorphism of formal Hopf algebras
$ \, {\uepsilonPgtilde}^* \twoheadrightarrow
{\widetilde{U}_1^P (\gerg)}^* \, $; composing the latter with the natural embedding $ \, \uepsilonQhhat \hookrightarrow {\uepsilonPgtilde}^* \, $ provided by the (specialized) quantum Poisson pairing
$ \; \pi^\varphi_{\varepsilon,H^\tau} \colon \uepsilonphiQhhat \otimes \uepsilonphiPgtilde \longrightarrow k \; $ yields a morphism
$$ \frhatgerhtau \colon \uepsilonphiQhhat \longrightarrow
{\uunophiPgtilde}^* \; . $$
\indent Then the very construction gives
$$ \left\langle \frhatgerhtau(h), g \right\rangle =
\pi^\varphi_{1,H^\tau} \left( \frhatgerhtau(h), g \right) =
\pi^\varphi_{\varepsilon,H^\tau} \left( h, \frtildegergtau(g) \right) $$
for all $ \, h \in \uepsilonphiQhhat $, $ g \in
\uunophiPgtilde $, from which one immediately gets that
$ \frhatgerhtau $ is described by the formulas above and
has image $ \uunophiQhhat $, q.~e.~d. $ \square $
\enddemo
\vskip7pt
With similar arguments we also prove next result, which
is the analog of (3.9); we set
$$ \eqalign{
\uunophiPinfhtilde & := \, \uqphiPinfhtilde \Big/ (q - 1) \uqphiPinfhtilde \, \cong \, \uqphiPinfhtilde
\otimes_{\kqqm} k \cr
\uepsilonphiPinfhtilde & := \, \uqphiPinfhtilde \Big/ (q - \varepsilon) \uqphiPinfhtilde \, \cong \, \uqphiPinfhtilde \otimes_{\kqqm} k \, . \cr } $$
\vskip7pt
\proclaim{Theorem 8.10}
\hfill\break
\indent (a) There exists a formal Hopf algebra
monomorphism
$$ \frtildegerhtau \colon \finfgtau \cong \uunophiPinfhtilde \hookrightarrow
\uepsilonphiPinfhtilde $$
defined by
$$ \frtildegerhtau \colon \; \quad \fbar^\varphi_\alpha \Big\vert_{q=1} \mapsto
{(\fbar^\varphi_\alpha)}^\ell \Big\vert_{q=\varepsilon} \, , \; \quad L^\varphi_\lambda
\Big\vert_{q=1} \mapsto {(L^\varphi_\lambda)}^\ell \Big\vert_{q=\varepsilon} \, , \; \quad
\ebar^\varphi_\alpha \Big\vert_{q=1} \mapsto {(\ebar^\varphi_\alpha)}^\ell
\Big\vert_{q=\varepsilon} $$ for all $ \, \alpha \in R^+ $, $ \lambda \in P \, $. Moreover, $
\frtildegerhtau $ is the continuos extension of $ \frhatGtau $ (cf.~\S 4.5) and is adjoint of $
\frhatgergtau $ (cf.~\S 3.5) with respect to the (specialized) quantum Poisson pairings, viz.~(for
all $ \, h \in \uunophiPhtilde $, $ g \in \uepsilonphiQghat \, $)
$$ \pi^\varphi_{\varepsilon,G^\tau} \left( \frtildegerhtau(h), g \right) = \pi^\varphi_{1,G^\tau} \left( h, \frhatgergtau(g) \right) \, . $$
\hfill\break
\indent (b) The image $ \, Z_0 \, \left( \, \cong_{\frtildegerhtau} \uunophiPinfhtilde \right) $ of
$ \frtildegerhtau $ is a central formal Hopf subalgebra\break
of $ \uepsilonphiPinfhtilde $.
\hfill\break
\indent (c) The set of ordered PBW monomials
$$ \left\{\, \prod_{r=N}^1 {\left( \fbar^\varphi_{\alpha^r} \right)}^{\ell f_r} \cdot \prod_{i=1}^n {(L^\varphi_i)}^{\ell l_i} \cdot \prod_{r=1}^N {\left( \ebar^\varphi_{\alpha^r} \right)}^{\ell e_r} \, \bigg\vert \, f_1, \dots, f_N, l_1, \dots, l_n, e_1, \dots, e_N \in \N \,\right\} $$
is a pseudobasis of $ Z_0 $ (over $ k \, $).
\hfill\break
\indent (d) The set of ordered PBW monomials
$$ \left\{\, \prod_{r=N}^1
{\left( \fbar^\varphi_{\alpha^r} \right)}^{e_r} \cdot \prod_{i=1}^n {(L^\varphi_i)}^{l_i} \cdot \prod_{s=1}^N {
\left( \ebar^\varphi_{\alpha^s} \right)}^{e_s} \, \bigg\vert \, f_r, l_i, e_s = 0, 1, \ldots, \ell-1 \; \forall \, r, i, s \,\right\} $$
is a basis of $ \uepsilonphiPinfhtilde $ over $ Z_0 \, $;
thus $ \uepsilonphiPinfhtilde $ is a free module of rank
$ \ell^{dim(H)} $ over $ Z_0 \, $.
\endproclaim
\demo{Proof} As for {\it (a)\/}, consider the Hopf
algebra epimorphism $ \, \frhatgergtau \colon \uepsilonphiQghat \twoheadrightarrow
\uunophiQghat \cong \ugtau \, $; its (linear) dual is a monomorphism of formal Hopf algebras
$ \, {\uunophiQghat}^* \hookrightarrow {\uepsilonphiQghat}^* \, $; composing the latter with the
natural embedding $ \, \uunophiPinfhtilde \hookrightarrow {\uunophiQghat}^* \, $ provided by
the (specialized) quantum Poisson pairing $ \; \pi^\varphi_{1,G^\tau} \colon \uunophiPinfhtilde
\otimes \uunophiQghat @>>> k \; $ yields a morphism
$$ \frtildegerhtau \colon \uunophiPinfhtilde
\longrightarrow {\uepsilonphiQghat}^* \; . $$
Thus the very construction gives
$$ \left\langle \frtildegerhtau(h), g \right\rangle =
\pi^\varphi_{\varepsilon,G^\tau} \left( \frtildegerh(h), g \right) = \pi^\varphi_{1,G^\tau} \left( h, \frhatgerg(g) \right) $$
for all $ \, h \in \uunophiPhtilde $, $ x \in
\uepsilonQghat $: then from the definition of
$ \pi^\varphi_{q,G^\tau} $ one immediately gets that
$ \frtildegerhtau $ is described by the formulas above,
hence in particular its image is contained in
$ \uepsilonphiPhtilde $ (embedded in
$ {\uepsilonphiQghat}^* $ by means of
$ \pi^\varphi_{\varepsilon,G^\tau} $). Finally, since $ \; \frhatGtau \colon \fgtau \cong
\funophiPghat \hookrightarrow \fepsilonphiPghat \; $ is also defined as (Hopf) dual of
$ \; \frhatgergtau \colon \uepsilonphiQghat \twoheadrightarrow \uunophiQghat \cong \ugtau \; $
(cf.~[DL], Proposition 6.4, along with [CV-2], \S 3), then $ \; \frtildegerhtau \colon \finfgtau \cong
\uunophiPinfhtilde \hookrightarrow \uepsilonphiPinfhtilde \; $ is obviously an extension of
$ \; \frhatGtau \colon \fgtau \cong \funophiPghat \hookrightarrow \fepsilonphiPghat \; $, and it
is also clear that this extension is by continuity.
\par
Parts {\it (b)\/}, {\it (c)\/}, and {\it (d)\/} can be easily deduced from the analogous result
for $ \uepsilonPgtilde $ (namely [DP], Theorem 19.1 and its generalization) along with the very
definitions of $ \uqphiPg $ and $ \uqphiPinfh $ (otherwise one can directly use computations in
[DP], \S 19, or mimick the proof of Proposition 6.4 of [DL]). $ \square $
\enddemo
\vskip7pt
By the way, we remark that part {\it (d)\/} of Theorem
8.10 is well-related with [CV-2], Proposition 3.5
(extending [DL], Theorem 7.2), which claims that
$ \fepsilonphiPghat $ is a projective module over $ F_0 $
of rank $ \ell^{dim(G^\tau)} = \ell^{dim(H^\tau)} $.
\par
At last, we can prove the following counterpart of [CV-2], \S 3.3 (cf.~also [DL], Proposition 6.4
for the one-parameter case), which dealt with (4.6): notice in
particular that now we get a {\it surjective\/} morphism
instead of an {\it injective\/} one (like is in [CV-2] and
[DL]).
\vskip7pt
\proclaim{Theorem 8.11} \ There exists a Hopf algebra epimorphism
$$ \frtildehtau \colon \fepsilonphiQgtilde
\twoheadrightarrow \funophiQgtilde \cong \uhtau $$
dual of $ \; \frtildegergtau \colon \fhtau \cong
\uunophiPgtilde \hookrightarrow \uepsilonphiPgtilde \; $.
\endproclaim
\demo{Proof} Since $ \, \fepsilonphiQgtilde \leq
\uepsilonphiQhhat \, $, we can restrict $ \frhatgerhtau $
to $ \fepsilonphiQgtilde $, thus getting a Hopf algebra
morphism $ \, \frtildehtau \colon \fepsilonphiQgtilde
\longrightarrow \uunophiQhhat \cong \uhtau \, $. Now Theorem 5.13 {\it (b1)--(b2)\/} gives
(adapting notations to the multiparameter setting)
$$ \frtildehtau \left( \fepsilonphiQgtilde \right) =
\frhatgerhtau \left( \fepsilonphiQgtilde \left[
{\psi_{-\delta}^{-1}}\big\vert_{q = \varepsilon} \right]
\right) = \frhatgerhtau \left( \aphiQtilde\big\vert_{q =
\varepsilon} \right) $$
because\footnote{ Notice that the matrix coefficient
$ \psi_{-\delta} $ does not depend on $ \varphi $, but
its image $ \, \mu_\varphi^Q (\psi_{-\delta}) \in \aphiQtilde \, $ does. }
$ \psi_{-\delta}^{-1} \cong \mu_\varphi^Q (\psi_{-\delta}) =
K^\varphi_{-\delta} = \prod_{i=1}^n {\left( K^\varphi_i
\right)}^{-1} \, $ whence $ \, \frhatgerhtau \left(
{\psi_{-\delta}^{-1}}\big\vert_{q
= \varepsilon} \right) =\break
= \prod_{i=1}^n \frhatgerhtau \left(
{{\left( K^\varphi_i \right)}^{-1}}\big\vert_{q = \varepsilon} \right) = 1 \, $; moreover, it is
immediate from definitions that
$$ \frhatgerhtau \left( \aphiQtilde\big\vert_{q = \varepsilon} \right) = \frhatgerhtau \left(
\uepsilonphiQhhat \right) = \uunophiQhhat = \funophiQgtilde \; \, (\, \cong \uhtau) $$
hence we conclude that $ \, \frtildehtau \left(
\fepsilonphiQgtilde \right) = \funophiQgtilde \, $; the
thesis follows. $ \square $
\enddemo
\vskip7pt
Like in the case of $ \gerg $, we call also
$ \frhatgerhtau $, $ \frtildegerhtau $, and
$ \frtildehtau $ {\bf quantum Frobenius morphisms}, because for $ \, \ell = p \, $ prime they
can be thought of as lifting of classical Frobenius morphisms $ \, H^\tau_{\Z_p} \longrightarrow
H^\tau_{\Z_p} \, $, $ \, G^\tau_{\Z_p} \longrightarrow G^\tau_{\Z_p} \, $ to
characteristic
zero\footnote{ Here again
$ H^\tau_{\Z_p} $ denotes the Chevalley-type Poisson group-scheme over $ \Z_p $ that one can
clearly build up from the "Cartan datum" associated to $ H^\tau $ (mimicking the
usual procedure one follows for $ G_{\Z_p} \, $); as a group-scheme it is isomorphic to
$ H_{\Z_p} := H^0_{\Z_p} $, the difference residing only in the Poisson structure. }.
\vskip7pt
{\bf 8.12 Specializing the quantum Poisson pairing.} \
From the analysis in \S\S 8.2--7 it follows that the Hopf
pairing $ \, \pi^\varphi_{q,H^\tau} \colon \uqphiQinfhhat
\otimes \uqphiPgtilde \rightarrow \kqqm \, $, resp. $ \,
\pi^\varphi_{q,G^\tau} \colon \uqphiPinfhtilde \otimes
\uqphiQghat \rightarrow \kqqm \, $ specializes to the
natural Hopf pairing $ \, \pi_{H^\tau} \colon \uhtau
\otimes \fhtau \rightarrow k \, $, resp. $ \, \pi_{G^\tau}
\colon \finfgtau \otimes \ugtau \rightarrow k \, $ given by
evaluation: thus the quantum Poisson pairing can be seen as a quantization of the previous
pairings at the classical level. Now we show that it can also be seen as a quantization of
the classical Poisson pairing $ \, \pi^\tau_{\Cal P} \colon \gerh^\tau
\otimes \gerg^\tau \rightarrow k \, $.
\par
To begin with, we define a grading on $ \uqphiQinfh $ by giving to PBW monomials the degree
$$ \partial \left(\prod_{r=N}^1 {(F^\varphi_{\alpha^r})}^{(m_r)} \cdot \prod_{i=1}^n
\left( K^\varphi_i; 0 \atop t_i \right) \cdot \prod_{r=1}^N {(E^\varphi_{\alpha^r})}^{(n_r)} \right)
:= \sum_{r=N}^1 m_r + \sum_{i=1}^n t_i + \sum_{r=1}^N n_r $$
and extending by linearity (this is a grading as a $ \kq $-vector space). Then we set
$$ \pi^\varphi_{q,{\Cal P}} (h,g) := \, {(q-1)}^{\partial (h)} \cdot \pi^\varphi_q (h,g) $$
for all tensors $ \, h \otimes g \, $ of PBW monomials in
$ \, \uqphiQhhat \otimes \uqphiQghat \; \left( \, \subseteq \uqphiPinfh \otimes \uqphiQg \, \right) $,
and finally extend by linearity: this gives a perfect (linear) pairing
$ \, \pi^\varphi_{q,{\Cal P}} \colon \uqphiQinfh \otimes \uqphiQg \rightarrow \kq \, $ such that $ \,
\pi^\varphi_{q,{\Cal P}} \left( \uqphiQhhat,\uqphiQghat \right) \subseteq \kqqm_{(q-1)} \, $ (where
$ \kqqm_{(q-1)} $ is the localization of $ \kqqm $ at the principal prime ideal $ (q-1) \, $), hence
(by restriction) a pairing
$$ \pi^\varphi_{q,{\Cal P}} \colon \uqphiQhhat \otimes \uqphiQghat \rightarrow \kqqm_{(q-1)} $$
which can be specialized at $ q=1 $.
\vskip7pt
\proclaim{Theorem 8.13} The pairing $ \, \pi^\varphi_{q,{\Cal P}} \colon \uqphiQhhat \otimes
\uqphiQghat \rightarrow \kqqm_{(q-1)} \, $ specializes to a perfect linear pairing
$$ \pi^\tau_{\Cal P} \colon \uhtau \otimes \ugtau
\rightarrow k $$
which extends the Lie bialgebra pairing $ \, \pi^\varphi_{{\Cal P}} \colon \gerh^\tau \otimes
\gerg^\tau \rightarrow k \, $ (cf.~\S 1.2).
\endproclaim
\demo{Proof} It is clearly enough to check that $ \, \pi^\varphi_{q,{\Cal P}} (h,g){\big\vert}_{q=1}
= \pi^\tau_{\Cal P} \left( h{\big\vert}_{q=1}, g{\big\vert}_{q=1} \right) \, $ for Chevalley
generators and root vectors $ \, {\hbox{h}^\tau} = h{\big\vert}_{q=1} \, $ of $ \gerh^\tau $ and
$ {\hbox{g}} = g{\big\vert}_{q=1} $ of $ \gerg^\tau $, viz.~for elements
$ \, F^\varphi_\alpha{\big\vert}_{q=1} = \text{f}^\tau_\alpha \, $, $ \, \left( K^\varphi_i; \, 0
\atop 1 \right){\Big\vert}_{q=1} = \text{h}^\tau_i \, $,
$ \, E^\varphi_\alpha{\big\vert}_{q=1} = \text{e}^\tau_\alpha \, $ for $ \uqphiQhhat $
specializing to $ \uhtau $, and
$ \, F_\alpha{\big\vert}_{q=1} = f_\alpha \, $, $ \, \left(
K_i; \, 0 \atop 1 \right){\Big\vert}_{q=1} = h_i \, $, $ \,
E_\alpha{\big\vert}_{q=1} = e_\alpha \, $ for $ \uqQghat $
specializing to $ \ug $. The thesis then follows from
direct (and straightforward) computation, e.~g.~(cf.~(1.1--a/b))
$$ \pi^\varphi_{q,{\Cal P}} (F^\varphi_\alpha,
E_\beta){\big\vert}_{q=1} = (q-1) {-\delta_{\alpha \beta}
\over {(q_\alpha - q_\alpha^{-1})} }{\bigg\vert}_{q=1} = -
{1 \over 2} \delta_{\alpha \beta} d_\alpha^{-1} =
\pi^\tau_{{\Cal P}} (\text{f}^\tau_\alpha, e_\beta)
\, . \; \square $$
\enddemo
\vskip7pt
{\bf 8.14 The pairings $ \, \fgtau \otimes \fhtau
\longrightarrow k \, $, $ \, \finfgtau \otimes \fhtau
\longrightarrow k \, $.} \ The construction of \S 8.12
can be reversed as follows; we define a pairing \ $ \pi_q^{\varphi, \Cal P} \colon
\uqphiPinfhtilde \otimes \uqphiPgtilde \rightarrow \kqqm $ \ modifying
$ \pi^\varphi_q \, $: give to PBW monomials of $ \uqphiPinfhtilde $ or
$ \uqphiPgtilde $ the degree
$$ \partial \left(\prod_{r=N}^1
{(\fbar^\varphi_{\alpha^r})}^{m_r} \cdot \prod_{i=1}^n
{(L^\varphi_i)}^{l_i} \cdot \prod_{r=1}^N
{(\ebar^\varphi_{\alpha^r})}^{n_r} \right) := \sum_{r=N}^1
m_r + \sum_{i=1}^N n_r $$
and extend by linearity to define a (linear) grading of
$ \uqphiPinfh $ or $ \uqphiPg $. Now extend\break
$ \, \pi^\varphi_q \colon \uqphiPinfh \otimes \uqphiQg \rightarrow \kq \, $
(or $ \, \pi^\varphi_q \colon \uqphiQinfh \otimes \uqphiPg \rightarrow \kq \, $) to a perfect
pairing $ \, \pi^\varphi_q \colon \uqphiPinfh \otimes \uqphiPg \rightarrow
k\left(q^{1/d},q^{-{1/d}}\right) \, $ ($ \, d := det \, {(\aij)}_{i,j=\unon} \, $) by
$ \, \pi^\varphi_q (L^\varphi_\lambda, L_\mu) := q^{(\lambda \vert \mu)} \, $. Then set
$$ \pi_q^{\varphi, \Cal P} (h,g) := \, {(q-1)}^{-\partial
(h)} \cdot \pi^\varphi_q (h,g) = {(q-1)}^{-\partial (g)} \cdot \pi^\varphi_q (h,g) $$
for all $ \, h \in \uqphiPinfhtilde \, $ and $ \, g \in \uqphiPgtilde \, $ {\it
homogeneous\/}, with $ \, \partial (x) := \text{degree of } x \, $, and extend by
linearity to all of $ \, \uqphiPinfhtilde \otimes \uqphiPgtilde \, $: then $ \, \pi_q^{\varphi,
\Cal P} \left( \uqphiPinfhtilde, \uqphiPgtilde \right) \subseteq k\bigl[q^{1/d},q^{-{1/d}}\bigr]
\, $, thus we get perfect pairings
$$ \displaylines{
\pi_q^{\varphi, \Cal P} \colon \uqphiPinfhtilde \otimes \uqphiPgtilde = \fqphiPinfghat \otimes
\uqphiPgtilde \longrightarrow k\left[q^{1/d},q^{-{1/d}}\right] \cr
\pi_q^{\varphi, \Cal P} \colon \fqphiPghat \otimes \uqphiPgtilde \longrightarrow k\left[q^{1/d},q^{-{1/d}}\right] \cr } $$
(the latter arising by restricting the former): these can be specialized at $ q^{1/d} = 1 $, yielding
\vskip7pt
\proclaim{Theorem 8.15} The perfect pairings \
$ \pi_q^{\varphi, \Cal P} \colon \uqphiPinfhtilde \otimes
\uqphiPgtilde \rightarrow k\left[q^{1/d},q^{-{1/d}}\right] \; $
and\break
$ \; \pi_q^{\varphi, \Cal P} \colon \fqphiPghat \otimes
\uqphiPgtilde \rightarrow k\left[q^{1/d},q^{-{1/d}}\right] $ \ specialize to perfect pairings
$$ \pi^{\tau, \Cal P} \colon \finfgtau \otimes \fhtau
\rightarrow k \, , \quad \pi^{\tau, \Cal P} \colon \fgtau
\otimes \fhtau \rightarrow k \, . \; \square $$
\endproclaim
\vskip7pt
{\bf 8.16 Final remarks.} \ {\it (a)\/} Most of our
results --- as far as most of the results we quoted
from [DL], [DP], [CV-1/2], etc. --- are
in fact available also for quantum (enveloping/function)
algebras with any weight lattice $ M $ (with $ \, Q \leq
M \leq P \, $), e.~g.~$ U_{q,\varphi}^{M,\infty} (\gerh) $, $ F_{q,\varphi}^M [G] $, etc.
\par
{\it (b)\/} The specialization $ \, \uqphiPinfhtilde @>{q \rightarrow 1}>> \finfgtau \, $
(Theorem 8.7) can be seen as a particular concrete realization of the equivalence among the category of
{\it quantum universal enveloping algebras\/} and the category of {\it quantum formal series Hopf
algebras\/} described in [Dr], \S 7.
\par
{\it (c)\/} Constructions and results about integer forms have been formulated with $ \kqqm $
as ground ring for simplicity; nevertheless, one could check that same forms can in fact be defined
(and results proved) also on the smaller ring $ \Z[q,\qm] $, exactly like it occurs for
integer forms of $ U_{q,\varphi}^M(\gerg) $ and $ F_{q,\varphi}^M[G] $ (cf.~[DL] and [CV-2]).
\vskip2truecm
\vfill\break
\centerline{ \bf Appendix: the case $ \, G = SL(2,k)
\, $ }
\vskip10pt
{\bf A.1.} \ The present appendix is devoted to the
particularly simple case of $ \, G = SL(2,k) \, $: we will
exhibit some explicit formulas by direct computation ---
following the general pattern described in the main text
--- and then relate this approach with the well known
technique of studying $ F_q[SL(2,k)] $ by generators and relations; it will then be clear that such a comparison can be done more in general for the group $ \, G = SL(n+1,k)
\, $ too.
\vskip7pt
{\bf A.2.} \ From \S 3 we recall that $ \,
D_q^P(\gerg) = \uqPbp \otimes \uqQbm \, $ (resp.~$ \, D_q^Q(\gerg) = \uqQbp \otimes $\break
$ \otimes \uqQbm \, $) is generated by $ \, E \otimes 1 $, $ L \otimes 1 $ (resp.~$ K \otimes 1
$), $ 1 \otimes K $, $ 1 \otimes F $, with defining relations (3.2) (where $ \, E = E \otimes
1 \, $, and so on). Using these and relations (3.3) we find the following "straightening laws"
for the product of monomials of a PBW basis of $ D_q^P(\gerg) $
$$ \eqalign{
\Big( \ebar^r L^\ell \otimes & K^k \fbar^s \Big) \cdot \Big(
\ebar^{r'} L^{\ell'} \otimes K^{k'} \fbar^{s'} \Big) = \cr
= \sum_{t \geq 0}^{t \leq r,s} & q^{(\ell + 2k)(r'-t) +
(\ell' + 2k')(s-t)} \cdot {\left[ r' \atop t \right]}_{\! q} \cdot {\bigg[ {s \atop t} \bigg]}_{\! q} \cdot
{{[t]}_q \! !}^2 \cdot {(q - q^{-1})}^{2t} \cdot \cr
& \cdot \ebar^{r+r'-t} L^{\ell + \ell'} \cdot {\left[
{K^{-1}}^\otimes_\otimes ; 2t-r'-s \atop t \right]}_q \cdot
K^{k+k'} \cdot \fbar^{s+s'-t} \cr } \eqno (A.1) $$
and for the product of monomials of a PBW basis of $ D_q^Q(\gerg) $
$$ \eqalign{
\Big( E^{(r)} K^h \otimes & K^k F^{(s)} \Big) \cdot \Big( E^{(r')}
K^{h'} \otimes K^{k'} F^{(s')} \Big) = \cr
= \sum_{t \geq 0}^{t \leq r,s} & q^{2((h+k)(r'-
t)+(h'+k')(s-t))} \cdot {\left[ r+r'-t \atop r \right]}_{\!
q} \cdot {\left[ s+s'-t \atop s' \right]}_{\! q} \cdot \cr
& \cdot E^{(r+r'-t)} K^{h+h'} \cdot {\left[ {K^{-
1}}^\otimes_\otimes ; 2t-r'-s \atop t \right]}_q \cdot
K^{k+k'} F^{(s+s'-t)} \; . \cr } \eqno (A.2) $$
\vskip7pt
{\bf A.3.} \ Following \S 5, we identify the
$ \kq $-algebra $ \, {U_q^M (\gerb_-)}_{op} \otimes {\uqPbp}_{op} \, $ with a
subalgebra of $ {D_q^{M'}(\gerg)}^* $, and denote by $ \, F_q^M[D] := {U_q^M (\gerb_-)}_{op}
\otimeshat {U_q^P (\gerb_+)}_{op} \, $ its completion; we shall now describe by explicit
formulas the formal Hopf algebra structure of $ A^{0,M}_\infty $ --- for $ M = P, Q $, by
restriction of that of $ F_q^M[D] $. We already know that
$$ \eqalign{
\varepsilon (F \otimes 1) = 0 \, , & \; \quad \varepsilon (L^{\pm 1} \otimes 1) = 1 \, , \; \quad \varepsilon (1 \otimes L^{\pm 1}) = 1 \, , \; \quad \varepsilon (1 \otimes E) = 0 \cr
\text{resp. \ \ \ } \varepsilon (F \otimes 1) = 0 \, , & \; \quad \varepsilon (K^{\pm 1} \otimes 1) = 1 \, , \; \quad \varepsilon (1 \otimes L^{\pm 1}) = 1 \, , \; \quad \varepsilon (1 \otimes E) = 0 \cr } $$
thus the counity of $ F_q^P[D] $, resp.~$ F_q^Q[D] $, hence of $ A^{0,P}_\infty $,
resp.~$ A^{0, Q}_\infty $, is completely determined. As for comultiplication
$ \Delta_{{D_q^{M'}\!(\gerg)}^*} \! := {\left( m_{D_q^{M'}\!(\gerg)} \right)}^* \colon
{D_q^{M'}\!(\gerg)}^* \rightarrow {D_q^{M'}\!(\gerg)}^* \otimeshat {D_q^{M'}\!(\gerg)}^* $, we
perform direct computation. We begin with $ \, \fbar \otimes 1 \in {D_q^Q(\gerg)}^* \, $: we have
$$ \displaylines{
\left\langle \Delta \left( \fbar \otimes 1 \right), \left( E^{(r)} K^h \otimes K^k F^{(s)} \right)
\otimes \left( E^{(r')} K^{h'} \otimes K^{k'} F^{(s')} \right) \right\rangle = \hfill {\ } \cr
{\ } \hfill = \langle \fbar \otimes 1, \left( E^{(r)} K^h \otimes K^k F^{(s)} \right) \cdot
\left( E^{(r')} K^{h'} \otimes K^{k'} F^{(s')} \right) = \cr } $$
$$ \displaylines{
= \sum_{t \geq 0}^{t \leq r',s} q^{2((h+k)(r'-
t)+(h'+k')(s-t))} \cdot {\left[ r+r'-t \atop r \right]}_{\! q} \cdot {\left[ s+s'-t \atop s' \right]}_{\! q}
\cdot \hfill {\ } \cr
{\ } \hfill \cdot \left\langle \fbar \otimes 1, E^{(r+r'-t)} K^{h+h'} \cdot {\left[
{K^{-1}}^\otimes_\otimes ; 2t-r'-s \atop t \right]}_q \cdot K^{k+k'} F^{(s+s'-t)} \right\rangle =
\cr } $$
$$ = \delta_{s',0} \cdot \delta_{r+r'-s,1} \cdot
q^{2(h+k)(r'-s)} \cdot {\left[ 1 \atop r \right]}_{\! q} \cdot \left\langle \fbar \otimes 1, E^{(1)}
K^{h+h'} \cdot {\left[ {K^{-1}}^\otimes_\otimes ; s-r' \atop s \right]}_q \cdot K^{k+k'}
\right\rangle = $$
$$ {\ } \hfill = \delta_{s',0} \cdot {(-1)}^{s+1}
\cdot \left( \delta_{r,0} \cdot \delta_{r',s+1} \cdot q^{2(h+k)} + \delta_{r,1} \cdot \delta_{r',s} \cdot
\delta_{s,0} \right) \; ; $$
now the element $ \, \left( \fbar \otimes 1 \right) \otimes (1 \otimes 1) + \sum_{n=0}^\infty
q^{-n} \cdot \left( K^{-1} \otimes K \ebar^n \right) \otimes \left( \fbar^{n+1} \otimes 1 \right)
\, \left( \in {F_q^P[D]}^{\otimes 2} \right) \, $ takes on PBW monomials $ \, \left( E^{(r)}
K^h \otimes K^k F^{(s)} \right) \otimes \left( E^{(r')} K^{h'} \otimes K^{k'} F^{(s')} \right)
\, $ the same values as $ \Delta \left( \fbar \otimes 1 \right) $, hence
$$ \Delta \left( \fbar \otimes 1 \right) = \left( \fbar \otimes 1 \right) \otimes (1 \otimes 1) +
\sum_{n=0}^\infty q^{-n} \cdot \left( K^{-1} \otimes K \ebar^n \right) \otimes \left( \fbar^{n+1}
\otimes 1 \right) \; . $$
With exactly the same technique we get formulas
$$ \eqalign{
\Delta \left( L^{-1} \otimes L \right) & = \sum_{n=0}^\infty \left( L^{-1} \otimes L \ebar^n
\right) \otimes \left( \fbar^n L^{-1} \otimes L \right) \cr
\Delta \left( L \otimes L^{-1} \right) & = \left( L \otimes L^{-1} \right) \otimes \left( L
\otimes L^{-1} \right) - \left( L \otimes L^{-1} \ebar \right) \otimes \left( \fbar L \otimes
L^{-1} \right) \cr
\Delta \left( 1 \otimes \ebar \right) & = (1 \otimes 1) \otimes \left( 1 \otimes \ebar \right)
+ \sum_{n=0}^\infty q^{+n} \cdot \left( 1 \otimes \ebar^{n+1} \right) \otimes \left( \fbar^n K^{-1}
\otimes K \right) \; ; \cr } $$
by the way, this also proves $ \, \Delta \left( A^{0,P}_\infty \right) \leq A^{0,P}_\infty \otimeshat A^{0,P}_\infty \, $
and\footnote{ In fact one can also compute (in the same way) the values $ \Delta
\left( L^{\pm 1} \otimes 1 \right) $, $ \Delta \left( 1 \otimes L^{\pm 1} \right) $, and so directly
verify that $ \, \Delta \left( F_q^P[D] \right) \leq F_q^P[D] \otimeshat F_q^P[D] \, $. } $ \, \Delta
\left( {\widetilde A}^{0,P}_\infty \right) \leq {\widetilde A}^{0, P}_\infty \otimeshat {\widetilde A}^{0,P}_\infty
\, $.
\par
The same procedure for $ A^{0,Q}_\infty $ --- using formula (A.2) --- gives
$$ \eqalign{
\Delta \left( F^{(1)} \otimes 1 \right) & = \left( F^{(1)} \otimes 1 \right) \otimes (1 \otimes 1) +
\hfill {\ } \cr
{\ } \hfill + & \sum_{n=0}^\infty q^{-n} \cdot {(q -
q^{-1})}^{2n} \cdot {[n]}_q ! \cdot {[n+1]}_q ! \cdot \left( K^{-1} \otimes K \ebar^n \right) \otimes
\left( \fbar^{n+1} \otimes 1 \right) \cr
\Delta \left( K^{-1} \otimes K \right) & = \sum_{n=0}^\infty {(q - q^{-1})}^{2n} \cdot {{[n]}_q !}^2
\cdot {[n+1]}_q \cdot \left( K^{-1} \otimes K E^{(n)} \right) \otimes \left( F^{(n)} K^{-1} \otimes K
\right) \cr
\Delta \left( K \otimes K^{-1} \right) & = \left( K \otimes K^{-1} \right) \otimes \left( K \otimes K^{-1} \right) - \hfill {\ } \cr
- & {(q - q^{-1})}^2 \cdot {[2]}_q \cdot \, \left( K \otimes K^{-1} E^{(1)} \right) \otimes \left( F^{(1)} K \otimes K^{-1} \right) + \cr
{\ } \ \ & + {(q - q^{-1})}^4 \cdot {{[2]}_q}^2 \!
\cdot \left( K \otimes K^{-1} E^{(2)} \right) \otimes \left( F^{(2)} K \otimes K^{-1} \right) \cr
\Delta \left( 1 \otimes E^{(1)} \right) & = (1 \otimes 1) \otimes \left( 1 \otimes E^{(1)} \right) +
\hfill {\ } \cr
{\ } \hfill + & \sum_{n=0}^\infty q^{+n} \cdot {(q -
q^{-1})}^{2n} \cdot {[n+1]}_q ! \cdot {[n]}_q ! \cdot \left( 1 \otimes E^{(n+1)} \right) \otimes
\left( F^{(n)} K^{-1} \otimes K \right) \, ; \cr } $$
notice that, as explained in the main text (\S 5.15), we deal with series which are convergent in the
$ (q-1) $-adic topology; moreover, formulas above explicitly show that
$ \, \Delta \left( A^{0,Q}_\infty \right) \leq $\break
$ \leq A^{0,Q}_\infty \otimeshat A^{0,Q}_\infty \, $
and\footnote{ As above, one can compute
$ \Delta \left( K^{\pm 1} \otimes 1 \right) $,
$ \Delta \left( 1 \otimes L^{\pm 1} \right) $, and get
$ \, \Delta \left( F_q^Q[D] \right) \leq F_q^Q[D]
\otimeshat F_q^Q[D] \, $. }
$ \, \Delta \left( {\widehat A}^{0,Q}_\infty \right) \leq {\widehat A}^{0,Q}_\infty \otimeshat
{\widehat A}^{0,Q}_\infty \, $.
\par
Similarly we proceed with the antipode; recall that
$ S_{{D_q^{M'}\!(\gerg)}^*} $ is defined as
$ \, S_{{D_q^{M'}\!(\gerg)}^*} := {\left( S_{D_q^{M'}\!(\gerg)} \right)}^* \colon
{D_q^{M'}\!(\gerg)}^* \rightarrow {D_q^{M'}\!(\gerg)}^* \, $; then direct computation gives
$$ \displaylines{
\left\langle S \left( \fbar \otimes 1 \right), \left( E^{(r)} K^h \otimes K^k F^{(s)} \right)
\right\rangle = \left\langle \fbar \otimes 1, S \left( E^{(r)} K^h \otimes K^k F^{(s)} \right)
\right\rangle = \cr
= {(-1)}^{r+s} \cdot q^{-s(s-1)-r(r+1)-2rh-2rk+2rs} \cdot
\hfill {\ } \cr
{\ } \hfill \cdot \sum_{t \geq 0}^{t \leq r,s} q^{2((r-
h)+(s-k))(s-t)} \cdot \left\langle \fbar \otimes 1, E^{(r-
t)} \cdot K^{r-h} \left[ {K^{-1}}^\otimes_\otimes ; 2t - r
- s \atop t \right] K^{s-k} \cdot F^{(s-t)} \right\rangle =
\cr
= \delta_{r,s+1} \cdot q^{-2(1+h(s+1)+k(s+1))} \cdot
{(-1)}^s \cr } $$
that is we have
$$ \left\langle S \left( \fbar \otimes 1 \right), E^{(r)} K^h \otimes K^k F^{(s)} \right\rangle =
\delta_{r,s+1} \cdot q^{-2(1+h(s+1)+k(s+1))} \cdot {(-1)}^s \; ; $$
now, the element $ \, - q^{-2} \cdot \sum_{n=0}^\infty \fbar^{n+1} L^{2(n+1)} \otimes
L^{-2(n+1)} \ebar^n \, \left( \in F_q^P[D] \right) \, $ when evaluated on PBW elements $ \,
E^{(r)} K^h \otimes K^k F^{(s)} \, $ of $ D_q^Q(\gerg) $ takes the same values as $ S \left(
\fbar \otimes 1 \right) $, therefore
$$ S(\fbar \otimes 1) = - q^{-2} \cdot \sum_{n=0}^\infty
\fbar^{n+1} L^{2(n+1)} \otimes L^{-2(n+1)} \ebar^n \; . $$
With exactly the same technique we get formulas
$$ \eqalign{
S \left( L^{-1} \otimes L \right) & = \sum_{n=0}^\infty \fbar^n L^{2n+1} \otimes L^{-2n-1}
\ebar^n \cr
S \left( L \otimes L^{-1} \right) & = L^{-1} \otimes L - \fbar L \otimes L^{-1} \ebar \cr
S \left( 1 \otimes \ebar \right) & = - q^{+2} \cdot \sum_{n=0}^\infty \fbar^n L^{2(n+1)} \otimes
L^{-2(n+1)} \ebar^{n+1} \; ; \cr } $$
by the way, this explicitly shows that $ \, S \left( A^{0,P}_\infty \right) \leq
A^{0,P}_\infty \, $
and\footnote{ Similarly one can compute $ S \left( L^{\pm 1} \otimes 1 \right) $,
$ S \left( 1 \otimes L^{\pm 1} \right) $, and so verify that $ \, S \left( F_q^P[D]
\right) \leq F_q^P[D] \, $. }
$ \, S \left( {\widetilde A}^{0,P}_\infty \right) \leq {\widetilde A}^{0, P}_\infty \, $. In
particular we directly proved that $ A^{0,P}_\infty $ is a Hopf subalgebra of $ F_q^P [D] $,
and\footnote{ Similarly we could directly prove that $ F_q^P[D] $ is a Hopf subalgebra
of $ {D_q^Q(\gerg)}^* $. }
$ {\widetilde A}^{0, P}_\infty $ is a $ \kqqm $-integer form of $ A^{0,P}_\infty $.
\par
The same procedure for $ A^{0,Q}_\infty $ --- formula (A.2) --- gives
$$ \eqalign{
S \left( F^{(1)} \otimes 1 \right) & = - q^{-2} \cdot \sum_{n=0}^\infty {(q-\qm)}^{2n} \cdot
{[n+1]}_q ! \cdot {[n]}_q ! \cdot F^{(n+1)} K^{n+1} \otimes K^{-(n+1)} E^{(n)} \cr
S \left( K^{-1} \otimes K \right) & = \sum_{n=0}^\infty {(q-\qm)}^{2n} \cdot {{[n]}_q !}^2 \cdot
{[n+1]}_q \cdot F^{(n)} K^{2n+1} \otimes K^{-2n-1} E^{(n)} \cr
S \left( K \otimes K^{-1} \right) & = K^{-1} \!\otimes\! K - {[2]}_q \!\cdot\!
{(q-\qm)}^2 \!\cdot\! F^{(1)} \!\otimes\! E^{(1)} + {(q-\qm)}^2 \!\cdot\! F^{(1)} K
\!\otimes\! K^{-1} E^{(1)} \cr
S \left( 1 \otimes E^{(1)} \right) & = - q^{+2} \cdot \sum_{n=0}^\infty {(q-\qm)}^{2n} \cdot
{[n]}_q ! \cdot {[n+1]}_q ! \cdot F^{(n)} K^{n+1} \otimes K^{-(n+1)} E^{(n+1)} \; ; \cr } $$
here again we remark that the involved series are convergent
in the $ (q-1) $-adic topology; moreover, we explicitly showed that
$ \, S \left( A^{0,Q}_\infty \right) \leq A^{0,Q}_\infty \, $
and\footnote{ Similarly we can also compute $ S \left( K^{\pm 1} \otimes 1 \right) $,
$ S \left( 1 \otimes L^{\pm 1} \right) $, and get $ \, S \left( F_q^Q[D] \right) \leq F_q^Q[D] \, $. }
$ \, S \left( {\widehat A}^{0,Q}_\infty \right) \leq {\widehat A}^{0,Q}_\infty \, $, thus
$ A^{0,Q}_\infty $ is a Hopf subalgebra of $ F_q^Q[D] $
and\footnote{ Similarly we could directly prove that $ F_q^Q[D] $ is a Hopf subalgebra of $ {D_q^P(\gerg)}^* $. }
$ {\widehat A}^{0,Q}_\infty $ is a $ \kqqm $-integer form of $ A^{0,Q}_\infty $.
\par
Finally we record that --- thanks to Theorem 6.2 ---
the substitutions
$$ F \otimes 1 \mapsto F \, , \; \qquad L^{\mp 1} \otimes L^{\pm 1} \mapsto L^{\pm 1} \, , \; \qquad K^{\mp 1} \otimes K^{\pm 1} \mapsto K^{\pm 1} \, , \; \qquad 1 \otimes E \mapsto E $$
turn the above formulas defining the "Hopf operations"
$ \Delta $, $ \varepsilon $, and $ S $ of
$ A^{0,M}_\infty $ into analogous formulas defining $ \Delta $, $ \varepsilon $, and $ S $ for
$ U_q^{M,\infty} (\gerh) $, for $ \, M= P, Q \, $.
\vskip7pt
{\bf A.4.} \ Let $ \, \gerg := {\hbox{{\cs sl}}
(2)} \, $, and consider $ \, \uqQg = U_q^Q({\hbox{{\cs sl}} (2)}) \, $; consider the $ 2 $-dimensional vector
space $ V $ over $ \kq $ with (ordered) basis $ \, \{
v_+ , v_-\} \, $: the standard representation \ $ \uqQg
\rightarrow End_{\kq} (V) $ \ is defined by\hfill\break
\centerline{ $ E \mapsto \left( 0 \ \ 1 \atop 0 \ \ 0
\right) \quad \qquad K^{\pm 1} \mapsto \left( q^{\pm 1}
\ \! 0 \, \, \atop \, 0 \ \, \, q^{\mp 1} \! \right) \quad
\qquad F \mapsto \left( 0 \ \ 0 \atop 1 \ \ 0
\right) $ }
where matrices are relative to the ordered basis $ \, \{
v_+ , v_-\} \, $; letting $ \, \{ \phi_+, \phi_-\} \, $
be the basis of $ V^* $ dual of $ \, \{ v_+ , v_-\}
\, $, we denote the matrix coefficients of this
representation by
$$ \eqalign{
a:= \langle \phi_+, \text{--} .v_+ \rangle \quad (x
\mapsto \langle \phi_+, x.v_+ \rangle) \, ,
& \qquad b:= \langle \phi_+, \text{--} .v_- \rangle \quad
(x \mapsto \langle \phi_+, x.v_-
\rangle) \cr
c:= \langle \phi_-, \text{--} .v_+ \rangle \quad (x
\mapsto \langle \phi_-, x.v_+ \rangle) \, ,
& \qquad d:= \langle \phi_-, \text{--} .v_- \rangle \quad
(x \mapsto \langle \phi_-, x.v_-
\rangle) \, . \; \cr } $$
\indent It is well known (cf.~for instance [APW],
Appendix, or [DL], \S 1) that the quantum function
algebra $ \, \fqPg = F_q[{SL(2,k)}] \, $ is generated by
the matrix coefficients of the standard representation,
(this is true more in general for $ \, G = SL(n+1,k) \, $) i.~e.~$ \, a $, $ b $, $ c $, $ d $. Thus $ \, \fqPg \, $ is the associative $ \kq
$-algebra with $ 1 $ with generators $ \, a $, $ b $, $ c $, $ d $ and relations
$$ \eqalign{
a b = q b a \; , \qquad c d = q d c \; , & \qquad a c
= q c a \; , \qquad b d = q d b \cr
b c = c b \; , \qquad a d - d a = (q & -\qm) \, b c
\; , \qquad a d - q b c = 1 \; ; \cr } $$
with Hopf algebra structure defined by formulas
$$ \eqalign{
\Delta (a) = a \otimes a + b \otimes c \; , & \qquad
\varepsilon (a) = 1 \; , \qquad S(a) = d \cr
\Delta (b) = a \otimes b + b \otimes d \; , &
\qquad \varepsilon (b) = 0 \; , \qquad S(b) = -q b \cr
\Delta (c) = c \otimes a + d \otimes c \; , & \qquad
\varepsilon (c) = 0 \; , \qquad S(c) = -\qm c \cr
\Delta (d) = c \otimes b + d \otimes d \; , & \qquad
\varepsilon (d) = 1 \; , \qquad S(d) = a \cr } \eqno (A.3) $$
and $ \fqPghat $ is nothing but the $ \kqqm $-subalgebra generated by $ a $, $ b $, $ c $, $ d $.
\par
Similarly $ \, F_q^P[B_+] \, $, resp.~$ \, F_q^P[B_-]
\, $, has the following presentation: it is the associative
$ \kq $-algebra with $ 1 $ with generators $ a_+ $,
$ b_+ $, $ d_+ $, resp. $ a_- $, $ c_- $, $ d_- $, and
relations
$$ \eqalign{
a_+ b_+ = q \, b_+ a_+ \; , & \qquad b_+ d_+ = q \,
d_+ b_+ \; , \qquad a_+ d_+ = 1 = d_+ a_+ \; , \cr
\text{ resp. } \qquad \qquad a_- c_- = q \, c_- a_- \; , & \qquad c_- d_- = q \, d_- c_- \; , \qquad a_- d_- = 1 = d_- a_- \; , \cr } $$
with the Hopf algebra structure given by
$$ \eqalign{
\Delta (a_+) = a_+ \otimes a_+ \; , & \qquad
\varepsilon (a_+) = 1 \; , \qquad S(a_+) = d_+ \cr
\Delta (b_+) = a_+ \otimes b_+ + b_+ \otimes d_+ \; , &
\qquad \varepsilon (b_+) = 0 \; , \qquad S(b_+) = -q b_+ \cr
\Delta (d_+) = d_+ \otimes d_+ \; , & \qquad
\varepsilon (d_+) = 1 \; , \qquad S(d_+) = a_+
\cr } $$
resp.
$$ \eqalign{
\Delta (a_-) = a_- \otimes a_- \; , & \qquad
\varepsilon (a_-) = 1 \; , \qquad S(a_-) = d_- \cr
\Delta (c_-) = c_- \otimes a_- + d_- \otimes c_- \; , &
\qquad \varepsilon (c_-) = 0 \; , \qquad S(c_-) = -\qm c_- \cr
\Delta (d_-) = d_- \otimes d_- \; , & \qquad
\varepsilon (d_-) = 1 \; , \qquad S(d_-) = a_-
\; ; \cr } $$
and the integer form $ \fqPbphat $, resp.~$ \fqPbmhat $,
is nothing but the $ \kqqm $-subalgebra generated by
$ a_+ $, $ b_+ $, $ d_+ $, resp. $ a_- $, $ c_- $,
$ d_- $.
\par
The Hopf algebra epimorphisms
$$ \rho_+ \colon \fqPg \twoheadrightarrow F_q^P[B_+] \; ,
\qquad \rho_- \colon \fqPg \twoheadrightarrow \fqPbm $$
given by restriction (cf.~the proof of Theorem 5.7) are then
described by
$$ \rho_+ \colon \, a \mapsto a_+ \, , \; b \mapsto b_+
\, , \; c \mapsto 0 \, , \; d \mapsto d_+ \, , \qquad \rho_- \, \colon a \mapsto a_- \, , \; b \mapsto 0 \, , \; c \mapsto c_- \, , \; d \mapsto d_- \; . $$
\vskip7pt
{\bf A.5.} \ We shall now explicitly realize the
embedding of (formal) Hopf algebras (cf.~\S 5)
$$ \mu^P \colon \fqPg \hookrightarrow \aPinf \, . $$
\indent From the previous subsection it is easy to check that the Hopf algebra
isomorphisms\footnote{ {\it Remark:\/} beware of the difference among these isomorphisms and the ones in [DL], \S 1, which are only {\it algebra\/} isomorphisms; for this reason the embedding $ \, \fqPg \hookrightarrow {\uqPbm}_{op} \otimes {\uqPbp}_{op} \, $ (see below) defined in [DL] is only an {\it algebra\/} monomorphism, while ours is a formal Hopf algebra monomorphism. }
$$ \vartheta_+ \colon \fqPbp \cong {\uqPbm}_{op} \; \text{ (induced by $ \pi_- $) }\!, \; \vartheta_- \colon \fqPbm \cong {\uqPbp}_{op} \; \text{ (induced by $ \overline{\pi_-} $) } $$
(cf.~(4.1)) are described by formulas
$$ \vartheta_+ \colon \, a_+ \mapsto L^{-1} \, , \; b_+
\mapsto - \fbar L \, , \; d_+ \mapsto L \, , \qquad \vartheta_- \, \colon a_- \mapsto L \, , \; c_- \mapsto L^{-1} \ebar \, , \; d_- \mapsto L^{-1} \; . $$
\indent As we saw in the proof of Theorem 5.7,
$ \mu^P $ is nothing but the composition
$$ \fqPg @>{\Delta}>> \fqPg \otimes \fqPg @>{\rho_+ \otimes \rho_-}>> F_q^P[B_+] \otimes \fqPbm @>{\vartheta_+ \otimeshat \vartheta_-}>> {U_q^P(\gerb_-)}_{op} \otimeshat {U_q^P(\gerb_+)}_{op} $$
(whose image lies in $ \aPinf \, $) hence it is described
by
$$ \mu^P \colon \quad a \mapsto L^{-1} \otimes L - \fbar
L \otimes L^{-1} \ebar \, , \; \; b \mapsto - \fbar L \otimes L^{-1} \, , \; \; c \mapsto L \otimes L^{-1} \ebar \, , \; \; d \mapsto L \otimes L^{-1} \, ; $$
then by direct comparison between formulas (A.3) and the
formulas we gave for $ \, F_q^P[D] := $\break
$ := {\uqPbm}_{op} \otimes {\uqPbp}_{op} \, $
one immediately checks that $ \mu^P $ preserves the (formal) Hopf structure, i.~e.~it is a
monomorphism {\it of (formal) Hopf algebras\/}.
\vskip2,1truecm
\Refs
\endRefs
\vskip7pt
[APW] \ H.~H.~Andersen, P.~Polo, Wen Kexin, {\it
Representations of quantum algebras\/}, Invent.~Math.~{\bf
104} (1991), 1--59.
\vskip4pt
[CV-1] \ M.~Costantini, M.~Varagnolo, {\it Quantum double
and multiparameter quantum group\/}, Comm.~in Alg.~{\bf 22
(15)} (1994), 6305--6321.
\vskip4pt
[CV-2] \ M.~Costantini, M.~Varagnolo, {\it Multiparameter
Quantum Function Algebra at Roots of 1\/}, Preprint hep-th 9410106.
\vskip4pt
[DD] \ I.~Damiani, C.~De Concini, {\it Quantum groups and
Poisson groups\/}, in W.~Baldoni, M.~Picardello (eds.),
{\it Representations of Lie groups and quantum groups\/},
Longman Scientific $ \and $ Technical.
\vskip4pt
[Di] \ J.~Dieudonn\'e, {\it Introduction to the theory of
formal groups\/}, Pure and Applied Mathematics {\bf 20},
1973.
\vskip4pt
[DK] \ C.~De Concini, V.~G.~Kac, {\it Representations of
Quantum Groups at Roots of 1\/}, Colloque Dixmier 1989,
Progr.~in Math.~{\bf 92} (1990), 471 -- 506.
\vskip4pt
[DKP] \ C.~De Concini, V.~G.~Kac, C.~Procesi, {\it
Quantum coadjoint action\/}, Jour.~Am. Math.~Soc.~{\bf 5}
(1992), 151 -- 189.
\vskip4pt
[DL] \ C.~De Concini, V.~Lyubashenko, {\it Quantum
function algebra at roots of 1\/}, Adv. Math.~{\bf 108}
(1994), 205 -- 262.
\vskip4pt
[DP] \ C.~De Concini, C.~Procesi, {\it Quantum groups\/},
in {\it D-modules, Representation Theory, and Quantum
Groups\/}, L.~Boutet de Monvel, C.~De Concini, C.~Procesi,
P.~Schapira, M.~Vergne eds., Lectures Notes in Mathematics
vol.~1565, Springer $ \and $ Verlag, Berlin--Heidelberg--
New York, 1993.
\vskip4pt
[Dr] \ V.~G.~Drinfeld, {\it Quantum groups\/}, Proc.~ICM
Berkeley 1 (1986), 789--820.
\vskip4pt
[EK-1] \ P.~Etingof, D.~Kazhdan, {\it Quantization of Lie
bialgebras, I\/}, Preprint\break
q-alg\@eprints.math.duke.edu (July 24$^{\text{th}}$, 1995).
\vskip4pt
[EK-2] \ P.~Etingof, D.~Kazhdan, {\it Quantization of Poisson algebraic groups and Poisson homogeneous spaces\/}, Preprint q-alg\@eprints.math.duke.edu (October 20$^{\text{th}}$, 1995).
\vskip4pt
[Ji] \ M.~Jimbo, {\it A $q$-difference analogue of
$ \ug $ and the Yang-Baxter equation\/}, Lett. Math.
Phys.~{\bf 10} (1985), 63--69.
\vskip4pt
[Hu] \ J.~E.~Humphreys, {\it Introduction to Lie Algebras
and Representation Theory\/}, Grad. Texts in Math.~{\bf 9}
(1972), Springer $ \and $ Verlag, Berlin--Heidelberg--New
York.
\vskip4pt
[Lu-1] \ G.~Lusztig, {\it Quantum deformations of certain
simple modules over enveloping algebras\/}, Adv. Math. {\bf
70} (1988), 237--249.
\vskip4pt
[Lu-2] \ G.~Lusztig, {\it Quantum groups at roots of 1\/},
Geom.~Dedicata~{\bf 35} (1990), 89--113.
\vskip4pt
[Pa] \ P.~Papi, {\it A characterization of a good ordering
in a root system\/}, Proc.~Am.~Math. Soc.~{\bf 120}, no.~3
(1994), 661--665.
\vskip4pt
[Re] \ N.~Reshetikin, {\it Multiparameter Quantum Groups
and Twisted Quasitriangular Hopf Algebras\/},
Lett.~Math.~Phys.~{\bf 20} (1990), 331--335.
\vskip4pt
[Sw] \ M.~Sweedler, {\it Hopf algebras\/}, Mathematics
Lecture Notes Series, Benjamin, New York, 1969.
\vskip4pt
[Tk] \ M.~Takeuchi, {\it Hopf algebra techniques applied to the quantum group \hbox{ $ U_q(\hbox{{\cs sl}} (2)) $}\/}, in {\it Deformation Theory and Quantum Groups with Applications to Mathematical Physics\/}, M. Gerstenhaber and J.~Stasheff eds., Cont.~Math.~{\bf 134} (1992), 309--323.
\vskip4pt
[Tn] \ T.~Tanisaki, {\it Killing forms, Harish-Chandra
Isomorphisms, and Universal R-Matrices for Quantum
Algebras\/}, Internat.~J.~Modern Phys.~A {\bf 7}, Suppl.~1 B, (1992), 941--961.
\vskip0,75truecm
\enddocument