1: \documentclass{amsart}

2: 

3: \usepackage{amsmath}

4: \usepackage{amscd}

5: \usepackage{amsfonts}

6: \usepackage{amssymb}

7: \usepackage{latexsym}

8: \usepackage{pifont}

9: \usepackage{eepic}

10: \usepackage{array}

11: %\usepackage{draftcopy}      

12: %\draftcopySetGrey{0.95}     

13: 

14: \newcommand{\ncm}{\newcommand}

15: %\ncm{\rncm}{\renewcommand}

16: 

17: \ncm{\beq}{\begin{equation}}

18: \ncm{\eeq}{\end{equation}}

19: 

20: \newtheorem{thm}{Theorem}[section]

21: \newtheorem{pro}[thm]{Proposition}

22: \newtheorem{lem}[thm]{Lemma}

23: \newtheorem{cor}[thm]{Corollary}

24: \newtheorem{thm&def}[thm]{Theorem \& Definition}

25: 

26: \theoremstyle{definition}

27: \newtheorem{defi}[thm]{Definition}

28: \newtheorem{exa}[thm]{Example}

29: \newtheorem{xca}[thm]{Exercise}

30: 

31: \theoremstyle{remark}

32: \newtheorem{rmk}[thm]{Remark}

33: 

34: \numberwithin{equation}{section}

35: 

36: \def\Ob{\mathit{Ob}\,}

37: \def\Set{\mathsf{Set}}

38: \def\Ab{\mathsf{Ab}}

39: \def\Vec{\mathsf{Vec}}

40: \def\Ring{\mathsf{Ring}}

41: \def\Grp{\mathsf{Grp}}

42: \def\Top{\mathsf{Top}}

43: \def\Tangle{\mathsf{Tangle}}

44: \def\Alg{\mathsf{Alg}}

45: \def\Mat{\mathsf{Mat}}

46: \def\Path{\mathsf{Path}}

47: \def\Cat{\mathsf{Cat}}

48: \def\Mon{\mathsf{Mon}}

49: \def\Comon{\mathsf{Comon}}

50: \def\MonCat{\mathsf{MonCat}}

51: \def\Func{\mathsf{Func}}

52: \def\rep{\mathsf{rep}\,}

53: \def\FGP{\mathsf{FGP}}

54: \def\M{\mathsf{M}}

55: \ncm{\Cgd}{\mathsf{Cgd}}

56: \ncm{\BGD}{\mathsf{Bgd}}

57: \ncm{\BgdMap}{\mathsf{BgdMap}}

58: \ncm{\Bim}{\mathbb{BIM}}

59: \ncm{\YD}{\mathcal{YD}}

60: \ncm{\Gal}{\mathsf{Gal}}

61: \ncm{\Fib}{\mathsf{Fib}}

62: \ncm{\DHR}{\mathsf{DHR}}

63: 

64: \def\Mod{\mathtt{Mod}}

65: \def\A{\mathcal{A}}

66: \def\B{\mathcal{B}}

67: \def\C{\mathcal{C}}

68: \def\D{\mathcal{D}}

69: \def\E{\mathcal{E}}

70: \ncm{\F}{\mathcal{F}}

71: \ncm{\G}{\mathcal{G}}

72: \def\T{\mathcal{T}}

73: \ncm{\V}{\mathcal{V}}

74: \ncm{\W}{\mathcal{W}}

75: \ncm{\Z}{\mathcal{Z}}

76: 

77: \ncm{\asso}{\mathbf{a}}

78: \ncm{\luni}{\mathbf{l}}

79: \ncm{\runi}{\mathbf{r}}

80: 

81: \ncm{\End}{\operatorname{End}}

82: \ncm{\Hom}{\operatorname{Hom}}

83: %\def\Hom{\mbox{\rm Hom}\,}

84: \ncm{\BiEnd}{\operatorname{BiEnd}}

85: \def\Iso{\mbox{\rm Iso}\,}

86: \def\Ker{\mbox{\rm Ker}\,}

87: \def\id{\mbox{\rm id}\,}

88: \def\Center{\mbox{\rm Center}\,}

89: 

90: \newcommand{\ci}{\circ}

91: \newcommand{\bo}{\,\Box\,}

92: \def\o{\otimes}

93: \def\x{\times}

94: 

95: 

96: \ncm{\amalgo}[1]{\underset{\scriptscriptstyle #1}{\o}}

97: \ncm{\ractB}{\underset{\scriptscriptstyle B}{\ract}}

98: \ncm{\ractT}{\underset{\scriptscriptstyle T}{\ract}}

99: \ncm{\mash}{\Pisymbol{psy}{35}}

100: \ncm{\mashed}[1]{\underset{\scriptscriptstyle #1}{\Pisymbol{psy}{35}}}

101: 

102: \ncm{\oA}{\amalgo{A}}

103: \ncm{\oB}{\amalgo{B}}

104: \ncm{\oC}{\amalgo{C}}

105: \ncm{\oR}{\amalgo{R}}

106: \ncm{\oT}{\amalgo{T}}

107: \ncm{\oL}{\amalgo{L}}

108: \ncm{\oS}{\amalgo{S}}

109: \ncm{\oN}{\amalgo{N}}

110: \ncm{\oH}{\amalgo{H}}

111: \ncm{\oQ}{\amalgo{Q}}

112: \ncm{\ex}[1]{\underset{\scriptscriptstyle #1}{\x}}

113: \def\bra{\langle}

114: \def\ket{\rangle}

115: \ncm{\rarr}[1]{\stackrel{#1}{\longrightarrow}}

116: \ncm{\larr}[1]{\stackrel{#1}{\longleftarrow}}

117: \def\cop{\Delta}

118: \def\del{\delta}

119: \def\eps{\varepsilon}

120: \ncm{\op}{\mathrm{op}}

121: \ncm{\coop}{\mathrm{coop}}

122: \ncm{\co}{\mathrm{co}}

123: \ncm{\Fi}{\varphi}

124: \ncm{\OR}{\overrightarrow}

125: \ncm{\OL}{\overleftarrow}

126: 

127: \def\zeroT{^{(0)}}

128: \def\oneT{^{(1)}}

129: \def\twoT{^{(2)}}

130: \def\threeT{^{(3)}}

131: \def\zeroB{_{(0)}}

132: \def\oneB{_{(1)}}

133: \def\twoB{_{(2)}}

134: \def\threeB{_{(3)}}

135: \def\fourB{_{(4)}}

136: \def\oneR{^{[1]}}

137: \def\twoR{^{[2]}}

138: \def\threeR{^{[3]}}

139: \def\oneL{_{[1]}}

140: \def\twoL{_{[2]}}

141: \def\threeL{_{[3]}}

142: \def\zerozeroT{^{(0)(0)}}

143: \def\zerooneT{^{(0)(1)}}

144: \def\onezeroT{^{(1)(0)}}

145: \def\oneoneT{^{(1)(1)}}

146: \def\zerotwoT{^{(0)(2)}}

147: \def\twozeroT{^{(2)(0)}}

148: \def\onetwoT{^{(1)(2)}}

149: \def\twooneT{^{(2)(1)}}

150: \def\twotwoT{^{(2)(2)}}

151: \def\zerozeroR{^{[0][0]}}

152: \def\zerooneR{^{[0][1]}}

153: \def\onezeroR{^{[1][0]}}

154: \def\oneoneR{^{[1][1]}}

155: \def\zerotwoR{^{[0][2]}}

156: \def\twozeroR{^{[2][0]}}

157: \def\onetwoR{^{[1][2]}}

158: \def\twooneR{^{[2][1]}}

159: \def\twotwoR{^{[2][2]}}

160: \ncm{\coa}[1]{^{\langle #1\rangle}}

161: \ncm{\coabar}[1]{^{\langle \overline{#1}\rangle}}

162: 

163: \def\tr{\mbox{\rm tr}}

164: \def\PL{\pi_{\scriptscriptstyle L}}

165: \def\PR{\pi_{\scriptscriptstyle R}}

166: \ncm{\I}{\mathcal{I}}

167: \def\la{\!\rightharpoonup\!}

168: \def\ra{\!\leftharpoonup\!}

169: \def\du1{\hat 1}

170: \def\iso{\stackrel{\sim}{\rightarrow}}

171: \def\iiso{\rarr{\sim}}

172: \def\ract{\triangleleft}

173: \def\lact{\triangleright}

174: \def\rcross{\rtimes}

175: \ncm{\under}{\mbox{\rm\_}\,}

176: %\ncm{\under}{\text{\underscore}}

177: \ncm{\Cnt}{\mathsf{C}}

178: 

179: \ncm{\ZZ}{\mathbb{Z}}

180: 

181: \ncm{\Oo}{\mathcal{O}}

182: \ncm{\Ha}{\mathcal{H}}

183: \ncm{\Ve}{\mathcal{V}}

184: 

185: \ncm{\sqrM}{$\sqrt{\text{Morita}}$}

186: \ncm{\ld}{\,.\,}

187: \ncm{\ud}{\,^.\,}

188: \ncm{\Mor}{\underset{k}{\sim}}

189: \ncm{\into}{\hookrightarrow}

190: %\ncm{\Inv}{\operatorname{Inv}}

191: \ncm{\coinv}[1]{{\co\text{-}#1}}

192: \ncm{\lZ}{\overrightarrow{\mathcal{Z}}}

193: \ncm{\rZ}{\overleftarrow{\mathcal{Z}}}

194: \ncm{\fgp}{\mathrm{fgp}}

195: \ncm{\adj}{\dashv}

196: \ncm{\Mref}{{\scriptscriptstyle M\text{-}\mathrm{ref}}}

197: 

198: 

199: 

200: \begin{document}

201: 

202: \title{Finitary Galois extensions over noncommutative bases}

203: \author[I. B\'alint, K. Szlach\'anyi]{Imre B\'alint \and Korn\'el Szlach\'anyi}

204: \address{Research Institute for Particle and Nuclear Physics, Budapest}

205: \email{balint@rmki.kfki.hu, szlach@rmki.kfki.hu}

206: \thanks{Supported by the Hungarian Scientific Research Fund, OTKA T-034 512

207: and T-043 159}

208: 

209: \begin{abstract}

210: We study Galois extensions $M^{(\co\text{-})H}\subset M$ for 

211: $H$-(co)module algebras $M$ if $H$ is a Frobenius Hopf algebroid.

212: The relation between the action and coaction pictures is analogous to that found

213: in Hopf-Galois theory for finite dimensional Hopf algebras over fields. 

214: So we obtain generalizations of various classical theorems of 

215: Kreimer-Take\-uchi, Doi-Takeuchi and Cohen-Fischman-Montgomery. 

216: We find that the Galois extensions $N\subset M$ over some Frobenius Hopf 

217: algebroid are precisely the balanced depth 2 Frobenius extensions. 

218: We prove that the Yetter-Drinfeld categories over $H$ are always braided and 

219: their braided commutative algebras play the role of noncommutative scalar 

220: extensions by a slightly generalized Brzezi\'nski-Militaru Theorem. 

221: %In this way we have an analogue of Yetter's 

222: %Theorem in a situation when the base category $_R\M_R$ is not braidable. 

223: Contravariant "fiber functors" are used to prove an analogue of Ulbrich's 

224: Theorem and to get a monoidal embedding of the module category $\M_E$ of the 

225: endomorphism Hopf algebroid $E=\End\,_NM_N$ into $_N\M_N^\op$.

226: \end{abstract}

227: 

228: 

229: \maketitle

230: 

231: 

232: 

233: 

234: 

235: \section{Introduction}

236: 

237: The problem of extending Hopf Galois theory to quantum groupoids has been 

238: attracting some attention in recent years. That this theory should possess  

239: interesting new applications even for finite quantum groupoids is manifest 

240: already from the pioneering work of D. Nikshych and L. Vainerman \cite{N-V}. 

241: A pure algebraic Galois theory for weak Hopf algebras has been proposed 

242: by S. Caenepeel and E. de Groot \cite{Caenepeel-DeGroot}. 

243: As the next step of generalization, this paper is devoted to 

244: developing a Galois theory for Frobenius Hopf algebroids. These quantum 

245: groupoids are the analogues of finite dimensional Hopf algebras over a 

246: field or Frobenius Hopf algebras over a commutative ring. Therefore it is not 

247: surprising that we obtain generalizations of the classical theorems of 

248: Kreimer-Takeuchi \cite{Kreimer-Takeuchi}, Doi-Takeuchi \cite{Doi-Takeuchi} and 

249: Cohen-Fischman-Montgomery \cite{CFM} (see Theorems \ref{thm: def Gal} and 

250: \ref{thm: WS}). Our results partly overlap with those of the recent paper 

251: \cite{B: Gal} by G. B\"ohm who studies Galois theory for general Hopf 

252: algebroids using previous results from the theory of corings 

253: \cite{Brz: corings,Brz-Wis}. In our approach the double algebraic structure 

254: \cite{Sz: DA} of Frobenius Hopf algebroids is particularly useful e. g. in 

255: proving that Yetter-Drinfeld categories are braided (Proposition \ref{pro: YD 

256: braided}), in this way generalizing a result of \cite{Caenepeel-Wang-Yin}, or in 

257: obtaining an instrinsic characterization of Galois extensions as being the depth 

258: 2, balanced, Frobenius extensions (Theorem \ref{thm: bD2F}). 

259: 

260: 

261: 

262: \subsection{Modules and comodules over bialgebroids}

263: 

264: Let $k$ be a commutative ring. We choose  the category $\M=\M_k$ of $k$-modules 

265: as our base category. This means that all objects and morphisms we use

266: have an underlying $k$-module or $k$-module morphism, respectively.

267: In particular, algebras are always meant to be $k$-algebras and unadorned $\o$ 

268: means tensor product in $\M_k$.

269: 

270: Let $T$ be an algebra and let $T^e:=T^\op\o T$ be its enveloping algebra. 

271: A right bialgebroid over $T$ consists of

272: \begin{itemize}

273: \item an algebra $A$

274: \item a $T^e$ ring structure on $A$, i.e., an algebra morphism 

275: $t_r\o s_r:T^e\to A$

276: \item and a $T$-coring structure $\bra A_{T^e},\cop_T,\Fi_T\ket$

277: \end{itemize}

278: subject to axioms, see e.g. \cite{K-Sz}. 

279: 

280: If $A$ is a right bialgebroid over $T$ then

281: a right module over $A$ is the same thing as a right module over the $k$-algebra 

282: $A$ and a right $A$-module map is defined accordingly. The $T^e$-ring structure 

283: $T^e\to A$ endows the category $\M_A$ of right $A$-modules with a (monadic) 

284: forgetful functor $U:\M_A\to\,_T\M_T$ by identifying $\M_{T^e}$ with $_T\M_T$.

285: The coring structure of $A$ serves to make $\M_A$ a monoidal category. The 

286: monoidal product of the $A$-modules $V$ and $W$ is the $k$-module $V\oT W$ 

287: together with the right $A$-action $(v\oT w)\ract a:=(v\ract a\oneT)\oT(w\ract 

288: a\twoT)$. In this way the forgetful functor $U$ becomes strict monoidal. 

289: 

290: Left bialgebroids and their category of left modules can be defined by passing 

291: to the opposite algebra in all occurences of an algebra in the definition of a 

292: right bialgebroid and their right modules. So let $B$ be an algebra which stands 

293: for $T^\op$ and let $B^e:=B\o B^\op$. Then a left bialgebroid over $B$ consists 

294: of

295: \begin{itemize}

296: \item an algebra $A$

297: \item a $B^e$-ring structure on $A$, i.e., an algebra morphism 

298: $s_l\o t_l:B^e\to A$

299: \item and a $B$-coring structure $\bra\,_{B^e}A,\cop_B,\Fi_B\ket$.

300: \end{itemize}

301: The category of left $A$-modules has a monoidal product $V\oB W$ such that the 

302: forgetful functor $_AV\mapsto \,_{B^e}V\equiv\,_BV_B$ is strict monoidal.

303: 

304: Right comodules can be defined for both left and right bialgebroids as follows.

305: Let $A$ be a right bialgebroid over $T$. Then a right $A$-comodule consists of

306: \begin{itemize}

307: \item a right $T$-module $X$

308: \item a right $T$-module map $\delta:X\to X\oT A$

309: \end{itemize}

310: such that

311: \begin{align*}

312: (\delta\oT A)\circ\delta&=(X\oT\cop_T)\circ\delta\\

313: (X\oT\Fi_T)&=X

314: \end{align*}

315: suppressing the coherence isomorphisms of $_T\M_T$.

316: A morphism of comodules $\tau:\bra X,\gamma\ket\to \bra Y,\delta\ket$ is a right 

317: $T$-module map $\tau:X\to Y$ satisfying $(\tau\oT 

318: A)\circ\gamma=\delta\circ\tau$. The category of right $A$-comodules is denoted 

319: $\M^A$.

320: 

321: The above definition of comodules disguises the fact that $\M^A$ is 

322: monoidal with a strict monoidal forgetful functor $\M^A\to\,_T\M_T$.

323: Notice that although $M$ is not a left $T$-module, $M\oT A$ is by setting

324: $t\cdot(x\oT a)=x\oT s_r(t)a$.

325: \begin{pro} \label{pro: coactions are bim}

326: Let $\bra X,\delta\ket$ be a right comodule over the rigt bialgebroid $A$.

327: Then $X$ has a unique left $T$-module structure such that $\delta$ is a left 

328: $T$-module map. With this left module structure

329: \begin{enumerate}

330: \item $X$ is a $T$-$T$-bimodule,

331: \item $\delta$ is a $T$-$T$-bimodule map,

332: \item $\delta(X)\subset X\x_T A$,

333: \item and every arrow $\tau\in\M^A$ is a $T$-$T$-bimodule map.

334: \end{enumerate}

335: In (3) we used Takeuchi's $\x$-product which is defined by

336: \begin{equation*}

337: X\x_T A:=\{\sum_i x_i\oT a_i\in X\oT A\,|\,\sum_i t\cdot 

338: x_i\oT a_i=\sum_i x_i\oT t_r(t)a_i\,\forall t\in T\}.

339: \end{equation*}

340: \end{pro}

341: \begin{proof}

342: If $X$ is a left $T$-module and $\delta$ is a left $T$-module map then

343: \begin{align*}

344: t\cdot x&=(t\cdot x)\zeroT\cdot\Fi_T((t\cdot x)\oneT)\\

345: &=x\zeroT\cdot\Fi_T(s_r(t)x\oneT)

346: \end{align*}

347: This proves uniqueness. If we use the above formula to define $t\cdot x$ then we 

348: find that it is a left action because $s_r:T\to A$ is an algebra 

349: homomorphism. It commutes with the right $T$-action 

350: \begin{equation*}

351: t\cdot (x\cdot t')=x\zeroT\cdot\Fi_T(s_r(t)as_r(t'))=(t\cdot x)\cdot t'

352: \end{equation*}

353: so $X$ is a $T$-$T$-bimodule and the coaction is a bimodule map,

354: \begin{align*}

355: \delta(t\cdot x\cdot t')=x\zeroT\oT s_r(t)x\oneT s_r(t').

356: \end{align*}

357: Now the Takeuchi property (3) holds automatically,

358: \begin{align*}

359: t\cdot x\zeroT\oT x\oneT&=x\zeroT\cdot\Fi_T(s_r(t)x\oneT)\oT x\twoT\\

360: &=x\zeroT\oT\Fi_T(t_r(t)x\oneT)\cdot x\twoT\\

361: &=x\zeroT\oT t_r(t)x\oneT\,.

362: \end{align*}

363: If $\tau:X\to Y$ is a comodule morphism then

364: \begin{align*}

365: \tau(t\cdot x)&=\tau(x\zeroT)\cdot\Fi_T(s_r(t)x\oneT)=

366: \tau(x)\zeroT\cdot\Fi_T(s_r(t)\tau(x)\oneT)\\

367: &=t\cdot\tau(x)\,.

368: \end{align*}

369: \end{proof}

370: 

371: The tensor product of right comodules $X$ and $Y$ can now be defined as $X\oT Y$ 

372: with coaction

373: \begin{equation}\label{tensor prod of comodules over right bgd}

374: (x\oT y)\zeroT\oT (x\oT y)\oneT=(x\zeroT\oT y\zeroT)\oT x\oneT y\oneT\,.

375: \end{equation}

376: This makes the category of right $A$-comodules $\M^A$ monoidal and the 

377: forgetful functor $\M^A\to\,_T\M_T$ strict monoidal.

378: 

379: 

380: 

381: For left bialgebroids $A$ over $B$ a right comodule is an arrow $\delta_A:M\to 

382: M\oB A\in\M_B$ satisfying coassociativity and counitality. A right comodule 

383: carries a left $B$-module structure such that $\delta_A$ is a $B$-$B$-bimodule 

384: map and such that $\M^A$ is a monoidal category with strict monoidal forgetful 

385: functor to $_B\M_B$. The monoidal product of two right comodules $X$ and $Y$ 

386: has coaction

387: \begin{equation}\label{tensor prod of comodules over left bgd}

388: (x\oB y)\zeroB\oB (x\oB y)\oneB=(x\zeroB\oB y\zeroB)\oB y\oneB x\oneB\,.

389: \end{equation}

390: Note the different order compared to (\ref{tensor prod of comodules over 

391: right bgd}).

392: 

393: \subsection{Double algebras}

394: 

395: Studying module (co)algebras over bialgebroids one can obtain  

396: generalizations of certain theorems of Hopf-Galois theory and, 

397: therefore, hints toward the proper definition of bialgebroid Galois extensions

398: \cite{K-Sz,Sz: Strasbourg,Kadison}. For example, the behaviour of depth 2 

399: balanced extensions $N\subset M$ of algebras suggest that they are 

400: Galois extensions in the very noncommutative sense. Of course, in the absence of 

401: antipode, even in the finitely generated projective (fgp) case, many results of 

402: classical and Hopf Galois theory are far from reach.

403: 

404: This leads us to study Hopf algebroids instead and for finiteness we assume that 

405: it has an integral which is a Frobenius functional. These Frobenius Hopf 

406: algebroids were shown in \cite{Sz: DA} to possess a \textit{distributive double 

407: algebra} structure (DDA) by choosing a Frobenius integral. Here we summarize 

408: its basic properties. 

409: 

410: A double algebra is a $k$-module $A$ equipped with two associative unital 

411: multiplications: the vertical multiplication, denoted $a\ci a'$, has unit 

412: element $e$ and the horizontal multiplication, denoted $a\star a'$, has unit 

413: element $i$. So we have the horizontal and vertical algebras $H=\bra 

414: A,\star,i\ket$ and $V=\bra A,\ci,e\ket$, respectively. The multiplications with 

415: the wrong unit, i.e.,

416: \begin{alignat*}{2}

417: \Fi_L(a)&:=a\star e &\qquad\Fi_R(a)&:=e\star a \\

418: \Fi_B(a)&:=a\ci i &\qquad \Fi_T(a)&:=i\ci a 

419: \end{alignat*}

420: map onto subalgebras $L$ and $R$ of $V$ and $B$ and $T$ of $H$.

421: Assuming for $X=L,R,B,T$ that the algebra extensions $X\subset A$ are Frobenius 

422: with Frobenius homomorphism $\Fi_X$ we obtain the notion of Frobenius DA's.

423: In this way $A$ has Frobenius algebra structures in all the bimodule 

424: categories $_X\M_X$ for $X=L,R,B,T$ which implies four 

425: comultiplications

426: 

427: $\bra A, \Delta_B,\Fi_B\ket$ is a comonoid in $_B\M_B$, where

428:       $\Delta_B(a)\equiv a\oneB\oB a\twoB=a\star u_k\oB v_k$,

429: 

430: $\bra A, \Delta_L,\Fi_L\ket$ is a comonoid in $_L\M_L$, where

431:       $\Delta_L(a)\equiv a\oneL\oL a\twoL=a\ci x_j\oL y_j$,

432: 

433: $\bra A, \Delta_T,\Fi_T\ket$ is a comonoid in $_T\M_T$, where

434:       $\Delta_T(a)\equiv a\oneT\oT a\twoT=a\star u^k\oT v^k$,

435: 

436: $\bra A, \Delta_R,\Fi_R\ket$ is a comonoid in $_R\M_R$, where

437:       $\Delta_R(a)\equiv a\oneR\oR a\twoR=a\ci x^j\oR y^j$.

438: 

439: \noindent

440: where note the special notation for the dual bases of the base homomorphisms 

441: $\Fi_X$. It turns out \cite[Proposition 3.2]{Sz: DA} that vertical 

442: multiplication with the horizontal type of comultiplications $\cop_B$ and 

443: $\cop_T$ obey bialgebroid like relations. However, if we also postulate the 

444: distributivity rules

445: \begin{align}

446: a\ci(a'\star a'')&=(a\oneB\ci a')\star(a\twoB\ci a'') \label{eq: DB}\\

447: a\star(a'\ci a'')&=(a\oneL\star a')\ci(a\twoL\star a'') \label{eq: DL}\\

448: (a'\star a'')\ci a&=(a'\ci a\oneT)\star(a''\ci a\twoT) \label{eq: DT}\\

449: (a'\ci a'')\star a&=(a'\star a\oneR)\ci(a''\star a\twoR) \label{eq: DR}

450: \end{align}

451: in which case we say that $\bra A,\ci,e,\star,i\ket$ is a distributive double 

452: algebra (DDA), then $V$ and $H$ become Hopf algebroids \cite{B-Sz: Hgd} in 

453: duality. The underlying left bialgebroids are

454: \[

455: \bra V,B,\Fi_L|_B,\Fi_R|_B,\Delta_B,\Fi_B\ket \quad\text{and}\quad

456: \bra H,L,\Fi_B|_L,\Fi_T|_L,\Delta_L,\Fi_L\ket

457: \]

458: and the right bialgebroids are 

459: \[

460: \bra V,T,\Fi_R|_T,\Fi_L|_T,\Delta_T,\Fi_T\ket \quad\text{and}\quad

461: \bra H,R,\Fi_T|_R,\Fi_B|_R,\Delta_R,\Fi_R\ket

462: \]

463: The notation means e.g. that $V$ over $T$ has source map $s_r:t\mapsto 

464: \Fi_R(t)$, target map $t_r:t\mapsto \Fi_L(t)$ and counit $\Fi_T$. Or, $H$ over 

465: $R$ has source map $s_r:r\mapsto \Fi_T(r)$, target map $t_r:r\mapsto \Fi_B(r)$, 

466: and counit $\Fi_R$. The antipode of $V$ -- called the antipode of the double 

467: algebra -- is an antiautomorphism $S$ which is also an antiautomorphism of $H$ 

468: but the antipode of $H$ is $S^{-1}$. (There is a regrettable mistake in 

469: \cite[Theorem 7.4]{Sz: DA} where $H$ was claimed to have antipode also $S$; see 

470: \texttt{arXiv: math.QA/0402151 v2} for the corrected version.) The vertical Hopf 

471: algebroid has Frobenius integral $i$ and $H$ has $e$. 

472: 

473: 

474: 

475: 

476: \section{Modules and comodules over DDA's}

477: 

478: \subsection{Modules}

479: Let $\bra A,\ci,e,\star,i\ket$ be a double algebra. A \textit{right $A$-module} 

480: is a $k$-module $M$ together with an associative unital action $M\oR H\to M$ of 

481: the horizontal algebra $H=\bra A,\star,i\ket$ denoted $m\oR h\mapsto m\ract 

482: h$.

483: 

484: Equivalently, a right $A$-module can be formulated in the category $\M_{B\o 

485: T}$ as an object $M_{B\o T}$ and an arrow $M\amalgo{B\o T}A\to M$ satisfying 

486: associativity and unitality w.r.t the algebra $H$ in $_{B\o T}\M_{B\o T}$. The 

487: $T$ and $B$-actions are denoted by $m\ud t$ and $m\ld b$, respectively.

488: 

489: Analogously one can define \textit{left $A$-modules} as left modules over $H$ 

490: and \textit{bottom} and \textit{top $A$-modules} as "left", respectively 

491: "right", modules over the vertical algebra $V=\bra A,\ci,e\ket$.

492: 

493: 

494: \subsection{Comodules}

495: 

496: A \textit{right $A$-comodule} over a Frobenius double algebra consists of an 

497: object $M$ and two arrows $\delta_M\colon M\to M\oB A$, $\delta^M\colon M\to 

498: M\oT A$ in $\M_{B\o T}$ such that

499: \begin{itemize}

500: \item $\bra M_B,\delta_M\ket$ is a right comodule over the left bialgebroid $V$ 

501: over $B$, 

502: \item $\bra M_T,\delta^M\ket$ is a right comodule over the right 

503: bialgebroid $V$ over $T$

504: \item and the two coactions satisfy the mixed coassociativity 

505: conditions

506: \begin{align} \label{mixed coa 1}

507: {m\zeroT}\zeroB\oB {m\zeroT}\oneB\oT m\oneT

508: &=m\zeroB\oB {m\oneB}\oneT\oT {m\oneB}\twoT\\

509: {m\zeroB}\zeroT\oT {m\zeroB}\oneT\oB m\oneB

510: &=m\zeroT\oT {m\oneT}\oneB\oB {m\oneT}\twoB \label{mixed coa 2}

511: \end{align}

512: \end{itemize}

513: where we used the notation

514: \begin{align*}

515: \delta_M(m)&=m\zeroB\oB m\oneB\\

516: \delta^M(m)&=m\zeroT\oT m\oneT

517: \end{align*}

518: for $m\in M$.

519: 

520: A \textit{right $A$-comodule morphism} $\tau:X\to Y$ is a right  $B\o T$-module 

521: map  

522: %satisfying

523: %\begin{align}

524: %\delta_Y(\tau(x))&=\tau(x\zeroB)\oB x\oneB\\

525: %\delta^Y(\tau(x))&=\tau(x\zeroT)\oT x\oneT,

526: %\end{align}

527: which is a right comodule morphism for both the left bialgebroid $V_B$ and the 

528: right bialgebroid $V_T$. The category of right $A$-comodules is denoted by 

529: $\M^V$.

530: The occurence of two compatible coactions in the definition of an $A$-comodule 

531: is precisely what we need to identify $\M^V$ and $\M_H$ in case of DDA's. 

532: 

533: \begin{lem} \label{lem: act-coact}

534: Let $A$ be a DDA and let

535: $\delta_M$ and $\delta^M$ be two coactions of $V_B$, respectively $V_T$, on 

536: $M$. They then determine two right $H$-actions on $M$,

537: \begin{align}

538: m\ractB h&=m\zeroB \ld\Fi_B(m\oneB\star h) \label{eq: coa->act B}\\

539: m\ractT h&=m\zeroT \ud\Fi_T(m\oneT\star h)\,. \label{eq: coa->act T}

540: \end{align}

541: The two actions coincide if and only if the two coactions satisfy the mixed 

542: coassociativity condition (\ref{mixed coa 1}) and (\ref{mixed coa 2}).

543: \end{lem}

544: \begin{proof}

545: The inverses of (\ref{eq: coa->act B}) and (\ref{eq: coa->act T}) can be given 

546: in terms of the dual bases of $\Fi_B$ and $\Fi_T$ as

547: \begin{align}

548: m\zeroT\oT m\oneT&=m\ractT u^k\oT v^k \label{eq: act->coa B}\\

549: m\zeroB\oB m\oneB&=m\ractB u_k\oT v_k \label{eq: act->coa T}

550: \end{align}

551: Therefore if $\ractB=\ractT$ then

552: \begin{align*}

553: {m\zeroT}\zeroB\oB {m\zeroT}\oneB\oT m\oneT

554: &=(m\ractT u^k)\ractB u_l\oB v_l\oT v^k

555: =m\ractB(u^k\star u_l)\oB v_l\oT v^k\\

556: &=m\ractB u_l\oB v_l\star u^k\oT v^k

557: =m\zeroB\oB {m\oneB}\oneT\oT {m\oneB}\twoT

558: \end{align*}

559: and similarly for (\ref{mixed coa 2}). On the other hand, if mixed 

560: coassociativity holds then

561: \begin{align*}

562: m\ractT h&=(m\ractT h)\zeroB\ld\Fi_B((m\ractT h)\oneB)

563: ={m\zeroT}\zeroB\ld\Fi_B({m\zeroT}\oneB\star\Fi_T(m\oneT\star h))\\

564: &=m\zeroB\ld\Fi_B({m\oneB}\oneT\star\Fi_T({m\oneB}\twoT\star h))

565: =m\zeroB\ld\Fi_B(m\oneB\star h)\\

566: &=m\ractB h

567: \end{align*}

568: \end{proof}

569: 

570: 

571: If $M$ is a right module over the DDA $A$ then it is a right $V$-comodule $M^V$ 

572: and a right $H$-module $M_H$ at the same time. The invariants of $M_H$, 

573: \begin{align}

574: M^H&{:=}\{ n\in M | n\ract h=n\ract\Fi_T\Fi_R(h),\ h\in H\}\\

575: &=\{ n\in M | n\ract h=n\ract\Fi_B\Fi_R(h),\ h\in H\}

576: \end{align}

577: and the coinvariants of $M^V$,

578: \begin{align}

579: M^\coinv{V}&{:=}\{n\in M|n\zeroT\oT n\oneT=n\oT e\}\\

580: &=\{n\in M|n\zeroB\oB n\oneB=n\oB e\},\notag

581: \end{align}

582: yield one and the same $k$-submodule of $M$. This is an instance of the more 

583: general identification between the categories of $H$-modules, $V_B$-comodules,

584: and $V_T$-comodules. Since $\Fi_T$ and $\Fi_B$ restrict to 

585: algebra isomorphisms $R\to T$ and $R^\op\to B$, respectively \cite[Lemma 

586: 2.2]{Sz: DA}, the identifications between $H$-modules and $V$-comodules provide 

587: a monoidal category isomorphism $\M^{V_T}\cong\M_H$ and the antimonoidal 

588: category isomorphism $\M^{V_B}\cong\M_H$.  We can use these isomorphisms to 

589: introduce $\oR$ both in $\M^{V_T}$ and $\M^{V_B}$ as the monoidal product while 

590: keeping $\oT$ and $\oB$ to appear in the coactions. One  advantage of this 

591: convention is that the difference between (\ref{tensor prod of comodules over 

592: left bgd}) and (\ref{tensor prod of comodules over right bgd}) disappears, viz. 

593: (\ref{eq: comalg T}) and (\ref{eq: comalg B}). Now the $R$ becomes a monoidal 

594: unit in three senses: As a right ideal in $H$ it is the trivial right 

595: $H$-module, $r\ract h=r\star h$. But it is also a right comodule over $V_T$ via 

596: $r\zeroT \oT r\oneT=e\oT r$ and a right comodule over $V_B$ via $r\zeroB \oB 

597: r\oneB=e\oB r$.

598: 

599: 

600: \subsection{Module (co)algebras} \label{ss: ext}

601: 

602: Comodule algebras over $V$ are monoids in $\M^V$ and therefore they are the same 

603: as monoids in $\M_H$, i.e., module algebras over $H$.

604: 

605: Hence a right $H$-module algebra $M$ consists of an algebra map 

606: $\eta:R\to M$ inducing the bimodule structure $_RM_R$ and a bimodule map 

607: $\mu:M\oR M\to M$, $m\oR m'\mapsto mm'$, satisfying $(mm')\ract h=(m\ract 

608: h\oneR)(m\ract h\twoR)$.

609: In the language of the $V$-coactions (\ref{eq: act->coa B}), 

610: (\ref{eq: act->coa T}) these correspond to the right comodule algebra relations

611: \begin{alignat}{2}

612: (mm')\zeroT\oT(mm')\oneT&=

613: m\zeroT{m'}\zeroT\oT m\oneT\ci{m'}\oneT

614: &\quad

615: 1\zeroT\oT 1\oneT&=1\oT e   \label{eq: comalg T}\\

616: (mm')\zeroB\oB(mm')\oneB&=

617: m\zeroB{m'}\zeroB\oB m\oneB\ci{m'}\oneB

618: &\quad

619: 1\zeroB\oB 1\oneB&=1\oB e   \label{eq: comalg B}

620: \end{alignat}

621: respectively. 

622: Just as in the case of Hopf algebras the invariants of a module algebra form a 

623: subalgebra. More precisely we have the following

624: \begin{lem} \label{lem: R->M}

625: For any right $H$-module $M$ there is a unique $k$-module map \newline

626: $\Hom_H(R,M)\to M^H$ that makes the diagram

627: \[

628: \begin{CD}

629: \Hom_H(R,M)@>\Hom(\Fi_R,M)>>\Hom_H(H,M)\\

630: @VVV @VV{f\mapsto f(i)}V\\

631: M^H@>\subset>>M

632: \end{CD}

633: \]

634: commutative. This $k$-module map is an isomorphism. If $M_H$ is a module algebra 

635: then the diagram is in the category of $k$-algebras. In particular, $M^H\subset 

636: M$ is a subalgebra which is isomorphic to the convolution algebra $\Hom_H(R,M)$. 

637: \end{lem}

638: 

639: The smash product $H\mash M$ for a right $H$-module algebra is defined to be

640: to $k$-module $H\oR M$ equipped with multiplication

641: \begin{equation}

642: (h\mash m)(h'\mash m')=h\star {h'}\oneR\mash (m\ract {h'}\twoR)m'

643: \end{equation}

644: and unit element $i\mash 1$.

645: 

646: Next we consider extensions. Let $N\to M^H\subset M$ be an algebra map. Then we 

647: have left actions $\lambda$ of $N$ and $\lambda$ of $M^H$ on $M$. Denoting

648: $\E:=\End(\,_NM)$ we have an algebra map $H\mash M\to\E$ by 

649: \begin{equation} \label{eq: Gamma}

650: m'\cdot(h\mash m):=(m'\ract h)m

651: \end{equation}

652: so that $M$ becomes an $N$-$(H\mash M)$-bimodule. We have the inclusions

653: \begin{equation} \label{eq: to be bal}

654: \lambda(N)\subset \End_\E(M)\subset \End_{H\mash M}(M)=\lambda(M^H)

655: \end{equation}

656: where the last equality can be proven exactly as in the Hopf algebra case 

657: \cite[8.3.2]{Montgomery}.

658: \begin{defi}

659: An algebra homomorphism $\eta:N\to M$ is called a right $A$-extension for some 

660: DDA $A$ if $M$ is a right module algebra over $A$ and $\eta$ factorizes through 

661: $M^H\subset M$ via an algebra isomorphism $N\iso M^H$.

662: \end{defi}

663: Later on an $A$-extension will be meant in the narrower sense that $N=M^H$ but 

664: sometimes, as in Section \ref{s: fiber} we need this more categorical 

665: definition.

666: \begin{lem} \label{lem: ext bal}

667: Let $A$ be a DDA and $N\to M$ be a right $A$-extension. Then 

668: \begin{enumerate}

669: \item $_NM$ is balanced, i.e., $\BiEnd(\,_NM)=\lambda(N)$ and

670: \item $_NM_{H\mash M}$ is faithfully balanced iff the canonical map $H\mash 

671: M\to\E$ given by (\ref{eq: Gamma}) is an isomorphism. 

672: \end{enumerate}

673: \end{lem}

674: \begin{proof}

675: Both statements are immediate consequences of the fact that all the inclusions 

676: in (\ref{eq: to be bal}) reduce to equalities in case of $A$-extensions.

677: \end{proof}

678: 

679: 

680: 

681: 

682: 

683: 

684: 

685: 

686: \section{Galois extensions}

687: 

688: 

689: \subsection{The coaction picture}

690: 

691: Let $M$ be a right comodule algebra over the Hopf algebroid $V$ and let 

692: $N:=M^\coinv{V}$. Then the maps

693: \begin{align}

694: \gamma^M&:M\oN M\to M\oT V\,,\qquad m\oN m'\mapsto m{m'}\zeroT\oT {m'}\oneT\\

695: \gamma_M&:M\oN M\to M\oB V\,,\qquad m\oN m'\mapsto m\zeroB m'\oB m\oneB

696: \end{align}

697: are $M$-$M$-bimodule maps if we endow $M\oT V$ and $M\oB V$ with the structure

698: \begin{align}

699: m'\cdot(m\oT v)\cdot m''&=m'm{m''}\zeroT\oT v\ci {m''}\oneT\,,\\

700: m'\cdot(m\oB v)\cdot m''&=m'\zeroB mm''\oB m'\oneB\ci v\,,

701: \end{align}

702: respectively. They are also right $V$-comodule maps, i.e., belong to $\M^V$, 

703: because they can be written as composites of $\mu_M$ and $\delta^M$, 

704: respectively $\mu_M$ and $\delta_M$.

705: 

706: \begin{lem} \label{lem: 2 gamma}

707: Let $M$ be a right $V$-comodule algebra over the Hopf algebroid $V$. Then

708: $\gamma^M$ is epimorphism iff $\gamma_M$ is and $\gamma^M$ is 

709: isomorphism iff $\gamma_M$ is.

710: %\begin{alignat*}{3}

711: %&\gamma^M \text{ is epi}&\quad &\Leftrightarrow&\quad &\gamma_M \text{ is 

712: %epi,}\\

713: %&\gamma^M \text{ is iso}&\quad &\Leftrightarrow&\quad &\gamma_M \text{ is 

714: %iso.}

715: %\end{alignat*}

716: \end{lem}

717: \begin{proof}

718: Let $\phi$ denote the composite 

719: \begin{align}

720: &\begin{CD}

721: M\oT V@>\delta_M\oT V>>M\oB V\oT V@>M\oB V\o S>>M\oB V\oR V

722: @>M\oB\mu_V>>M\oB V

723: \end{CD}\\

724: &\ \ \ \ m\oT v\mapsto m\zeroB\oB m\oneB S(v) \notag

725: \end{align}

726: where $S$ is the antipode of the Hopf algebroid $V$. Then $\phi$ has inverse

727: \[

728: \phi^{-1}(m\oB v)=m\zeroT\oT S^{-1}(v) m\oneT\,.

729: \]

730: and one obtains that $\phi\ci\gamma^M=\gamma_M$.

731: \end{proof}

732: 

733: The next result is an immediate generalization of \cite[Theorem 

734: 8.3.1]{Montgomery}.

735: \begin{pro} \label{pro: epi implies iso}

736: Assume that $V$ is a Frobenius Hopf algebroid and $M$ is a right $V$-comodule 

737: algebra with coinvariant subalgebra $N$. Then $\gamma^M$ being epi implies that 

738: $\gamma^M$ is an isomorphism and $M_N$ is finitely generated projective.

739: \end{pro}

740: \begin{proof}

741: Let $V$ and $H$ be the vertical and horizontal Hopf algebroid of a distributive 

742: double algebra $\bra A,\ci,e,\star,i\ket$. Then $M$ is a right $H$-module 

743: algebra and $e$, the unit of $V$, is an integral for $H$, therefore $m\ract e\in 

744: N$, $m\in M$. By the hypothesis there exists $\sum_j m_j\oN m'_j\in M\oN M$ such 

745: that 

746: \[

747: \sum_j m_j{m'}_j\zeroT\oT {m'}_j\oneT=1\oT i\,.

748: \]

749: Therefore we can write for arbitrary $m\in M$ that

750: \begin{align*}

751: \sum_j m_j((m'_jm)\ract e)&=\sum_j m_j(m'_j\ract e\oneR)(m\ract e\twoR)\\

752: &=\sum_j m_j\left({m'}_j\zeroT\ud\Fi_T({m'}_j\oneT\star e\oneR)\right)(m\ract 

753: e\twoR)\\

754: &=(1\ract (i\ci e\oneR))(m\ract e\twoR)=(1\ract i\oneR)(m\ract i\twoR)\\

755: &=m\ract i\ =\ m

756: \end{align*}

757: proving that $(m'_j\under)\ract e$ is a dual basis of $m_j$ for $M_N$, thus 

758: $M_N$ is fgp.

759: 

760: Next we show that $\gamma_M$ is mono. Suppose $\sum_i z_i\oN 

761: w_i\in\Ker\gamma_M$. Then 

762: \[

763: \sum_i {z_i}\zeroB w_i\oB{z_i}\oneB\ =\ 0\,.

764: \]

765: Using the dual bases for $M_N$ we find that

766: \begin{align*}

767: \sum_i z_i\oN w_i&=\sum_i\sum_j m_j((m'_jz_i)\ract e)\oN w_i\\

768: &=\sum_j m_j\oN\sum_i\left({m'_j}\zeroB{z_i}\zeroB\ld\Fi_B(({m'_j}\oneB\ci 

769: {z_i}\oneB)\star e)\right)w_i\\

770: &=\sum_j m_j\oN\sum_i {m'_j}\zeroB{z_i}\zeroB w_i\ld\Fi_B({m'_j}\oneB\ci 

771: {z_i}\oneB)\\

772: &=0\,.

773: \end{align*}

774: Therefore $\gamma_M$ is mono. But it is also epi because $\gamma^M$ is. 

775: Therefore $\gamma_M$ is iso, and so is $\gamma^M$.

776: \end{proof}

777: 

778: 

779: 

780: 

781: \subsection{The action picture}

782: 

783: For a right bialgebroid $H$ over $R$ and an $H$-module algebra $M$ there are 

784: canonical maps

785: \begin{align}

786: \Gamma^M&:M\oR H\to\End(M_N)\,\qquad m\oR h\mapsto\{m'\mapsto m(m'\ract h)\}\\

787: \Gamma_M&:H\oR M\to\End(\,_NM)\,\qquad h\oR m\mapsto\{m'\mapsto (m'\ract h)m\}

788: \end{align}

789: being algebra maps from the smash products $M\mash H^\op$ and $H\mash M$, 

790: respectively, where in the latter case $\End(\,_NM)$ is considered with 

791: multiplication that arises from its natural right action on $M$. 

792: 

793: Note that if 

794: $H$ is the horizontal Hopf algebroid of a DDA and the right $H$ action arises 

795: from a right $V$-coaction as in Lemma \ref{lem: act-coact} then $M$ being a left 

796: $H^\op$-module algebra is in complete agreement with the familiar Hopf algebraic 

797: situation since it is $H^\op$ which is the dual of $V$. 

798: %The double algebraic 

799: %formalism, however, dictates to consider the right $H$-action on $M$ to be the 

800: %more natural. This is why the multiplication rule in $H\mash M$ looks nicer,

801: 

802: 

803: 

804: \begin{thm&def} \label{thm: def Gal}

805: Let $A$ be a distributive double algebra and $M$ a right $H$-module algebra, 

806: equivalently a right $V$-comodule algebra, over the horizontal, resp. vertical 

807: Hopf algebroid of $A$. Let $N= M^H \equiv M^\coinv{V}$. Then $N\subset M$ is 

808: called an $A$-Galois extension if any one of the following equivalent conditions 

809: hold:

810: %\begin{enumerate}

811: \begin{tabular}{llll}

812: (1)&$\gamma^M$ is epi.&(2)& $\gamma_M$ is epi.\\

813: (3)& $\gamma^M$ is iso.&(4)& $\gamma_M$ is iso.\\

814: (5)& $\Gamma^M$ is iso and $M_N$ is fgp.&

815: (6)& $\Gamma_M$ is iso and $_NM$ is fgp.

816: \end{tabular}

817: \end{thm&def}

818: \begin{proof}

819: Equivalence of the first four conditions follows from Proposition \ref{pro: epi 

820: implies iso} and Lemma \ref{lem: 2 gamma}.

821: 

822: $(3)\Rightarrow(5)$ Considering it as a right $M$-module map, $\gamma^M$ induces 

823: the isomorphism (of left $M$-modules)

824: \[

825: {\gamma^M}^*\colon\Hom_{-M}(M\oT V,M)\iso\Hom_{-M}(M\oN M,M)\,.

826: \]

827: If $\chi\in\Hom_{-M}(M\oT V,M)$ then $\chi(1\oT\under)\in \Hom(V_T,M_T)$ 

828: because

829: \begin{align*}

830: \chi(1\oT v\star t)&=\chi(1\oT v\ci\Fi_R(t))\\

831: &=\chi\left(j(\Fi_R(t))\zeroT\oT v\ci j(\Fi_R(t))\oneT\right)\\

832: &=\chi(1\oT v)j(\Fi_R(t))\,.

833: \end{align*}

834: Thus we have a well defined map (of left $M$-modules)

835: \begin{align}\label{map 1}

836: \Hom_{-M}(M\oT V,M)&\to \Hom(V_T,M_T)\\

837: \chi&\mapsto \chi(1\oT\under)\notag

838: \end{align}

839: We claim that this map is an isomorphism with inverse

840: \[

841: \kappa\mapsto \{m\oT v\mapsto\kappa(v\ci S^{-1}(m\oneB))m\zeroB\}

842: \]

843: This follows from the computation 

844: \begin{align*}

845: \chi(m\oT v)&=\chi(m\zeroT\oT\Fi_T(m\oneT)\star v)

846: =\chi(m\zeroT\oT v\ci \Fi_L\Fi_T(m\oneT))\\

847: &=\chi(m\zeroT\oT v\ci S^{-1}({m\oneT}\twoB)\ci {m\oneT}\oneB)\\

848: &=\chi({m\zeroB}\zeroT\oT v\ci S^{-1}(m\oneB)\ci{m\zeroB}\oneT)\\

849: &=\chi(1\oT v\ci S^{-1}(m\oneB))\,m\zeroB

850: \end{align*}

851: on the one hand and on the other hand from $\delta_M(1)=1\oB e$.

852: Composing the map (\ref{map 1}) with the isomorphism

853: \begin{align}

854: \Hom(V_T,M_T)&\to M\oR H\\

855: \kappa&\mapsto \kappa(x^j)\oR y^j\notag

856: \end{align}

857: where $x^j\oR y^j\equiv\cop_R(e)$ is the dual basis of $\Fi_R$, we obtain the 

858: left vertical arrow in the diagram

859: \begin{equation}\label{dia: g-G 1}

860: \begin{CD}

861: \Hom_{-M}(M\oT V,M)@>{\gamma^M}^*>>\Hom_{-M}(M\oN M,M)\\

862: @VVV @VVV\\

863: M\oR H@>\Gamma^M>> \End(M_N)

864: \end{CD}

865: \end{equation}

866: The vertical arrow on the right is the isomorphism $\sigma\mapsto 

867: \sigma(\under\oN 1)$ therefore the composite along the top and right is

868: $\chi\mapsto \chi(\under\oT e)$. The other two compose to give 

869: \[

870: \chi\mapsto \chi(1\oT x^j)\oR y^j\mapsto \chi(1\oT x^j)(\under\ract y^j) \,.

871: \]

872: In order to see commutativity of the diagram we need a calculation.

873: \begin{align*}

874: \chi(1\oT x^j)(m\ract y^j)&=

875: \chi(1\oT x^j)\,m\zeroT\ud\Fi_T(m\oneT\star y^j)\\

876: &=\chi(m\zeroT\oT x^j\ci m\oneT\ci\Fi_R\Fi_T(m\twoT\star y^j))\\

877: &=\chi(m\zeroT\oT x^j\ci(m\oneT\star\Fi_T(m\twoT\star y^j))\\

878: &=\chi(m\zeroT\oT x^j\ci (m\oneT\star y^j)\\

879: &=\chi(m\zeroT\oT\Fi_L\Fi_T(m\oneT))\\

880: &=\chi(m\zeroT\ud\Fi_T(m\oneT)\oT e)\ =\ \chi(m\oT e)\,,

881: \end{align*} 

882: where in the fifth equality we used \cite[Equation (4.16)]{Sz: DA}.

883: So (\ref{dia: g-G 1}) is commutative and therefore $\Gamma^M$ is an isomorphism.

884: 

885: The proof of $(4)\Rightarrow(6)$ goes similarly by proving commutativity of the 

886: diagram

887: \begin{equation}\label{dia: g-G 2}

888: \begin{CD}

889: \Hom_{M-}(M\oB V,M)@>{\gamma_M}^*>>\Hom_{M-}(M\oN M,M)\\

890: @VVV @VVV\\

891: H\oR M@>\Gamma_M>> \End(\,_NM)

892: \end{CD}

893: \end{equation}

894: with the left hand side arrow being the isomorphism $\chi\mapsto 

895: x^j\oR\chi(1\oB y^j)$ and the one on the right hand side being 

896: $\sigma\mapsto \sigma(1\oN \under)$. 

897: 

898: $(5)\Rightarrow(4)$ Consider the diagram

899: \begin{equation} \label{dia: G-g 1}

900: \begin{CD}

901: M\oN M@>\gamma_M>>M\oB V\\

902: @VVV @AAA\\

903: \Hom_{M-}(\End(M_N),M)@>{\Gamma^M}^*>>\Hom_{M-}(M\oR H,M)

904: \end{CD}

905: \end{equation}

906: The lower horizontal arrow is an isomorphism since $\Gamma^M$ is. The vertical 

907: arrow on the left, mapping $m\oN m'$ to the homomorphism $\alpha\mapsto 

908: \alpha(m)m'$, is an isomorphism because $M_N$ is fgp. The other vertical arrow

909: is the composite of two maps,

910: \[

911: \begin{CD}

912: \Hom_{M-}(M\oR H,M)@>>>\Hom(\,_RH,\,_RM)@>>>M\oB V

913: \end{CD}

914: \]

915: where the second one is the isomorphism $\kappa\mapsto \kappa(u^k)\oB v^k$ 

916: with $u^k\oB v^k\equiv \cop_B(i)$ denoting the dual basis of $\Fi_B$. The first 

917: one, $\chi\mapsto \chi(1\oR \under)$, is obviously invertible (in contrast to 

918: the similar map in the $(3)\Rightarrow(5)$ part) because the left $M$-module 

919: structure of $M\oR H$ we need here is the trivial one. It remains to show 

920: commutativity of (\ref{dia: G-g 1}). So we compute the action of the lower three 

921: arrows,

922: \begin{align*}

923: m\oN m'&\mapsto \{\alpha\mapsto \alpha(m)m'\} \mapsto\{m''\oR h\mapsto

924: m''(m\ract h)m'\}\\

925: &\mapsto \{h\mapsto(m\ract h)m'\} \mapsto (m\ract u^k)m'\oB v^k

926: \end{align*}

927: which is indeed $\gamma_M$ if we compare the right $H$-action with the right 

928: $V$-coaction $\delta_M$. This proves that $\gamma_M$ is invertible.

929: 

930: The proof of the implication $(6)\Rightarrow(3)$ can be done similarly by using 

931: the diagram

932: \begin{equation} \label{dia: G-g 2}

933: \begin{CD}

934: M\oN M@>\gamma^M>>M\oT V\\

935: @VVV @AAA\\

936: \Hom_{-M}(\End(\,_NM),M)@>{\Gamma_M}^*>>\Hom_{-M}(H\oR M,M)

937: \end{CD}

938: \end{equation}

939: where on the left hand side we have the map $m\oN m'\mapsto\{\alpha\mapsto 

940: m\alpha(m')\}$ which is an isomorphism because $_NM$ is fgp.

941: \end{proof}

942: \begin{rmk}

943: The terminology "right $A$-Galois extension" where $A$ is a distributive double 

944: algebra does not, by any means, imply that the choice of the integral $i$ in the 

945: vertical Hopf algebroid $V$ plays any role. This is clear from the 

946: coaction picture that uses $\gamma^M$ alone. Therefore we might as well call it

947: "right $V$-Galois extensions" which would then be in complete agreement with the 

948: Hopf-Galois terminology. Saying "$A$-Galois" we try to put the 

949: coaction and action pictures on equal footing. For example, "bottom $A$-Galois" 

950: and "top $A$-Galois" extensions correspond to the left and right $H$-Galois 

951: extensions in the Hopf-Galois language if $H$ denotes the horizontal Hopf 

952: algebroid of $A$.

953: \end{rmk}

954: 

955: 

956: \subsection{Weak and strong structure theorems}

957: 

958: For a Frobenius Hopf algebroid $A$ let $M$ be a right $H$-module algebra and $N= 

959: M^H $. The category  $(\M_H)_M$ of right $M$-modules in $\M_H$ (by the 

960: identification $\M_H=\M^V$ being the analogue of relative Hopf modules) is 

961: nothing but the category of right modules over the smash product,

962: \begin{equation}

963: (\M_H)_M\ =\ \M_{H\mash M}\,.

964: \end{equation}

965: Indeed, for any action $X\oR M\to X$, $x\oR m\mapsto x\cdot m$ in $\M_H$ one has

966: the smash product action

967: \[

968: X\o(H\mash M)\to X,\quad x\o(h\mash m)\mapsto (x\ract h)\cdot m\,.

969: \]

970: Vice versa, any $H\mash M$-module is an $H$-module and an $M$-module and the 

971: $M$-action is an $H$-module map. Considering $M$ as an $N$-$H\mash M$ bimodule, 

972: it defines an adjoint pair $F\dashv U$ of functors 

973: \begin{alignat*}{2}

974: F&:\M_N\to\M_{H\mash M}&\qquad X&\mapsto X\oN M\\

975: U&:\M_{H\mash M}\to \M_N&\qquad Y&\mapsto\Hom_{H\mash M}(M,Y)

976: \end{alignat*}

977: with counit and unit

978: \begin{align*}

979: \eps_Y&:\Hom_{H\mash M}(M,Y)\oN M\to M\qquad \chi\oN m\mapsto \chi(m)\\

980: \eta_X&:X\to\Hom_{H\mash M}(M,X\oN M)\qquad x\mapsto\{m\mapsto x\oN m\}

981: \end{align*}

982: We note that $UY$ is isomorphic to the submodule of invariants via

983: \begin{equation} \label{eq: Inv Y /1}

984: \Hom_{H\mash M}(M,Y)\iso\Hom_H(R,Y)\iso Y^H\,.

985: \end{equation}

986: \begin{lem} \label{lem: when eta eps iso}

987: For any $A$-extension $N\subset M$ 

988: \begin{enumerate}

989: \item if $M_{H\mash M}$ is fgp then $\eta$ is invertible,

990: \item if $_NM$ is fgp and $\Gamma_M$ is invertible then $\eps$ is invertible.

991: \end{enumerate}

992: \end{lem}

993: \begin{proof}

994: (1) Apply \cite[20.10]{A-F} to the last arrow in the decomposition of $\eta_X$

995: \[

996: \begin{CD} 

997: X@>\sim>> X\oN M@>\sim>>X\oN\Hom_{H\mash M}(M,M)@>>>

998: \Hom_{H\mash M}(M,X\oN M)

999: \end{CD}

1000: \]

1001: (2) Apply \cite[20.11]{A-F} to the first arrow in the decomposition of $\eps_Y$

1002: \[

1003: \begin{CD} 

1004: \Hom_{H\mash M}(M,Y)\oN M@.@.\\

1005: @VVV@.@.\\

1006: \Hom_{H\mash M}(\Hom_{N-}(M,M),Y)@>\Hom(\Gamma_M,Y)>>

1007: \Hom_{H\mash M}(H\mash M,Y)@>{\sim}>>Y

1008: \end{CD}

1009: \]

1010: \end{proof}

1011: \begin{thm} \label{thm: WS}

1012: Let $A$ be a distributive double algebra.

1013: \begin{enumerate}

1014: \item For an $A$-extension $N\subset M$ the following conditions are equivalent:

1015: \begin{enumerate}

1016: \item $\eps:FU\to\M_{H\mash M}$ is an isomorphism.

1017: \item $N\subset M$ is $A$-Galois.

1018: \item $_NM$ is fgp and $_NM_{H\mash M}$ is faithfully balanced.

1019: \item $M_{H\mash M}$ is a generator.

1020: \end{enumerate}

1021: \item  For an $A$-Galois extension $N\subset M$ the following conditions are 

1022: equivalent:

1023: \begin{enumerate}

1024: \item $F\dashv U$ is an adjoint equivalence.

1025: \item $\eta:\,_N\M\to UF$ is an isomorphism.

1026: \item $_NM_{H\mash M}$ is a Morita equivalence bimodule.

1027: \item $M_{H\mash M}$ is fgp.

1028: \item $_NM$ is a generator.

1029: \item $_NN\subset\,_NM$ is a direct summand.

1030: \end{enumerate}

1031: \end{enumerate}

1032: \end{thm}

1033: \begin{proof}

1034: $(1a) \Leftrightarrow(1b)$: The $\Leftarrow$ follows from Lemma \ref{lem: 

1035: when eta eps iso} (2). As for the $\Rightarrow$ direction consider $\eps_Y$ 

1036: for $Y=M\oT V$ which is a $H\mash M$-module via

1037: \[

1038: (m'\oT v)\cdot (h\mash m):=m'm\zeroT\oT(v\star h)\ci m\oneT\,.

1039: \]

1040: This is a well-defined action due to

1041: \[

1042: (m\ract h)\zeroT\oT (m\ract h)\oneT =m\zeroT\oT m\oneT\star h\,.

1043: \]

1044: Now consider the map

1045: \begin{equation} \label{eq: Inv Y /2}

1046: Y^H\to M\qquad

1047: \sum_j\ m_j\oT w_j\mapsto \sum_j\ m_j\ud\Fi_T(w_j) 

1048: \end{equation}

1049: which has inverse $m\mapsto m\oT e$. As a matter of fact,

1050: \begin{align*}

1051: m\ud\Fi_T(e)&=m\,\eta\Fi_R\Fi_T(e)=m\eta(e)=m\\

1052: \sum_j m_j\ud\Fi_T(w_j)\oT e&=\sum_jm_j\oT\Fi_L\Fi_T(w_j)=\sum_jm_j\oT(i\ci 

1053: w_j)\star e\\

1054: &=\sum_jm_j\oT(i\star e\oneR)\ci(w_j\star e\twoR)\\

1055: &=\sum_jm_j\oT e\oneR\ci(w_j\star\Fi_B\Fi_R(e\twoR))\\

1056: &=\sum_jm_j\oT e\oneR\ci\Fi_R(e\twoR)\ci w_j=\sum_j m_j\oT w_j\,.

1057: \end{align*}

1058: Composing $\eps_Y$ with the inverses of (\ref{eq: Inv Y /1}) and (\ref{eq: Inv 

1059: Y /2}) we obtain the mapping

1060: \[

1061: m\oN m'\mapsto (m\oT e)\oN m'\mapsto (m\oT e)\cdot m'=m{m'}\zeroT\oT{m'}\oneT

1062: =\gamma^M(m\oN m')

1063: \]

1064: Therefore $\gamma^M$ is invertible.

1065: 

1066: $(1b)\Leftrightarrow(1c)$: This is Lemma \ref{lem: ext bal} (2) together with 

1067: the Theorem \ref{thm: def Gal} (6).  

1068: 

1069: $(1c)\Leftrightarrow(1d)$: Since $N\subset M$ is an extension, $_NM$ is 

1070: balanced by  Lemma \ref{lem: ext bal} (1). So (c) is equivalent to that $_NM$ 

1071: is fgp and $M_{H\mash M}$ is faithful and balanced. But these are the necessary 

1072: and sufficient conditions for (d) by \cite[Theorem 17.8]{A-F}.

1073: 

1074: $(2a)\Leftrightarrow(2b)$: This is clear from the equivalence of (1a) and 

1075: (1b).

1076: 

1077: $(2b)\Leftrightarrow(2c)$: Consider the composite  

1078: \[

1079: \begin{CD}

1080: \Hom_N(N,X)@.\Hom_N(M\amalgo{H\mash M}\Hom(\,_NM,\,_NN),X)\\

1081: @VV{\wr}V @A{\wr}AA\\

1082: X@>\eta>>\Hom_{H\mash M}(M,X\oN M)

1083: \end{CD}

1084: \]

1085: of natural isomorphisms where the last isomorphism exists because 

1086: $_NM$ is fgp. By the Yoneda lemma this determines an isomorphism

1087: \[

1088: M\amalgo{H\mash M}\Hom(\,_NM,\,_NN)\ \to\ N\qquad\in\,_N\M

1089: \]

1090: which is nothing but the evaluation associated to the right dual of the bimodule 

1091: $_NM_{H\mash M}$. Postulating the usual right $N$-module structure on 

1092: $\Hom(\,_NM,\,_NN)$ it becomes in fact an $N$-$N$-bimodule isomorphism. 

1093: Another hom-tensor relation for fgp $_NM$ and the isomorphism $\Gamma_M$ compose 

1094: to give

1095: \[

1096: \Hom(\,_NM,\,_NN)\oN M\iso\Hom(\,_NM,\,_NM)\iso H\mash M\qquad

1097: \in\,_{H\mash M}\M_{H\mash M}\,.

1098: \]

1099: Thus $\Hom(\,_NM,\,_NN)$ is the inverse equivalence of $_NM_{H\mash M}$.

1100: It follows from Morita theory that both $_NM$ and $M_{H\mash M}$ are 

1101: progenerators which prove that $(2c)\Rightarrow(2d)$ and 

1102: $(2c)\Rightarrow(2e)$.

1103: 

1104: $(2d)\Leftrightarrow(2b)$ follows from Lemma \ref{lem: when eta eps iso} 

1105: (1). 

1106: 

1107: In order to show $(2e)\Rightarrow(2f)$ we use that an $N$-module $M$

1108: is a generator iff a finite direct sum of $M$'s contains the regular 

1109: object as a summand, i.e., there exist $N$-module maps 

1110: $N\rarr{\iota_k}M\rarr{\pi_k}N$ such that $\sum_k\pi_k\circ\iota_k=N$. In this

1111: case $\{m\mapsto\sum_k\pi_k(m\,\iota_k(1))\}\in\Hom(\,_NM,\,_NN)$ splits 

1112: the inclusion $N\subset M$. The implication $(2f)\Rightarrow(2e)$ is now 

1113: obvious.

1114: 

1115: Finally $(2e)\Rightarrow(2d)$ follows from that $_NM_{H\mash M}$ is 

1116: faithfully balanced by Lemma \ref{lem: ext bal}. 

1117: \end{proof}

1118: 

1119: 

1120: 

1121: 

1122: 

1123: 

1124: 

1125: 

1126: 

1127: 

1128: 

1129: 

1130: 

1131: 

1132: 

1133: 

1134: 

1135: 

1136: 

1137: 

1138: 

1139: 

1140: 

1141: 

1142: 

1143: 

1144: 

1145: 

1146: 

1147: 

1148: 

1149: 

1150: 

1151: 

1152: \subsection{An intrinsic characterization of finitary Galois extensions}

1153: 

1154: \begin{thm} \label{thm: bD2F}

1155: For an algebra extension $N\subset M$ the following conditions are equivalent.

1156: \begin{enumerate}

1157: \item There is a Frobenius Hopf algebroid $V$ and a coaction of $V$ on $M$ such 

1158: that $N\subset M$ is $V$-Galois.

1159: \item $N\subset M$ is of depth 2 and Frobenius and $M_N$ is balanced.

1160: \end{enumerate}

1161: \end{thm}

1162: \begin{proof}

1163: $(1)\Rightarrow$ $N\subset M$ \textit{is Frobenius}: Consider the composite 

1164: \begin{equation} \label{map 3}

1165: \begin{CD}

1166: M\oN M@>\gamma^M>>M\oT V@>M\o S>>M\oR H@>\Gamma^M>>\End(M_N)

1167: \end{CD}

1168: \end{equation}

1169: where the middle arrow is meaningful in the double algebraic picture because $V$ 

1170: and $H$ have the same underlying $k$-module $A$ and $S(t\star a)=S(a)\star 

1171: \Fi_B\Fi_R(t)=\Fi_R(t)\ci a$ holds for all $a\in A$, $t\in T$, see \cite[Lemma 

1172: 5.4]{Sz: DA}. Computing the value of the map (\ref{map 3}) on $m\oN m'$ we 

1173: obtain

1174: \begin{align*}

1175: m{m'}\zeroT(m''\ract S({m'}\oneT))&=

1176: m{m'}\zeroT{m''}\zeroT\ud \Fi_T({m''}\oneT\star S({m'}\oneT))\\

1177: &=m{m'}\zeroT{m''}\zeroT\ud \Fi_T\Fi_L({m'}\oneT\ci{m''}\oneT)\\

1178: &=m(m'm'')\zeroT\ud\Fi_T((m'm'')\oneT\star e)\\

1179: &=m((m'm'')\ract e)

1180: \end{align*}

1181: Therefore (\ref{map 3}) has the familiar form $m\oN m'\mapsto m\psi m'$ in terms 

1182: of the $N$-$N$-bimodule map $\psi=\under \ract e$ from $M$ into $N$. Since 

1183: (\ref{map 3}) is isomorphism it follows that $\psi$ is a Frobenius homomorphism 

1184: with dual basis obtained from $\id_M$ by applying the inverse of (\ref{map 3}) .

1185: 

1186: $(1)\Rightarrow$ $N\subset M$ \textit{is D2}: Since $_TV$ is fgp and $\gamma^M$ 

1187: provides an $M$-$N$-bimodule isomorphism $M\oN M\iso (_MM_N)\oT V$, it 

1188: follows that $N\subset M$ is right D2. Similarly, the existence of the 

1189: isomorphism $\gamma_M$ and the $_BV$ being fgp imply that $N\subset M$ is left 

1190: D2.

1191: 

1192: $(1)\Rightarrow$ $N\subset M$ \textit{is balanced}: This follows from that every 

1193: $V$-extension is balanced, see Lemma \ref{lem: ext bal}.

1194: 

1195: $(2)\Rightarrow(1)$: The endomorphism algebra $H^\op:=\End(\,_NM_N)$ has a 

1196: natural structure of a Frobenius Hopf algebroid, see \cite[Subsection 8.6]{Sz: 

1197: DA} or \cite{BSz: D2F}. Moreover, the natural action of $H^\op$ on $M$ makes it 

1198: a left $H^\op$-module algebra and the corresponding smash product $M\mash H^\op$ 

1199: is isomorphic to $\End(M_N)$ via $\Gamma^M$ by \cite[Corollary 4.5]{K-Sz}. So 

1200: $N\subset M$ will be $V$-Galois, for $V$ the dual of $H^\op$, provided  

1201: $N= M^H $. But this is equivalent to $M_N$ being balanced.

1202: \end{proof}

1203: 

1204: Note that in the presence of the Frobenius condition left D2 is equivalent to 

1205: right D2 and in the presence of the D2 Frobenius condition $M_N$ is balanced iff 

1206: $_NM$ is balanced. 

1207: 

1208: 

1209: %\subsection{The endomorphism Hopf algebroid}

1210: 

1211: 

1212: 

1213: 

1214: 

1215: 

1216: 

1217: 

1218: 

1219: 

1220: 

1221: 

1222: 

1223: 

1224: \section{Noncommutative scalar extensions}

1225: 

1226: The Hopf algebroid $V$ making a given algebra extension $V$-Galois is highly 

1227: nonunique. This phenomenon can be observed already for Hopf Galois extensions. 

1228: As Greither and Pareigis have shown \cite{Greither-Pareigis} certain separable 

1229: field extensions can be $H$-Galois for two different Hopf algebras $H$ and $H'$.

1230: By an appropriate extension $k\subset K$ of the scalars, however, they become

1231: isomorphic, $K\o H\cong K\o H'$, as $K$-Hopf algebras. The $k$-Hopf algebras 

1232: $H$, $H'$ for which such a (commutative, faithfully flat) $k$-algebra $K$ exists 

1233: are called forms of each other \cite{Pareigis: forms}.

1234: 

1235: If we admit Hopf algebroids to appear in place of Hopf algebras then an 

1236: interesting generalization of scalar extension is provided by the 

1237: Brzezi\'nski-Militaru theorem \cite{Brz-Mil} constructing a Hopf algebroid 

1238: structure on the smash product $M\mash H$ if $M$ is a braided commutative 

1239: algebra in the Yetter-Drinfeld category $_H\YD^H$ over the Hopf algebra $H$.

1240: As we shall see the Brzezi\'nski-Militaru theorem holds also for $H$ a 

1241: bialgebroid or Frobenius Hopf algebroid. Since the base algebra 

1242: of $M\mash H$ is just $M$, the braided commutative algebras (BCA's) play the 

1243: role of (noncommutative) scalars.

1244: 

1245: If $N\subset M$ is a Galois extension for some Frobenius Hopf algebroid $H$ then 

1246: the center $C=M^N$ of the extension is a BCA over $H$ (Corollary \ref{cor: C 

1247: BCA}) and the scalar extension $H\mash C$ is the endomorphism Hopf algebroid 

1248: $E$ (Proposition \ref{pro: E as smash}). Therefore all Frobenius Hopf algebroids 

1249: $H$ for which $N\subset M$ is $H$-Galois are forms of each other.

1250: 

1251: \subsection{Braided commutative algebras}

1252: 

1253: Yetter-Drinfeld modules over bialgebroids have been introduced in 

1254: \cite{Schauenburg: ddqg}. They form a prebraided monoidal category, the 

1255: weak center of the category of modules over the bialgebroid.

1256: In this subsection we adapt the weak center construction to the double algebraic 

1257: notation and describe the (braided) center $\Z(\M_H)$ as `double' 

1258: Yetter-Drinfeld modules $^H\YD^H_H$ with two related coactions.

1259: 

1260: For a right bialgebroid $H$ over $R$ the weak 

1261: center $\lZ(\M_H)$ is defined as follows. The objects $\bra Z,\theta\ket$ are 

1262: $H$-modules equipped with a natural transformation $\theta_Y:Z\oR Y\to Y\oR Z$ 

1263: satisfying

1264: \begin{equation} \label{eq: theta}

1265: \theta_{X\oR Y}=(X\oR\theta_Y)\circ(\theta_X\oR Y)\quad\text{and}\quad 

1266: \theta_R=Z

1267: \end{equation}

1268: where the coherence isomorphisms are not written out explicitly.

1269: An arrow $\bra Z,\theta\ket\to\bra Z',\theta'\ket$ is an $H$-module map 

1270: $\alpha:Z\to Z'$ such that

1271: \begin{equation} \label{eq: arrows in lZ}

1272: (Y\oR \alpha)\circ\theta_Y=\theta'_Y\circ(\alpha\oR Y)

1273: \end{equation}

1274: for all objects $Y\in\M_H$. This category has a monoidal product which is 

1275: defined for objects by

1276: \[

1277: \bra Z,\theta\ket\oR\bra Z',\theta'\ket=\bra Z\oR Z',(\theta_-\oR Z')\circ

1278: (Z\oR\theta'_-\ket

1279: \]

1280: and for arrows by taking the ordinary tensor product in $\M_H$. The category 

1281: $\lZ(\M_H)$ is prebraided with

1282: \[

1283: \beta_{\bra Z,\theta\ket,\bra Z',\theta'\ket}=\theta_{Z'}\,.

1284: \]

1285: 

1286: Given an object $\bra Z,\theta\ket\in\lZ(\M_H)$ one can introduce

1287: \begin{equation}

1288: \tau:Z\to H\oR Z\,,\qquad \tau(z):=\theta_H(z\oR i)=z\coa{-1}\oR z\coa{0}

1289: \end{equation}

1290: which, as being the composite

1291: \begin{equation} \label{eq: tau}

1292: \begin{CD}

1293: Z@>\sim>>Z\oR R@>Z\oR\Fi_B>>Z\oR H@>\theta_H>>H\oR Z\,,

1294: \end{CD}

1295: \end{equation}

1296: preserves the left $R$-module structures inherited from $\M_H$. By naturality of 

1297: $\theta$, the $\tau$ determines $\theta_X$ for all $X$ by the formula

1298: \begin{equation} \label{eq: theta from tau}

1299: \theta_X(z\oR x)= x\ract z\coa{-1}\oR z\coa{0}\,.

1300: \end{equation}

1301: Using this formula it is easy to show that (\ref{eq: theta}) implies that 

1302: $\tau$ is coassociative and counital, thereby making $Z$ a left 

1303: $H$-comodule. We not only have Takeuchi's centrality property 

1304: \begin{align} \label{eq: T for tau}

1305: \Fi_T(r)\star z\coa{-1}\oR z\coa{0}&=\Fi_T(r)\ract z\coa{-1}\oR z\coa{0}

1306: =\theta_H(z\oR\Fi_T(r)) \notag\\

1307: &=\theta_H((z\oR i)\ract \Fi_T(r))=\tau(z)\ract\Fi_T(r) \notag\\

1308: &=z\coa{-1}\oR z\coa{0}\ract \Fi_T(r)\qquad r\in R,\ z\in Z

1309: \end{align}

1310: but also

1311: \begin{align} 

1312: \Fi_B(r)\star z\coa{-1}\oR z\coa{0}&=\Fi_B(r)\ract z\coa{-1}\oR z\coa{0}

1313: =\theta_H(z\oR\Fi_B(r)) \notag\\

1314: &=\theta_H(z\ract \Fi_T(r)\oR i)=\tau(z\ract\Fi_T(r))=\tau(z\cdot r)\,. 

1315: \end{align}

1316: The latter means that the right $R$-action we could construct from the left 

1317: $R$-action - in analogy with the left action we had in Proposition \ref{pro: 

1318: coactions are bim} for right comodules - would be the same as the original right 

1319: $R$-module structure inherited from (\ref{eq: tau}). In other words, 

1320: the requirement for (\ref{eq: tau}) to be an $R$-$R$-bimodule map defines a 

1321: right $R$-action on $H\oR Z$ which is conveyed by naturality of $\theta$ and not 

1322: by $\theta_H$ being an arrow in $_R\M_R$.

1323: 

1324: Given a left $H$-comodule $Z$ which is also a right $H$-module

1325: (with the same underlying $R$-$R$-bimodule structure) the condition for 

1326: (\ref{eq: theta from tau}) to determine an $H$-module map is precisely the 

1327: Yetter-Drinfeld condition given below.

1328: 

1329: Summarizing, one has a prebraided monoidal isomorphism $\lZ(\M_H)\cong 

1330: \,^H\YD_H$ with the following Yetter-Drinfeld category:

1331: \begin{defi}

1332: For a right bialgebroid $\bra H,\star,i,R,\Fi_T,\Fi_B,\cop_R,\Fi_R\ket$ the 

1333: category $^H\YD_H$ has objects $\bra Z,\ract,\tau\ket$ where 

1334: \begin{enumerate}

1335: \item $\bra Z,\ract\ket$ is a right $H$-module, hence also an 

1336: $R$-$R$-bimodule via $r\cdot z\cdot r'=z\ract(\Fi_B(r)\star\Fi_T(r'))$.

1337: \item $\bra Z,\tau\ket$ is a left $H$-coaction, that is to say,

1338: \begin{enumerate}

1339: \item $\tau:Z\to H\oR Z$ is an $R$-$R$-bimodule map in the sense of

1340: \begin{equation} \label{eq: tau bim}

1341: (r\cdot z\cdot r')\coa{-1}\oR(r\cdot z\cdot r')\coa{0}

1342: =\Fi_B(r')\star z\coa{-1}\star\Fi_B(r)\oR z\coa{0}\,,

1343: \end{equation}

1344: \item $\tau$ is coassociative and counital,

1345: \begin{align*}

1346: z\coa{-1}\oR {z\coa{0}}\coa{-1}\oR{z\coa{0}}\coa{0}&=

1347: {z\coa{-1}}\oneR\oR{z\coa{-1}}\twoR\oR z\coa{0}\\

1348: \Fi_R(z\coa{-1})\cdot z\coa{0}&=z

1349: \end{align*}

1350: \item $\tau$ factorizes through $H\ex{R}Z\subset H\oR Z$, i.e., 

1351: (\ref{eq: T for tau}) holds.

1352: \end{enumerate}

1353: \item The action and coaction satisfy the Yetter-Drinfeld condition

1354: \begin{equation*}

1355: h\twoR\star(z\ract h\oneR)\coa{-1}\oR(z\ract h\oneR)\coa{0}

1356: =z\coa{-1}\star h\oneR\oR z\coa{0}\ract h\twoR\,.

1357: \end{equation*}

1358: \end{enumerate}

1359: The arrows are the $H$-module $H$-comodule maps $Z\to Z'$. The monoidal product 

1360: of two Yetter-Drinfeld modules $Z$ and $Z'$ is $Z\oR Z'$ equipped with 

1361: \begin{align*}

1362: (z\oR z')\ract h&=(z\ract h\oneR)\oR(z'\ract h\twoR)\\

1363: (z\oR z')\coa{-1}\oR(z\oR z')\coa{0}&=

1364: {z'}\coa{-1}\star z\coa{-1}\oR(z\coa{0}\oR {z'}\coa{0})

1365: \end{align*}

1366: The monoidal unit is $R$ with $r\ract h=r\star h$ and $r\coa{-1}\oR 

1367: r\coa{0}=\Fi_B(r)\oR e$. The prebraiding is defined by

1368: \[

1369: \beta_{Z,Z'}:Z\oR Z'\to Z'\oR Z,\qquad z\oR z'\mapsto z'\ract z\coa{-1}\oR 

1370: z\coa{0}\,.

1371: \]

1372: \end{defi}

1373: 

1374: There is a coopposite version $\rZ(\M_H)= \lZ(\M_H^\coop)=\lZ(\M_{H^\coop})$ of 

1375: the left weak center, called the right weak center, in which an object $\bra 

1376: Z,\bar\theta\ket$ has natural transformation $\bar\theta_Y:Y\oR Z\to Z\oR Y$ 

1377: satisfying $\bar\theta_{X\oR Y}=(\bar\theta_X\oR Y)\circ(X\oR\bar\theta_Y)$.

1378: This determines a right coaction

1379: \[

1380: \bar\tau:Z\to Z\oR H\,,\qquad z\mapsto z\coa{0}\oR z\coa{1}=\bar\theta_H(i\oR z)

1381: \]

1382: and is determined by this coaction,

1383: \begin{equation} \label{eq: bartheta from bartau}

1384: \bar\theta_Y(y\oR z)=z\coa{0}\oR y\ract z\coa{1}\,.

1385: \end{equation}

1386: The center $\Z(\M_H)$ is the full subcategory of $\lZ(\M_H)$ in which the 

1387: objects $\bra Z,\theta\ket$ have invertible $\theta$. For such objects $\bra 

1388: Z,\theta^{-1}\ket$ is an object in $\rZ(\M_H)$ in which $\bar\theta$ is 

1389: invertible. The center is braided monoidal. In the language of 

1390: Yetter-Drinfeld modules the objects of the center are two-sided Yetter-Drinfeld 

1391: modules $\bra Z,\ract,\tau,\bar\tau\ket\in\,^H\YD^H_H$ in which the two 

1392: coactions are inverse to each other, i.e.,

1393: \begin{align}

1394: {z\coa{0}}\coa{0}\oR z\coa{-1}\star{z\coa{0}}\coa{1}&=z\oR i

1395: \label{eq: tau-bartau 1}\\

1396: z\coa{1}\star{z\coa{0}}\coa{-1}\oR{z\coa{0}}\coa{0}&=i\oR z\,.

1397: \label{eq: tau-bartau 2}

1398: \end{align}

1399: 

1400: \begin{defi}

1401: For a right bialgebroid $H$ the commutative monoids in $\Z(\M_H)$ are called 

1402: BCA's (braided commutative algebras) over $H$. The commutative monoids in 

1403: $\lZ(\M_H)$ and $\rZ(\M_H)$ are called left and right pre-BCA's over $H$, 

1404: respectively. 

1405: \end{defi}

1406: Therefore a left pre-BCA consists of

1407: an algebra $Q$ with an algebra map $\eta:R\to Q$ and

1408: a Yetter-Drinfeld module structure $\bra Q,\ract,\tau\ket\in\,^H\YD_H$ 

1409: such that

1410: \begin{align}

1411: \eta(r)\,q\,\eta(r')&=r\cdot q\cdot r'\\

1412: (qq')\ract h&=(q\ract h\oneR)(q'\ract h\twoR)\\

1413: 1\ract h&=\eta\,\Fi_R(h)\\

1414: (qq')\coa{-1}\oR(qq')\coa{0}&={q'}\coa{-1}\star q\coa{-1}\oR 

1415: q\coa{0}{q'}\coa{0}\\

1416: \eta(r)\coa{-1}\oR \eta(r)\coa{0}&=\Fi_B(r)\oR 1

1417: \end{align}

1418: and the prebraided commutativity

1419: \begin{equation}

1420: (q'\ract q\coa{-1})q\coa{0}=qq'

1421: \end{equation}

1422: holds. If $Q$ is a BCA then there exists also a right coaction $\bar\tau$ with 

1423: which $\bra Q,\ract,\bar\tau\ket\in\YD^H_H$ and which is inverse to $\tau$ in 

1424: the sense of equations (\ref{eq: tau-bartau 1}), (\ref{eq: tau-bartau 2}).

1425: 

1426: We note that the ground ring $R$ of the bialgebroid is always a BCA with the 

1427: structure $\bra R,\mu_R,R\ket$ that comes from $R$ being the monoidal unit of 

1428: $\lZ(\M_H)$.

1429: 

1430: \subsection{The centralizer of a Galois extension}

1431: 

1432: Interesting examples for BCA's are obtained from considering 

1433: centralizers $M^N$ of Galois extensions. 

1434: \begin{pro}

1435: Let $M$ be a monoid in $\M_H$ over the right bialgebroid $H$ and let $N=M^H$.

1436: Assume that $H_R$ is fgp and that the canonical map

1437: $\Gamma_M:H\mash M\to\End(\,_NM)$ is an isomorphism.

1438: Then the centralizer $M^N=\{c\in M\,|\,nc=cn,\ n\in N\}$ of the extension 

1439: $N\subset M$ is a left pre-BCA over $H$ with $H$-module algebra structure 

1440: inherited from $M^N\subset M$ (the Miyashita-Ulbrich action) and with left 

1441: coaction $\tau(c):=\Gamma_M^{-1}(\lambda_M(c))$ where $\lambda_M(c)=\{m\mapsto 

1442: cm\}$.

1443: \end{pro}

1444: \begin{proof}

1445: For each $h\in H$ the action $\under\ract h$ is an $N$-$N$-bimodule map.

1446: Therefore $M^H\subset M$ is a sub-$H$-module algebra. As such the unit 

1447: $\eta:R\to M$ has image in $M^H$. 

1448: Since $\Gamma_M$ is an $N$-$N$-bimodule map, it restricts to an isomorphism 

1449: $(H\oR M)^N\iso\End(\,_NM_N)$ between the centralizers. The $H_R$ being fgp we 

1450: have $(H\oR M)^N=H\oR M^N$. Since $\lambda(c)$ for $c\in M^N$ belongs to 

1451: $\End(\,_NM_N)$, the $\tau$ is a map $M^N\to H\oR M^N$. The $\tau$ is uniquely 

1452: determined by the equation

1453: \begin{equation}  \label{eq: tau in action}

1454: (m\ract c\coa{-1})c\coa{0}=cm\,,\qquad m\in M

1455: \end{equation}

1456: from which the bimodule property (\ref{eq: tau bim}) and the centrality 

1457: (\ref{eq: T for tau}) easily follow. The calculation 

1458: \begin{align*}

1459: c(mm')&=((mm')\ract c\coa{-1})c\coa{0}=(m\ract {c\coa{-1}}\oneT)

1460: (m'\ract{c\coa{-1}}\twoT)c\coa{0}\\

1461: (cm)m'&=(m\ract c\coa{-1})c\coa{0}m'=

1462: (m\ract c\coa{-1})(m'\ract{c\coa{0}}\coa{-1}){c\coa{0}}\coa{0}

1463: \end{align*}

1464: will imply coassociativity after verifying the next

1465: \begin{lem} \label{lem: isomorphisms}

1466: Under the assumptions of the Proposition and with the 

1467: notations $E:=\End(\,_NM_N)$, $C:=M^N$ the maps

1468: \begin{align}

1469: E\oC E&\to \Hom_{N\text{-}N}(M\oN M,M) \label{eq: fork rule}\\

1470: \alpha\oC\alpha'&\mapsto\{m\oN m'\mapsto\alpha(m)\alpha'(m')\} \notag\\

1471: H\oR H\oR C&\to \Hom_{N\text{-}N}(M\oN M,M) \label{eq: fine fork}\\

1472: h\oR h'\oR c&\mapsto\{m\oN m'\mapsto(m\ract h)(m'\ract h')c\}\notag

1473: \end{align}

1474: are isomorphisms.

1475: \end{lem}

1476: \begin{proof}

1477: Using both the isomorphism $\Gamma_M$ and its restriction $H\mash C\iso E$  

1478: we have a sequence of isomorphisms

1479: \begin{align*}

1480: E\oC E&\iso(H\oR C)\oC E\iso H\oR E= H\oR\Hom_{N\text{-}N}(M,M)\\

1481: &\iso\Hom_{N\text{-}N}(M,H\oR M)\iso\Hom_{N\text{-}N}(M,\Hom_{N\text{-}}(M,M))\\

1482: &\iso\Hom_{N\text{-}N}(M\oN M,M)

1483: \end{align*}

1484: The action of these isomorphisms can be computed by inserting 

1485: $\alpha=(\under\ract h)c$ and $\alpha'=(\under\ract h')c'$:

1486: \begin{align*}

1487: \alpha\oC\alpha'&\mapsto(h\oR c)\oC\alpha'\mapsto h\oR c\,\alpha'(\under)\mapsto

1488: \{m\mapsto h\oR c\,\alpha'(m)\}\\

1489: &\mapsto\{m\mapsto\{m'\mapsto\alpha(m')\alpha'(m)\}\}\mapsto

1490: \{m'\oN m\mapsto\alpha(m')\alpha'(m)\}

1491: \end{align*}

1492: This proves that (\ref{eq: fork rule}) is an isomorphism. The map in (\ref{eq: 

1493: fine fork}) is the composite

1494: \begin{equation*}

1495: \begin{CD}

1496: H\oR H\oR C@.@.\Hom_{N\text{-}N}(M\oN M,M)\\

1497: @V{H\oR\Gamma}VV @.  @AA{\cong}A\\

1498: H\oR E@>\cong>>H\oR C\oC E@>\Gamma\oC E>>E\oC E

1499: \end{CD}

1500: \end{equation*}

1501: of isomorphisms. 

1502: \end{proof}

1503: Returning to the proof of the Proposition counitality of $\tau$

1504: can be seen as

1505: \[

1506: \Fi_R(c\coa{-1})\cdot c\coa{0}=(1\ract c\coa{-1})c\coa{0}=c1=c\,.

1507: \]

1508: As for the Yetter-Drinfeld compatibility condition 

1509: it suffices to verify the equality

1510: \begin{align*}

1511: (m\ract c\coa{-1}\star h\oneT)(c\coa{0}\ract h\twoT)&=\left((m\ract 

1512: c\coa{-1})c\coa{0}\right)\ract h=(cm)\ract h\\

1513: =(c\ract h\oneT)(m\ract h\twoT)&=\left(m\ract h\twoT\star(c\ract 

1514: h\oneT)\coa{-1}\right)(c\ract h\oneT)\coa{0}

1515: \end{align*}

1516: In order to see compatibility of $\tau$ with multiplication and unit in $C$ it 

1517: suffices to check

1518: \[

1519: c'cm=c'(m\ract c\coa{-1})c\coa{0}=

1520: (m\ract c\coa{-1}\star {c'}\coa{-1}){c'}\coa{0}c\coa{0}\,.

1521: \]

1522: Finally, braided commutativity $(c'\ract c\coa{-1})c\coa{0}=cc'$

1523: follows from the more general relation (\ref{eq: tau in action}).

1524: \end{proof}

1525: 

1526: \begin{cor} \label{cor: C BCA}

1527: If $N\subset M$ is a right $A$-Galois extension for a distributive double 

1528: algebra $A$ then $M^N$ is a BCA over the horizontal Hopf algebroid $H$.

1529: \end{cor}

1530: \begin{proof}

1531: It suffices to prove that the prebraiding is invertible. Define the right 

1532: coaction $\bar\tau(c):=(\Gamma^M)^{-1}(\rho_M(c))$ where $\rho_M$ is 

1533: right multiplication on $M$. This is equivalent to $\bar\tau(c)=c\coa{0}\oR 

1534: c\coa{1}$ satisfying

1535: \begin{equation} \label{eq: bartau in action}

1536: c\coa{0}(m\ract c\coa{1})=mc\,,\qquad m\in M\,.

1537: \end{equation}

1538: Applying (\ref{eq: tau in action}) to (\ref{eq: bartau in action}) we obtain

1539: \[

1540: (m\ract i)c=mc=(m\ract c\coa{-1}\star{c\coa{0}}\coa{-1})\,{c\coa{0}}\coa{0}

1541: \]

1542: from which equation (\ref{eq: tau-bartau 2}) follows. Equation (\ref{eq: 

1543: tau-bartau 1}) can be seen similarly.

1544: \end{proof}

1545: Notice that this proof does not use very much from the Hopf algebroid structure.

1546: Therefore the Corollary holds true for any right bialgebroid for which both 

1547: $H_R$ and $_RH$ are fgp and for all extensions for which both $\Gamma^M$ and 

1548: $\Gamma_M$ are invertible.

1549: 

1550: 

1551: \subsection{Extensions by BCA's}

1552: 

1553: For any $H$-module algebra $Q$ over the right bialgebroid $H$ the category

1554: $\M_{H\mash Q}$ of modules over the smash product can be identified with the

1555: category of (internal) $Q$-modules $(\M_H)_Q$ in $\M_H$.

1556: 

1557: If $Q$ is also a pre-BCA then every right $Q$-module in $\M_H$ is also a left 

1558: $Q$-module by pre-braided commutativity. This defines an embedding of categories

1559: \begin{equation} \label{eq: BM 1}

1560: \M_{H\mash Q}=(\M_H)_Q\hookrightarrow \,_Q(\M_H)_Q

1561: \end{equation}

1562: into the monoidal category of internal $Q$-$Q$-bimodules. Since the 

1563: $Q$-$Q$-bimodule tensor product of diagonal bimodules $X,Y\in(\M_H)_Q$ is again 

1564: diagonal due to one of the hexagons, this embedding is actually strong monoidal.

1565: Composing (\ref{eq: BM 1}) with the strong monoidal forgetful functor 

1566: $_Q(\M_H)_Q\to \,_Q\M_Q$ we obtain a strong monoidal functor

1567: \begin{equation} \label{eq: BM 2}

1568: \M_{H\mash Q}=(\M_H)_Q\to \,_Q\M_Q\,.

1569: \end{equation}

1570: This functor is precisely the forgetful functor associated to the algebra map

1571: \begin{equation} \label{eq: forget along}

1572: Q^\op\o Q\to H\mash Q\,, \qquad q\o q'\mapsto q\coa{-1}\mash q\coa{0}q'

1573: \end{equation}

1574: therefore, by a theorem of Schauenburg \cite{Schauenburg: bnrsthb}, there is

1575: a unique bialgebroid structure on $H\mash Q$ such that the given monoidal 

1576: structure of $\M_{H\mash Q}$ is that of the module category of a bialgebroid. 

1577: This is the Brzezi\'nski-Militaru Theorem in disguise. More precisely this is 

1578: the "only if" part of \cite[Theorem 4.1]{Brz-Mil} generalized to bialgebroids 

1579: $H$.\begin{thm}

1580: Let $H$ be a right bialgebroid over $R$ and let $Q$ be a left pre-BCA over $H$. 

1581: Then the smash product $G:=H\mash Q$ is a right bialgebroid over $Q$ with 

1582: structure maps

1583: \begin{align}

1584: s_G(q)&=i\mash q \label{s_G}\\ 

1585: t_G(q)&=q\coa{-1}\mash q\coa{0}\label{t_G}\\

1586: \cop_G(h\mash q)&=(h\oneR\mash 1)\oQ(h\twoR\mash q)\label{cop_G}\\

1587: \eps_G(h\mash q)&=\eta(\eps_H(h))q\label{eps_G}

1588: \end{align}

1589: where $\eta:R\to Q$ is the unit of $Q$. Moreover, $h\mapsto h\mash 1$ is a

1590: bialgebroid map $\iota:H\to G$.

1591: 

1592: If $H$ is a Frobenius Hopf algebroid with Frobenius integral $e$ then $G$ is 

1593: also a Frobenius Hopf algebroid with $e_G=\iota(e)$ a Frobenius integral.

1594: \end{thm}

1595: \begin{proof}

1596: The observation made before the 

1597: formulation of the Theorem, in particular equation (\ref{eq: forget along}) 

1598: implies the formulae for $s_G$ and $t_G$. 

1599: In order to obtain the expressions for $\cop_G$ and $\eps_G$ at once,

1600: and also to prove the Frobenius Hopf algebroid case, the next Proposition, 

1601: however simple, is very useful.

1602: \begin{pro} \label{pro: tensor Q}

1603: If $H$ is a right bialgebroid over $R$ and $Q$ is a left pre-BCA over $H$ then

1604: the functor $\under\oR Q:\M_H\to(\M_H)_Q$ is strong monoidal.

1605: \end{pro}

1606: \begin{proof}

1607: The natural transformation

1608: \begin{align*}

1609: (Y\oR Q)\oQ(Y'\oR Q)&\to (Y\oR Y')\oR Q\\

1610: (y\oR q)\oQ(y'\oR q')&\mapsto(y\oR y'\ract q\coa{-1})\oR q\coa{0}q'

1611: \end{align*}

1612: has inverse $(y\oR y')\oR q\mapsto (y\oR 1)\oQ(y'\oR q)$. The $H\mash Q$-module 

1613: map

1614: \[

1615: Q\to R\oR Q\,,\qquad q\mapsto e\oR q

1616: \]

1617: is the unit part of the monoidal structure and is obviously invertible.

1618: \end{proof}

1619: Continuing the proof of the Theorem we take the comonoid $\bra 

1620: H,\cop_H,\eps_H\ket$ in $\M_H$ and apply the strong monoidal functor $\under\oR 

1621: Q$. It is easy to check that the result is precisely $\bra G,\cop_G,\eps_G\ket$ 

1622: which is then necessarily a comonoid in $\M_G$. This comonoid is obviously 

1623: strong \cite{Sz: MonMor} proving that $\bra G,Q,s_G,t_G,\cop_G,\eps_G\ket $ is a 

1624: bialgebroid. It is straightforward to verify that the pair $\bra \iota,\eta\ket$ 

1625: satisfies the four axioms \cite{Sz: Strasbourg,Sz: MonMor} for a bialgebroid map 

1626: $H\to G$.

1627: %Now consider the case of Hopf algebroids $H$. In order for the right 

1628: %bialgebroid $G$ to be a Hopf algebroid with antipode $S_G$ the following axioms 

1629: %should be satisfied:

1630: %\begin{align}

1631: %S_G\circ t_G&=s_G\\

1632: %S_G(g\twoR)\oneR\oQ g\oneR\,S_G(g\twoR)\twoR&=S_G(g)\oQ i_G\label{eq: anti 2}\\

1633: %g\twoR\,S_G^{-1}(g\oneR)\oneR\oQ S_G^{-1}(g\oneR)\twoR&=i_G\oQ S_G^{-1}(g)

1634: %\label{eq: anti 3}

1635: %\end{align}

1636: %for all $g\in G$ where $i_G=i\mash 1$ is the unit element of $G$.

1637: %Putting $g=h\mash q$ we can calculate the LHS of (\ref{eq: anti 3} as follows:

1638: %\begin{align*}

1639: 

1640: 

1641: %\end{align*}

1642: %which is the RHS. 

1643: 

1644: If $H$ is a Frobenius Hopf algebroid then it has a distributive double algebra 

1645: structure \cite{Sz: DA}. Therefore we may assume that $H$ is the horizontal Hopf 

1646: algebroid of $\bra A,\ci,e,\star,i\ket$. Then $\bra 

1647: H,\cop_R,\Fi_R,\ci,R\hookrightarrow H\ket$ is a Frobenius algebra in $\M_H$, so 

1648: it is mapped by the strong monoidal functor of Proposition \ref{pro: tensor Q} 

1649: to a Frobenius algebra in $\M_G$. The comonoid part of this Frobenius algebra 

1650: has already been determined to be $\bra G,\cop_G,\eps_G\ket$. The monoid part 

1651: will provide a convolution product with unit on $G$ which, together with the 

1652: smash product algebra structure, will make $G$ a distributive double algebra. 

1653: This convolution product (vertical multiplication) is obtained as the composite

1654: \[

1655: (h\mash q)\oQ(h'\mash q')\mapsto (h\oR h'\star q\coa{-1})\oR q\coa{0}q'

1656: \mapsto h\ci(h'\star q\coa{-1})\mash q\coa{0} q'

1657: \]

1658: and its unit element $e_G$ is the image of $1\in Q$ under the map

1659: \[

1660: Q\iso R\oR Q\to H\mash Q \,.

1661: \]

1662: So $e_G=e\mash 1$ is a two-sided Frobenius integral in $G$.

1663: \end{proof}

1664:  

1665: 

1666: 

1667: \begin{rmk}

1668: The construction of a vertical multiplication on $H\mash Q$ suggests the new 

1669: interpretation of the smash product as a double algebraic one. If $\bra 

1670: A,\ci,e,\star,i\ket$ is a DDA and $Q$ is a BCA over the bialgebroid 

1671: $H$ over $R$ then there is a smash product double algebra $A\mash Q$ with

1672: \begin{itemize}

1673: \item underlying $k$-module $A\oR Q$,

1674: \item horizontal multiplication

1675: $

1676: (a\mash q)\star(a'\mash q')=a\star {a'}\oneR\mash(q\ract {a'}\twoR)q'\,,

1677: $

1678: \item horizontal unit $i\mash 1$,

1679: \item vertical multiplication

1680: $

1681: (a\mash q)\ci(a'\mash q')=a\ci(a'\star q\coa{-1})\mash q\coa{0}q'\,,

1682: $

1683: \item and vertical unit $e\mash 1$.

1684: \end{itemize}

1685: \end{rmk}

1686: 

1687: As a biproduct of the double algebraic picture we obtain the following result.

1688: \begin{pro} \label{pro: YD braided}

1689: For Frobenius Hopf algebroids $H$ the prebraiding of the left weak center 

1690: $\lZ(\M_H)$ is a braiding. Therefore $\lZ(\M_H)=\Z(\M_H)=\rZ(\M_H)$ and every 

1691: pre-BCA is a BCA over $H$.

1692: \end{pro}

1693: \begin{proof}

1694: We claim that the inverse braiding encoded in the right coaction $\bar\tau$ by 

1695: (\ref{eq: bartheta from bartau}) is given by

1696: \begin{equation} \label{eq: inverse tau}

1697: q\coa{0}\oR q\coa{1}=\eta_Q\Fi_R\Fi_T(x^j\star q\coa{-1})q\coa{0}\oR y^j\,.

1698: \end{equation}

1699: The proof is motivated by the double algebraic structure on $H\mash Q$

1700: given in the above Remark but we do not use that the given structure maps 

1701: satisfy the axioms of a DDA. Let us compute the would-be $\Fi_R$ of $H\mash Q$.

1702: It is

1703: \[

1704: \Phi_R(h\mash q):=(e\mash 1)\star(h\mash q)=e\mash \eta_Q\Fi_R(h)q\,.

1705: \]

1706: One conjectures $(x^j\mash 1)\oQ (y^j\mash 1)$ to be its dual basis.

1707: Instead of proving that we prove its special case

1708: \begin{align*}

1709: \Phi_R((i\mash q)\ci(x^j\mash 1))\ci (y^j\mash 1)

1710: &=(e\mash\eta_Q\Fi_R\Fi_T(x^j\star q\coa{-1})q\coa{0})\ci(y^j\mash 1)\\

1711: &=y^j\star(\Fi_R\Fi_T(x^j\star q\coa{-2})\ci q\coa{-1})\mash q\coa{0}\\

1712: &=y^j\star q\coa{-1}\star \Fi_B\Fi_R\Fi_T(x^j\star q\coa{-2})\mash q\coa{0}\\

1713: &=\Fi_R(i\ci\Fi_R(q\coa{-1})\oneR)\ci\Fi_R(q\coa{-1})\twoR\mash q\coa{0}\\

1714: &=i\ci \Fi_R(q\coa{-1})\mash q\coa{0}\\

1715: &=i\mash q\,.

1716: \end{align*}

1717: Comparing the first row with the Ansatz (\ref{eq: inverse tau}) and then using 

1718: the vertical multiplication of $H\mash Q$ we arrive at

1719: \begin{align*}

1720: i\mash q&=(e\mash q\coa{0})\ci(q\coa{1}\mash 1)\\

1721: &=q\coa{1}\star{q\coa{0}}\coa{-1}\mash{q\coa{0}}\coa{0}

1722: \end{align*}

1723: which is equation (\ref{eq: tau-bartau 2}). The verification of  

1724: (\ref{eq: tau-bartau 1}) is a bit longer,

1725: \begin{align*}

1726: {q\coa{0}}\coa{0}\oR q\coa{-1}\star{q\coa{0}}\coa{1}

1727: &=\eta_Q\Fi_R\Fi_T(x^j\star q\coa{-1})q\coa{0}\oR q\coa{-2}\star y^j\\

1728: &=\eta_Q\Fi_R\Fi_T(S^{-1}(q\coa{-2})\star x^j\star q\coa{-1})q\coa{0}\oR y^j\\

1729: &=\eta_Q\Fi_R\Fi_T(S^{-1}(x^k)\star x^j\star(y^k\ci q\coa{-1}))q\coa{0}\oR y^j\\

1730: &=\eta_Q\Fi_R\Fi_T((S^{-1}(y^k)\ci(S^{-1}(x^k)\star x^j))\star q\coa{-1})

1731: q\coa{0}\oR y^j\\

1732: &=\eta_Q\Fi_R\Fi_T((x_k\ci(y_k\star x^j))\star q\coa{-1})q\coa{0}\oR y^j\\

1733: &=\eta_Q\Fi_R\Fi_T(\Fi_R\Fi_T(x^j)\star q\coa{-1})q\coa{0}\oR y^j\\

1734: &=\eta_Q\Fi_R(\Fi_T(x^j)\star q\coa{-1})q\coa{0}\oR y^j\\

1735: &=\eta_Q\Fi_R(q\coa{-1})q\coa{0}\ract \Fi_T(x^j)\oR y^j\\

1736: &=q\oR\Fi_R\Fi_T(x^j)\ci y^j\\

1737: &=q\oR i\,.

1738: \end{align*}

1739: \end{proof}

1740: %Recall that for Hopf algebras Yetter's formula \cite{Yetter} for the inverse 

1741: %braiding involves the symmetry of the category $\Vec_k$. Therefore it cannot 

1742: %have a direct generalization to Hopf algebroids for which the underlying 

1743: %category $_R\M_R$ is not even braidable. Formula (\ref{eq: inverse tau}) on the 

1744: %other hand contains the Frobenius integral $\Fi_T$ on $H$ and its dual  

1745: %integral $e$ in $H$ in the form of $\cop_R(e)=x^j\oR y^j$. Thus 

1746: 

1747: Extensions of quantum groupoids by BCA's are transitive in the following 

1748: sense.

1749: \begin{pro}

1750: If $Q$ is a BCA over $H$ and $P$ is a BCA over $H\mash

1751: Q$ then $P$ is a BCA over $H$, too. Furthermore, 

1752: $ (H\mashed{R} Q)\mashed{Q} P \cong H\mashed{R} P$.

1753: \end{pro}

1754: \begin{proof} (Sketch)

1755: The notation $P$ comprises the data $\bra\bra P,\theta^P\ket,\mu_P,\eta_P\ket$ 

1756: of a commutative monoid in $\lZ(\M_{H\mash Q})$. We use analogous notations for 

1757: $Q$. In order to obtain a monoid $\bra\bra P,\theta\ket,\mu,\eta\ket$ in 

1758: $\lZ(\M_H)$ we define

1759: \begin{align}

1760: \theta_X&:

1761: \begin{CD}

1762: P\oR X\iso P\oQ Q\oR X@>P\o\theta^Q_X>>P\oQ(X\oR Q)@>\theta^P_{X\o Q}>>

1763: X\oR Q\oQ P\iso X\oR P

1764: \end{CD}\\

1765: \mu&:P\oR P\to P\oQ P\rarr{\mu_P}P\\

1766: \eta&:R\rarr{\eta_Q}Q\rarr{\eta_P}P

1767: \end{align}

1768: For the $\theta_X$ to be a well-defined arrow in $\M_H$ notice that $\theta^Q_X$ 

1769: belongs to $_Q\M_H$ due to that $\mu_Q$ satisfies (\ref{eq: arrows in lZ}) and 

1770: is braided commutative.The pair $\bra P,\theta\ket$ will be established as an 

1771: object in the left weak center if we can show that (\ref{eq: theta}) is 

1772: satisfied. This is rather weary but straightforward utilizing that (\ref{eq: 

1773: theta}) is satisfied by both $\theta^Q$ and $\theta^P$. It is clear that the 

1774: triple $\bra P,\mu,\eta\ket$ is a monoid in $\M_H$, but we need to show that 

1775: $\mu$ and $\eta$ belong to $\lZ(\M_H)$. This is equivalent to checking  

1776: (\ref{eq: arrows in lZ}), i.e., the equations

1777: \begin{align}

1778: (X\oR\mu)\circ(\theta_X\oR P)\circ(P\oR \theta_X)&=\theta_X\circ(\mu\oR X)\\

1779: (X\oR \eta)\circ\theta^R_X&=\theta_X\circ(\eta\oR X)\,.

1780: \end{align}

1781: 

1782: 

1783: 

1784: \end{proof}

1785: 

1786: 

1787: 

1788: 

1789: A sort of converse to the previous proposition is the next proposition

1790: which we state without proof.

1791: \begin{pro}

1792: If $\eta:Q\to P$ is a monoid morphism in $\Z(\M_H)$ between commutative 

1793: monoids (i.e., BCA's over $H$) then there is a unique BCA structure on 

1794: $P$ over $H \mash Q$ the unit of which is $\eta$. 

1795: \end{pro}

1796: %\begin{proof}

1797: %The algebra map $\eta$ uniquely determines a monoid $\bra P,\mu,\eta\ket$ in 

1798: %$_Q(\M_H)_Q$. In order for $_QP_Q$ to be diagonal, i.e., 

1799: 

1800: 

1801: 

1802: 

1803: 

1804: 

1805: %\end{proof}

1806: 

1807: 

1808: 

1809: 

1810: \begin{pro} \label{pro: E as smash}

1811: Let $N\subset M$ be a Galois extension over the Frobenius Hopf algebroid $H$. 

1812: Then the restriction of the Galois map $\Gamma_M$ provides an isomorphism of 

1813: Hopf algebroids $H\mash C\cong E$ where $E$ is the endomorphism Hopf algebroid 

1814: of the extension.

1815: \end{pro}

1816: \begin{proof}

1817: The structure maps (\ref{s_G}), (\ref{t_G}), (\ref{cop_G}) and (\ref{eps_G}) of 

1818: the smash product are mapped by $\Gamma_M$ to 

1819: \begin{alignat}{2}

1820: s_E&:C\to E&\qquad c&\mapsto \{m\mapsto mc\}\\

1821: t_E&:C^\op\to E&\qquad c&\mapsto \{m\mapsto cm\}\\

1822: \cop_E&:E\to E\oC E\quad &\text{such that } 

1823: &\alpha\oneR(m)\alpha\twoR(m')=\alpha(mm')\\

1824: \eps_E&:E\to C&\qquad \alpha&\mapsto \alpha(1)

1825: \end{alignat}

1826: respectively, where note that multiplicativity of $\cop_E$ uniquely fixes it by 

1827: Lemma \ref{lem: isomorphisms}, (\ref{eq: fork rule}).

1828: Now it is easy to check that $\Gamma_M:H\mash C\to E$ satisfies the axioms 

1829: of bialgebroid maps.

1830: \end{proof}

1831: 

1832: 

1833: \section{Contravariant fiber functors} \label{s: fiber}

1834: 

1835: In this section we study functors from the module category of a Hopf algebroid 

1836: $A$ that correspond to $A$-Galois extensions of a given algebra $N$.

1837: In this sense we study generalizations of Ulbrich's Theorem \cite{Ulbrich} 

1838: relating Hopf-Galois extensions to fiber functors.

1839: Technically speaking, however, the functors we study here are very different 

1840: from the usual fiber functors. They are contravariant hom-functors 

1841: $\Hom_H(\under,M)$ from $\M_H$ to $_N\M_N$. So they are colimit preserving but 

1842: rarely faithful and exact. Still they have some properties that are worthy of

1843: discussion. As a preparation we prove 

1844: \begin{lem} \label{lem: fgp mon}

1845: For $H$ a Frobenius Hopf algebroid the full subcategory $\M_H^\fgp$ of $\M_H$ 

1846: the objects of which are the finitely generated projective $H$-modules is a 

1847: monoidal subcategory.

1848: \end{lem}

1849: \begin{proof}

1850: It suffices to show that $H\oR H$, the tensor square of the regular object in 

1851: $\M_H$ is a fgp module. This in turn follows from the existence of the 

1852: isomorphism \cite[(4.1)]{Sz: DA}

1853: \[

1854: \Gamma_{RB}:H\oR H\iso A\oB H\,,\qquad a\oR a'\mapsto a\oneB\oB a\twoB\ci a'\,.

1855: \]

1856: which happens to be an $H$-module map,

1857: \begin{align*}

1858: \Gamma_{RB}(a\ract h\oneR\oR a'\ract {h'}\twoR)

1859: &=a\star h\oneR\star u_k\oB v_k\ci(a'\star h\twoR)\\

1860: &=a\star u_k\oB (v_k\ci a')\star h

1861: \end{align*}

1862: thanks to right distributivity in $A$. Since $A_B$ is fgp, the statement is 

1863: proven.

1864: \end{proof} 

1865: 

1866: 

1867: 

1868: \begin{thm} \label{thm: U}

1869: Let $N$ be an algebra and $A$ a distributive double algebra. As usual, $H$ 

1870: denotes the horizontal Hopf algebroid of $A$.

1871: \begin{enumerate}

1872: \item The mappings

1873: \[

1874: M\mapsto F= \Hom_H(\under,M)\quad\text{respectively}\quad F\mapsto M=F(H_H)

1875: \]

1876: provide mutually inverse category equivalences between the following two 

1877: categories.

1878: \begin{itemize}

1879: \item The category of $H$-module algebras $M$ equipped with an algebra map 

1880: $\hat\eta:N\to  M^H \subset M$. The arrows 

1881: from $\bra M,\hat\eta\ket$ to $\bra M',\hat\eta'\ket$ are the $H$-module algebra 

1882: maps $\alpha:M\to M'$ for which $\alpha\circ\hat\eta=\hat\eta'$.

1883: \item The category of opmonoidal functors 

1884: $F:\M_H^\fgp\to\,_N\M_N^\op$ as objects and 

1885: monoidal natural transformations as arrows.

1886: \end{itemize}

1887: \item $M$ is an $A$-extension of $N$ iff $F$ is normal opmonoidal.

1888: \item $M$ is an $A$-Galois extension of $N$ iff $F$ is strong (op)monoidal.

1889: \item The full subcategories $A$-$\Gal(N)$ and $\F(\M_H^\fgp,\,_N\M_N^\op)$ of 

1890: those in (1) selected by the conditions of (3), respectively, are groupoids.

1891: \end{enumerate}

1892: In this way (1) and (3) establish a category equivalence $A$-$\Gal(N)\sim 

1893: \F(\M_H^\fgp,\,_N\M_N^\op)$ between $A$-Galois extensions of $N$ and strong 

1894: monoidal  functors $\M_H^\fgp\to \,_N\M_N^\op$.

1895: \end{thm}

1896: \begin{proof} 

1897: The construction of the functor $M\mapsto F$ goes as follows. Given $\bra 

1898: M,\hat\eta\ket$ the $M$ is an $N^e$-$H$-bimodule so $\Hom_H(\under,M)$ is a 

1899: contravariant functor from $H$-modules to $_N\M_N$.The monoid structure $\bra 

1900: M,\mu,\eta\ket$ defines an opmonoidal structure on this functor\footnote{The 

1901: arrows between $N$-$N$-bimodules are always considered in $_N\M_N$ and never in 

1902: $_N\M_N^\op$.} 

1903: \begin{align*}

1904: \Hom_H(Y,M)\oN\Hom_H(Y',M)&\to\Hom_H(Y\oR Y',M)\quad 

1905: \xi\oN\xi'\mapsto\mu\circ(\xi\oR\xi')\\

1906: N&\to\Hom_H(R,M)\qquad n\mapsto\{r\mapsto n\eta(r)\}

1907: \end{align*}

1908: An arrow $\alpha$ is mapped to the monoidal natural transformation 

1909: $\Hom_H(\under,\alpha)$. 

1910: 

1911: Now we construct the functor $F\mapsto M$.

1912: Any opmonoidal functor $\bra F,F^2,F^0\ket:\M_H^\fgp\to\,_N\M_N^\op$

1913: maps comonoids to comonoids. Therefore it maps $\bra H_H,\cop_R,\Fi_R\ket$ to a 

1914: monoid $M:=F(H_H)$ in $_N\M_N$. The unit of this monoid is the 

1915: composite

1916: \[

1917: \hat\eta:=\quad 

1918: \begin{CD}

1919: N@>F^0>> FR@>F(\Fi_R)>> M

1920: \end{CD}

1921: \]

1922: which becomes a $k$-algebra map by prolongation of the multiplication 

1923: \[

1924: \hat\mu:=\quad

1925: \begin{CD}

1926: M\oN M@>F^{H,H}>>F(H\oR H)@>F(\cop_R)>> M

1927: \end{CD}

1928: \]

1929: to a $k$-algebra multiplication. The $M$ also inherits a right $H$-module 

1930: structure from left multiplication $\lambda_h=h\star\under$ via 

1931: $H\rarr{\lambda}\End(H_H)\rarr{F}\End(\,_NM_N)$. We have

1932: \begin{align*}

1933: F\lambda_h\circ\hat\eta&=F(\Fi_R\circ\lambda_h)\circ 

1934: F^0=F(\Fi_R\circ\lambda_{\Fi_T\Fi_R(h)})\circ F^0\\

1935: &=F\lambda_{\Fi_T\Fi_R(h)}\circ\hat\eta

1936: \end{align*}

1937: implying that $\hat\eta$ factors uniquely through the inclusion 

1938: $ M^H \subset M$. 

1939: For each $h$ the action $F\lambda_h$ is an $N$-$N$ bimodule 

1940: map which makes $M$ an $N^e$-$H$-bimodule. By means of the isomorphism 

1941: $\iota:m\mapsto \{h\mapsto m\ract h\}$ the monoid 

1942: $\bra\,_NM_N,\hat\mu,\hat\eta\ket$ becomes isomorphic to the convolution monoid 

1943: $\Hom_H(H,M)$ associated to an $H$-module algebra $\bra M,\mu,\eta\ket$ 

1944: structure on $M$. Of course, the monoid $\bra M,\mu,\eta\ket$ arises from the

1945: $k$-algebra structure of $M$ just as the monoid $\bra M,\hat\mu,\hat\eta\ket$ 

1946: does.

1947: \begin{equation} \label{dia: mul-mul}

1948: \begin{CD}

1949: FH\o FH@>\iota\o\iota>>\Hom_H(H,M)\o\Hom_H(H,M)\\

1950: @VVV @VVV\\

1951: FH\oN FH@>\iota\oN\iota>>\Hom_H(H,M)\oN\Hom_H(H,M)\\

1952: @V{F^{H,H}}VV @VV{\mu\circ(\under\oR\under)}V\\

1953: F(H\oR H)@.\Hom_H(H\oR H,M)\\

1954: @V{F\cop_R}VV @VV{\under\circ\cop_R}V\\

1955: FH@>\iota>>\Hom_H(H,M)

1956: \end{CD}

1957: \end{equation}

1958: This yields the object map of the functor $F\mapsto M$. As 

1959: for the arrow map take any monoidal natural transformation $\nu:F\to F'$ and 

1960: define $\alpha:=\nu_H:M\to M'$. Then by the multiplicativity constraint 

1961: for $F$ the $\alpha$ is an $H$-module algebra morphism and the unit constraint 

1962: implies that

1963: \[

1964: \begin{CD}

1965: N@>F^0>>FR@>F\Fi_R>>M\\

1966: @| @VV{\nu_R}V @VV{\alpha}V\\

1967: N@>{F'}^0>>F'R@>F'\Fi_R>>M'

1968: \end{CD}

1969: \]

1970: is commutative, i.e., $\alpha\circ\hat\eta=\hat\eta'$. 

1971: 

1972: Now we construct a natural isomorphism $\nu$ from the identity functor $F\mapsto 

1973: F$ to the composite $F\mapsto M\mapsto F$. Choosing a direct 

1974: summand diagram $Y\rarr{\pi_k}H\rarr{\sigma_k}Y$ for each fgp $H$-module $Y$

1975: the isomorphism $\iota:M\to \Hom_H(H,M)$ for $M=FH$ extends to a natural 

1976: isomorphism $\nu:F\to\Hom_H(\under,M)$ by

1977: \[

1978: \begin{CD}

1979: FY@>F\sigma_k>>FH@>F\pi_k>>FY\\

1980: @V{\nu_Y}VV @VV{\iota}V @VV{\nu_Y}V\\

1981: \Hom_H(Y,M)@>\under\circ\sigma_k>>\Hom_H(H,M)@>\under\circ\pi_k>>

1982: \Hom_H(Y,M)

1983: \end{CD}

1984: \]

1985: This natural isomorphism will then be 

1986: automatically monoidal due to the interplay between the multiplications 

1987: $\hat\mu$ and $\mu$ seen on the diagram (\ref{dia: mul-mul}) .

1988: 

1989: The natural isomorphism from the identity functor $M\mapsto M$ to the composite 

1990: $M\mapsto F\mapsto M$ is just $\iota:M\to\Hom_H(H,M)$ viewed as a map of monoids 

1991: in $_N\M_N$. So in particular $\iota\circ\hat\eta$ is equal to $F\Fi_R\circ F^0$ 

1992: for the opmonoidal functor $F=\Hom_H(\under,M)$.

1993: This completes the proof of the equivalence in (1).

1994: 

1995: By Lemma \ref{lem: R->M} the unique arrow 

1996: $N\to  M^H $ factorizing $\hat\eta$ is an isomorphism iff $F^0$ is an 

1997: isomorphism, i.e., iff $F$ is normal. This proves (2).

1998: 

1999: Strong (op)monoidality of $F$ is equivalent to invertibility of $F^0$ and 

2000: $F^{H,H}$. By the natural isomorphism $F\cong\Hom_H(H,M)$ the latter is

2001: equivalent to invertibility of the left vertical arrow in the next diagram.

2002: \[

2003: \begin{CD}

2004: \Hom_H(H,M)\oN \Hom_H(H,M)@<\sim<< M\oN M\\

2005: @V{\mu\circ(\under\oR\under)}VV @VV{\gamma^M}V\\

2006: \Hom_H(H\oR H,M)@>\sim>>M\oT A

2007: \end{CD}

2008: \]

2009: where the lower horizontal arrow is given by a composition of isomorphisms

2010: \[

2011: \begin{CD}

2012: \Hom_H(H\oR H,M)@>\sim>>\Hom_H(H,M\oR H)@>\sim>> M\oR H

2013: @>M\o S^{-1}>> M\oT V

2014: \end{CD}

2015: \]

2016: performing the mappings

2017: \[

2018: \chi\mapsto \chi(\under\oR x^j)\oR y^j\mapsto \chi(i\oR x^j)\oR y^j\mapsto 

2019: \chi(i\oR u^j)\oT v^j\,.

2020: \]

2021: Commutativity of the diagram now follows from the simple calculation

2022: \[

2023: (m\ract i)(m'\ract u^k)\oT v^k=m{m'}\zeroT\oT {m'}\oneT=\gamma^M(m\oN m')

2024: \]

2025: Therefore $\gamma^M$ is invertible iff $F^{H,H}$ is invertible. Adding the 

2026: condition that $N\subset M$ is an $A$-extension we obtain (3).  

2027: 

2028: Since $H_H$ is a Frobenius algebra, it is a selfdual object in $\M_H$. Therefore 

2029: any monoidal natural transformation between strong monoidal functors from 

2030: $\M_H^\fgp$ is invertible at $H_H$ \cite{Saavedra-Rivano} and therefore it is 

2031: invertible everywhere. This proves (4).

2032: \end{proof}

2033: 

2034: \begin{cor}

2035: The map $M\mapsto \Hom_H(\under,M)$ is a category equivalence 

2036: between the category $A$-$\Gal(N)$ of $A$-Galois extensions of $N$ and

2037: the category $\F(\M_H,\,_N\M_N^\op)$ of colimit preserving opmonoidal functors

2038: the restrictions of which to $\M_H^\fgp$ is strong (op)monoidal.

2039: \end{cor}

2040: \begin{proof}

2041: If $F:\M_H\to \,_N\M_N^\op$ is colimit preserving then the corresponding 

2042: $F^\op:\M_H^\op\to\,_N\M_N$ is limit preserving and $H_H$ is a cogenerator for 

2043: $\M_H^\op$. The conditions for the special adjoint functor theorem 

2044: \cite{MacLane} hold, so $F^\op$ has a left adjoint. It follows that $F^\op$ is a 

2045: hom-functor, $F\cong\Hom_{\M_H^\op}(M\under)$, i.e., $F\cong\Hom_H(\under,M)$.

2046: Now Theorem \ref{thm: U} implies that $F$ has a strong restriction to the fgp 

2047: modules precisely when $N\subset M$ is $A$-Galois. Vice versa, every Galois 

2048: extension $M$ gives rise to a colimit preserving opmonoidal functor 

2049: $\Hom_H(\under, M)$ the resriction of which to $\M_H^\fgp$ is strong.

2050: \end{proof}

2051: 

2052: As an application of the strong monoidal functor $\Hom(\under,M)$ we present 

2053: here another characterization of Galois extensions over DDA's.

2054: In order to understand the terminology "left distributivity"

2055: let us look at multiplication of $M$ as a vertical one and the right $H$-action 

2056: $\ract$ as a partially defined horizontal multiplication between $M$ and $H$.

2057: \begin{pro}

2058: Let $A$ be a DDA and $M$ be a right $A$-module algebra with $N=M^H$.

2059: Then $N\subset M$ is $A$-Galois if and only if $\psi=\under\ract e:M\to N$ is a 

2060: Frobenius homomorphism and the "left distributivity" rule

2061: \[

2062: m\ract (a\ci a')=(m\oneL \ract a)(m\twoL\ract a')

2063: \]

2064: holds for all $m\in M$ and $a,a'\in A$. Here $m\oneL\oN m\twoL$ is the coproduct

2065: associated to the Frobenius structure on $N\subset M$ defined by $\psi$.

2066: \end{pro} 

2067: Note that "right distributivity" $(mm')\ract a=(m\ract a\oneR)(m'\ract a\twoR)$ 

2068: holds for all right module algebras. Note also that $\ci$ for $H$ plays the role

2069: of convolution product while the ordinary product is $\star$.

2070: \begin{proof}

2071: \textit{Necessity:} Consider the contravariant functor $\Hom_H(\under,M):\M_H\to 

2072: \,_N\M_N$. It is strong monoidal, so maps monoids to comonoids, comonoids to 

2073: monoids, and Frobenius algebras to Frobenius algebras. Therefore it maps 

2074: $\bra A,\cop_R,\Fi_R,\mu_V,R\hookrightarrow A\ket$ to some

2075: Frobenius algebra structure on $\Hom_H(A,M)\cong M\in\,_N\M_N$. Since a 

2076: Frobenius algebra structure in $_N\M_N$ is uniquely determined by the algebra 

2077: structure and by the Frobenius homomorphism, the counit, it is sufficient to 

2078: check that the image of $\bra A,\cop_R,\Fi_R\ket$ is the convolution algebra 

2079: $\Hom_H(A,M)$ and the image of the unit $R\hookrightarrow A$ is $\psi$. 

2080: Then the coproduct must have the form

2081: \[

2082: \cop_M(m)\equiv m\oneL\oN m\twoL=\sum_i me_i\oN f_i

2083: \]

2084: where $\sum_i e_i\oN f_i$ is the dual basis of $\psi$. This means that the

2085: composite

2086: \begin{equation}

2087: \begin{CD}

2088: \Hom_H(A,M)@>\Hom(\mu_V,M)>>\Hom_H(H\oR H,M)\\

2089: @. @V{\wr}V{[\mu\circ(\under\oR\under)]^{-1}}V\\

2090: @.\Hom_H(A,M)\oN\Hom_H(A,M)

2091: \end{CD}

2092: \end{equation}

2093: must be the map

2094: \[

2095: (m\ract\under)\mapsto(m\oneL\ract\under)\oN(m\twoL\ract\under)

2096: \]

2097: Applying $\mu\circ(\under\oR\under)$ we obtain left distributivity.

2098: 

2099: \textit{Sufficiency:} Consider the map $M\oT A\to M\oN M$ defined by $m\oT 

2100: a\mapsto m\oneL\oN m\twoL\ract a$. Then

2101: \begin{align*}

2102: \gamma^M(m\oneL\oN m\twoL\ract a)&=m\oneL (m\twoL\ract a)\zeroT\oT 

2103: (m\twoL\ract a)\oneT\\

2104: &=(m\oneL\ract i) (m\twoL\ract u^k)\oT v^k\star a=(m\ract (i\ci u^k))\oT 

2105: v^k\star a\\

2106: &=m\oT \Fi_T(u^k)\star v^k\star a=m\oT a

2107: \end{align*}

2108: proves that $\gamma^M$ is epi.

2109: \end{proof}

2110: 

2111: 

2112: \section{A monoidal duality}

2113: 

2114: Given a right bialgebroid $H$ over $R$, an $H$-module algebra $M$ and an 

2115: algebra map $N\to M^H $  we can 

2116: look for a duality between - full subcategories of - $\M_H$ and $_N\M_N$ in the 

2117: following form. The $M$ being an $N^e$-$H$-bimodule, it determines two functors

2118: \begin{alignat}{2}

2119: J&:\,_N\M_N^\op&\to\M_H\qquad X&\mapsto\Hom_{N^e}(X,M)\\

2120: K&:\M_H&\to \,_N\M_N^\op\qquad Y&\mapsto\Hom_H(Y,M),

2121: \end{alignat}

2122: the $M$-dual functors,

2123: that are in adjunction $K\adj J$. The counit and unit of the adjunction  

2124: are just the natural homomorphism to the double dual,

2125: \begin{align*}

2126: \sigma_X&:X\to\Hom_H(\Hom_{N^e}(X,M),M)\qquad\in \,_N\M_N\\

2127: \sigma_Y&:Y\to\Hom_{N^e}(\Hom_H(Y,M),M)\qquad\in\ \ \M_H 

2128: \end{align*}

2129: By definition they are isomorpisms precisely for the $M$-reflexive modules 

2130: \cite{A-F}. Either one of the $M$-dual functors map relexive modules to 

2131: reflexive ones, so the restriction of $J$ and $K$ to the $M$-reflexive modules 

2132: provides an adjoint equivalence 

2133: \begin{equation} \label{eq: adj equiv}

2134: \M_H^{\Mref}\sim(\,_N\M_N^\Mref)^\op\,,

2135: \end{equation}

2136: that is to say, a duality between the reflexive modules themselves.

2137: 

2138: Since $M$ has monoid structures both in $\M_H$ and $_N\M_N$, the functor $J$ is 

2139: monoidal and $K$ is opmonoidal, 

2140: \begin{align*}

2141: \left.

2142: \begin{alignedat}{2}

2143: J_{X,X'}&:JX\oR JX'\to J(X\oN X')&\qquad (\xi\oR \xi')&\mapsto 

2144: \hat\mu\circ(\xi\oN\xi')\\

2145: J_0&:R\to JN&\qquad r&\mapsto\{n\mapsto \hat\eta(n)\ract r\}\\

2146: \end{alignedat}

2147: \right\}&\in \M_H\\

2148: \left.

2149: \begin{alignedat}{2}

2150: K^{Y,Y'}&:KY\oN KY'\to K(Y\oR Y')&\qquad (\beta\oN\beta')&\mapsto

2151: \mu\circ(\beta\oR\beta')\\

2152: K^0&:N\to KR&\qquad n&\mapsto\{r\mapsto n\cdot\eta(r)\}

2153: \end{alignedat}

2154: \right\}&\in \,_N\M_N

2155: \end{align*}

2156: They are mates under the given adjunction $K\dashv J$, that is to say,

2157: \begin{align}

2158: K^{JX,JX'}\circ(\sigma_X\oN\sigma_{X'})&=KJ_{X,X'}\circ\sigma_{X\oN X'}\\

2159: K^0&=KJ_0\circ\sigma_N\\

2160: J_{KY,KY'}\circ(\sigma_Y\oR\sigma_{Y'})&=JK^{Y,Y'}\circ\sigma_{Y\oR Y'}\\

2161: J_0&=JK^0\circ\sigma_R

2162: \end{align}

2163: These equations are simple consequences of the fact that the two monoid 

2164: structures on $M$ come from the same $k$-algebra structure,

2165: \[

2166: \begin{CD}

2167: M\o M@>>>M\oR M\\

2168: @VVV @VV{\mu}V\\

2169: M\oN M@>\hat\mu>>M

2170: \end{CD}

2171: \qquad 

2172: \begin{CD}

2173: k@>>> R\\

2174: @VVV @VV{\eta}V\\

2175: N@>\hat\eta>>M

2176: \end{CD}

2177: \]

2178: We have, as in Theorem \ref{thm: U} (2), that $K$ is normal iff the map $N\to 

2179:  M^H $ is an isomorphism and $J$ is normal iff the map $R\to M^N$ is 

2180: an isomorphism.

2181: 

2182: In order to find monoidal subcategories in $\M_H$ and $_N\M_N$ that become 

2183: monoidally dual under (\ref{eq: adj equiv}) we have to make further assumptions. 

2184: Assume that the right bialgebroid $H$ is that of the horizontal Hopf algebroid 

2185: of a distributive double algebra $A$ and assume that $N\subset M$ is $A$-Galois. 

2186: We know from Lemma \ref{lem: fgp mon} and Theorem \ref{thm: U} that $\M_H^\fgp$ 

2187: is a full  monoidal subcategory and the restriction of $K$ to this subcategory 

2188: is strong opmonoidal. Therefore the restriction of $K$ will provide a monoidal 

2189: equivalence iff all the fgp $H$-modules are $M$-reflexive. Since the class of 

2190: reflexive modules is closed under taking direct summands and finite direct sums, 

2191: this happens precisely when the regular object $H_H$ is $M$-reflexive. 

2192: \[

2193: \sigma_H:H\to\Hom_{N^e}(\Hom_H(H,M),M)\iso\End\,_NM_N

2194: \]

2195: being just the canonical embedding to the endomorphism Hopf algebroid $E$ we are 

2196: left with considering the case when $H$ is $E$ and acts canonically on $M$.

2197: 

2198: \begin{thm} \label{thm: qsym}

2199: Let $E$ be the endomorphism Hopf algebroid associated to the balanced 

2200: depth 2 Frobenius extension $N\subset M$. Then the functor

2201: \[

2202: \Hom_E(\under, M):\M_E^\fgp\to \,_N\M_N^\op

2203: \]

2204: provides a monoidal duality between the categories of all fgp $E$-modules and 

2205: those $N$-$N$-bimodules that are direct summands of  finite direct sums of 

2206: $M$'s.

2207: \end{thm}

2208: \begin{proof}

2209: $E_E$ is $M$-reflexive by construction. Thus $\M_E^\fgp$ is 

2210: a full subcategory of the category of reflexive modules and (\ref{eq: adj 

2211: equiv}) restricts to a category equivalence $F$. Since $\M_E^\fgp$ is generated 

2212: by direct sums and direct summands from $E_E$, the same holds for the 

2213: $N$-$N$-bimodules in the image of $F$ and for $F(E_E)\cong M$. The 

2214: extension $N\subset M$ is $E$-Galois therefore $\Hom_E(\under,M)$ is strong 

2215: monoidal on fgp modules. 

2216: \end{proof}

2217: 

2218: Depending on the applications the content of the theorem varies from 

2219: trivialities to nontrivial statements. For example, if $k\subset K$ is a 

2220: separable field extension - including the case of classical Galois field 

2221: extensions - then $E=\End K_k$ is the Hopf algebroid version of the weak Hopf 

2222: algebra constructed in \cite{Sz: Strasbourg} and its representation category is 

2223: trivial: The theorem reduces to the statement that the category of finite 

2224: dimensional $k$-vector spaces is selfdual. 

2225: 

2226: If $N$ is a strongly $G$-graded $k$-algebra for a finite group $G$ and 

2227: $N_e\subset N$ has centralizer $k\cdot 1$ then choosing 

2228: $M=N\mash (kG)^*$ we obtain that $E$ is the group algebra $kG$ acting

2229: canonically on the smash product $M$.

2230: 

2231: If $N$ is the observable algebra in rational quantum field theory and $M$ is 

2232: the algebra of charge creating fields then $_NM$ is freely generated by 

2233: finitely many fields $f_q^i\in M$ each of them implementing a localized 

2234: endomorphism $\rho_q$ of $N$, i.e., $f^i_qn=\rho_q(n)f_q^i$, $n\in N$, 

2235: $i=1,\dots, I_q$. The Doplicher-Haag-Roberts category $\DHR(N)$ is the full 

2236: subcategory of $\End N$ the objects of which are finite direct sums of 

2237: $\rho_q$'s and is a monoidal category by composition of endomorphisms.

2238: One has a contravariant monoidal equivalence between $\DHR(N)$ and the category 

2239: of $N$-$N$-bimodules that are finite direct sums of the bimodules $Nf_q^i$. 

2240: Hence Theorem \ref{thm: qsym} gives a monoidal equivalence $\M_E^\fgp\simeq 

2241: \DHR(N)$ and therefore the Hopf algebroid $E$ can be interpreted as the 

2242: global gauge symmetry of the superselection sectors.

2243: 

2244: 

2245: \begin{thebibliography}{XX}

2246: \begin{small}

2247: 

2248: \bibitem{A-F} F.W. Anderson, K.R. Fuller, \textit{Rings and Categories of 

2249: Modules}, 2nd ed., Springer-Verlag New York, Inc., 1992

2250: 

2251: \bibitem{B: Gal} G. B\"ohm, \textit{Galois theory for Hopf algebroids},

2252: \texttt{arXiv:math.RA/0409513}

2253: 

2254: \bibitem{B-Sz: Hgd} G. B\"ohm, K. Szlach\'anyi, \textit{Hopf algebroids with 

2255: bijective antipodes: axioms, integrals and duals}, J. Algebra \textbf{274} 

2256: (2004) 708-750

2257: 

2258: \bibitem{BSz: D2F} G. B\"ohm, K. Szlach\'anyi, \textit{Hopf Algebroid Symmetry 

2259: of Abstract Frobenius Extensions of Depth 2}, Commun. Algebra \textbf{32} (2004) 

2260: 4433-4464

2261: 

2262: \bibitem{Brz: corings} T. Brzezi\'nski, \textit{The structure of corings},

2263: Algebra Represent. Theory \textbf{5} (2002) 389-410

2264: 

2265: \bibitem{Brz-Mil} T. Brzezi\'nski, G. Militaru, \textit{Bialgebroids,

2266: $\times_A$-bialgebras and duality}, J. Algebra \textbf{251} (2002) 279-294 

2267: 

2268: \bibitem{Brz-Wis} T. Brzezi\'nski, R. Wisbauer, \textit{Corings and Comodules},

2269: London Math. Soc. LNS 309, Cambridge Univ. Press 2003

2270: 

2271: \bibitem{Caenepeel-DeGroot} S. Caenepeel, E. de Groot, \textit{Galois theory for 

2272: weak Hopf algebras}, \newline\texttt{arXiv:math.RA/0406186} 

2273: 

2274: \bibitem{Caenepeel-Wang-Yin} S. Caenepeel, D.-G. Wang, Y.-M. Yin, 

2275: \textit{Yetter-Drinfeld modules over weak Hopf algebras and the center

2276: construction}, \texttt{arXiv:math.QA/0409599}

2277: 

2278: \bibitem{CFM} M. Cohen, D. Fischman, S. Montgomery, \textit{Hopf Galois 

2279: extensions, smash products, and Morita equivalence}, J. Algebra \textbf{133}

2280: (1990) 351-372

2281: 

2282: \bibitem{Doi-Takeuchi} Y. Doi, M. Takeuchi, \textit{Hopf-Galois extensions of 

2283: algebras, the Miyashita-Ulbrich action, and Azumaya algebras},

2284: J. Algebra \textbf{121} (1989) 488-516

2285: 

2286: \bibitem{Greither-Pareigis} C. Greither, B. Pareigis, \textit{Hopf Galois theory 

2287: for separable field extensions}, J. Algebra \textbf{106} (1987) 239-258

2288: 

2289: \bibitem{Kadison} L. Kadison, \textit{Normal Hopf subalgebras, depth two and 

2290: Galois extensions}, \newline\texttt{arXiv:math.QA/0411129} 

2291: 

2292: \bibitem{K-Sz} L. Kadison, K. Szlach\'anyi, \textit{Bialgebroid actions on depth

2293: two extensions and duality}, Advances in Mathematics \textbf{179} (2003) 75-121 

2294: 

2295: \bibitem{Kreimer-Takeuchi} H. F. Kreimer, M. Takeuchi, \textit{Hopf algebras and 

2296: Galois extensions of an algebra}, Indiana Univ. Math. J. \textbf{30} (1981)

2297: 675-692

2298: 

2299: \bibitem{MacLane} S. Mac Lane, \textit{Categories for the Working

2300: Mathematician}, 2nd edition, GTM 5, Springer-Verlag New-York Inc., 1998

2301: 

2302: \bibitem{Montgomery} S. Montgomery,

2303: \textit{Hopf Algebras and Their Actions on Rings},  CBMS Regional Conf.

2304: Series in Math. Vol. 82, AMS, Providence, 1993

2305: 

2306: \bibitem{N-V} D. Nikshych, L. Vainerman, \textit{A Galois correspondence for 

2307: actions of quantum  groupoids on II$_1$-factors}, J. Func. Analysis \textbf{178} 

2308: (2000) 113-142 

2309: 

2310: \bibitem{Pareigis: forms} B. Pareigis, \textit{Forms of Hopf algebras and Galois 

2311: theory}, Topics in Algebra, Banach Center Publications, Vol.  26, Part 1, pp 

2312: 75-93 (1990)

2313: 

2314: \bibitem{Saavedra-Rivano} N. Saavedra Rivano, \textit{Cat\'egories 

2315: Tannakiennes}, Lecture Notes in Mathematics 265, Springer-Verlag 

2316: Berlin-Heidelberg-New York 1972

2317: 

2318: \bibitem{Schauenburg: bnrsthb} P. Schauenburg, Bialgebras over noncommutative

2319: rings, and a structure theorem for Hopf bimodules, \textit{Applied Categorical

2320: Structures} \textbf{6} (1998) 193-222

2321: 

2322: \bibitem{Schauenburg: ddqg} P. Schauenburg, \textit{Duals and 

2323: doubles of quantum groupoids} in "New trends in Hopf algebra theory",

2324: Contemporary Mathematics {\bf 267},  p. 273, AMS 2000

2325: 

2326: \bibitem{Sz: Strasbourg} K. Szlach\'anyi, \textit{Galois actions by finite 

2327: quantum groupoids} in "Locally Compact Quantum Groups and Groupoids", 

2328: ed.: L. Vainerman  (IRMA Lectures in Mathematics and 

2329: Theoretical Physics 2) de Gruyter 2003

2330: 

2331: \bibitem{Sz: DA} K. Szlach\'anyi, \textit{The double algebraic view of finite 

2332: quantum groupoids}, Journal of Algebra \textbf{280} (2004) 249-294

2333: 

2334: \bibitem{Sz: MonMor} K. Szlach\'anyi, \textit{Monoidal Morita equivalence},

2335: \texttt{arXiv:math.QA/0410407}

2336: 

2337: \bibitem{Ulbrich} K.-H. Ulbrich, \textit{Galois extensions as functors of 

2338: comodules}, Manucripta Math. \textbf{59} (1987) 391-397

2339: \end{small}

2340: \end{thebibliography}

2341: 

2342: 

2343: 

2344: 

2345: 

2346: \end{document}

2347: 

2348: 

2349: 

2350: