COMPARING COMMUTATIVE AND ASSOCIATIVE UNBOUNDED DIFFERENTIAL GRADED ALGEBRAS OVER Q FROM HOMOTOPICAL POINT OF VIEW 4 1 0 ILIASAMRANI 2 n Abstract. In this paper we establish a faithfulness result, in a homotopical a sense,between asubcategory ofthemodelcategory ofaugmented differential J graded commutative algebras CDGA and a subcategory of the model cate- 8 goryofaugmented differential gradedalgebrasDGAover thefieldofrational 2 numbersQ. ] T A Introduction . h t It is well known that the forgetful functor from the category of commutative a k-algebras to the category of category of associative k-algebras is fully faithful. m We havean analogueresult betweenthe categoryof unbounded differential graded [ commutativek-algebrasdgCAlg andthecategoryofunboundeddifferentialgraded k 1 associative algebras dgAlgk. The question that we want explore is the following: v Supposethatk =Q,is ittruethatforgetfulfunctor U :dgCAlg →dgAlg induces k k 5 a fully faithful functor at the level of homotopy categories 8 2 RU :Ho(dgCAlgk)→Ho(dgAlgk). 7 The answer is no. A nice and easy counterexample was given by Lurie. He has . 1 considered k[x,y] the free commutative CDGA in two variables concentrated in 0 degree 0. It follows obviously that 4 1 Ho(dgCAlg )(k[x,y],S)≃H0(S)⊕H0(S), k : v while Xi Ho(dgAlgk)(k[x,y],S)≃H0(S)⊕H0(S)⊕H−1(S). r Something nice happens if we consider the category of augmented CDGA denoted a by dgCAlg∗ and augmented DGA denoted by dgAlg∗. k k Theorem 0.1(3.1). For anyR andS indgCAlg∗, theinducedmapby theforgetful k functor ΩMapdgCAlg∗(R,S)→ΩMapdgAlg∗(R,S), k k has a retract, in particular πiMapdgCAlg∗(R,S)→πiMapdgAlg∗(R,S) k k is injective ∀ i>0. 2000 Mathematics Subject Classification. Primary55,Secondary14,16,18. Key words and phrases. DGA, CDGA, Mapping Space, Noncommutative and Commutative DerivedAlgebraicGeometry, RationalHomotopyTheory. Supported by the project CZ.1.07/2.3.00/20.0003 of the Operational Programme Education forCompetitiveness oftheMinistryofEducation, YouthandSportsoftheCzechRepublic. 1 2 Let S be a differential graded commutative algebra which is a ”loop” of an other CDGA algebra A, i.e. S = Holim(k → A ← k), where the homotopy limit is taken in the model category dgCAlg . A direct consequence of our theorem k is that the right derived functor RU is a faithful functor i.e., the induced map Ho(dgCAlg∗)(R,S)→Ho(dgAlg∗)(R,S) is injective. k k Interpretation of the result in the derived algebraic geometry. Rationally, anypointedtopologicalX spacecanbe viewedasanaugmented(connective)com- mutative differential graded algebra via its cochain complex C∗(X,Q). In case where X is a simply connected rational space, the cochain complex C∗(X,Q) car- ries the whole homotopical information about X, by Sullivan Theorem [5]. More- over,thebarconstructionBC∗(X,Q)isidentified(asE -DGA)toC∗(ΩX,Q)and ∞ ΩC∗(X,Q) is identified (as E -DGA) to C∗(ΣX,Q) cf. [4]. This interpretation ∞ allows us to make the following definition: A generalized rational pointed space is anaugmentedcommutativedifferentialgradedQ-algebra(possiblyunbounded). In thesamespirit,wedefineapointedgeneralizednoncommutative rational space asanaugmenteddifferentialgradedQ-algebra(possiblyunbounded). LetAbeany augmented CDGA resp. DGA, we will call a CDGA resp. DGA of the form ΩA a op-suspended CDGA resp. DGA. Our theorem 3.1, can be interpreted as follows: The homotopy category of op-suspendedgeneralized commutative ratio- nal spaces is a subcategory of the homotopy category of op-suspended generalized noncommutative rational spaces. 1. DGA, CDGA and E -DGA. ∞ Weworkinthesettingofunboundeddifferentialgradedk-modulesdgMod . This k is a a symmetric monoidal closed model category (k is a commutative ring). We denote the category of (reduced) operads in dgMod by Op . We follow notations k k and definitions of [2], we say that an operad P is admissible if the category of P−dgAlg admitsamodelstructurewherethefibrationsaredegreewisesurjections k and weak equivalence are quasi-isomorphisms. For any map of operads φ : P →Q we have an induced adjunction of the corresponding categories of algebras: P−dgAlgk oo φ! // Q−dgAlgk. φ∗ AΣ-cofibrantoperadPisanoperadsuchthatP(n)isk[Σ ]-cofibrantindgMod . n k[Σn] AnycofibrantoperadP is aΣ-cofibrantoperad[2, Proposition4.3]. We denote the associative operad by Ass and the commutative operad by Com.The operad Ass is an admissible operad and Σ-cofibrant, while the operad Com is not admissible in general. In the rational case, when k = Q the operad Com is admissible but not Σ-cofibrant. More generally any cofibrant operad P is admissible [2, Proposition 4.1,Remark4.2]. Wedefineasymmetrictensorproductofoperadsbytheformulae [P⊗Q](n)=P(n)⊗Q(n), ∀ n∈N. Lemma 1.1. Suppose that φ : Ass → P is a cofibration of operads.The operad P is admissible and the functor φ∗ : P − dgAlg → dgAlg preserves fibrations, k k weak equivalences and cofibrations with cofibrant domain in the inderleing category dgMod . k 3 Proof. Firstofall,theoperadPisadmissible,indeedweusethecofibrantresolution r:E →Com and consider the following pushout in Op given by: ∞ k Ass (cid:31)(cid:127) //E ∞ ∞ ∼ α As(cid:15)(cid:15)(cid:15)(cid:15) s f //E′(cid:15)(cid:15) ∞ Where Ass in the cofibrant replacement of Ass in Op and Ass →E is a cofi- ∞ k ∞ ∞ bration. Since the category Op is left proper in the sense of [8, Theorem 3], we k have that α:E →E′ is an equivalence. We denote by I the unit interval in the ∞ ∞ category dgMod which is strictly coassociative. The opposite endomorphism op- k eradEndop(I)hasastructureofE -algbraandAss -algebrawhichfactorsthrough ∞ ∞ Ass i.e., we have two compatible maps of operads: Ass (cid:31)(cid:127) // E ∞ ∞✺ ✺ ✺ ✺ ∼ α ✺ As(cid:15)(cid:15)(cid:15)(cid:15) s(cid:31)❙(cid:127) ❙❙f❙❙❙❙//E❙❙′∞(cid:15)(cid:15) ❙❙❙❙✺✺❙✺✺❙✺❙✺✺))✺✺✺✺##✺(cid:26)(cid:26) Endop(I) bytheuniversalityofthepushout,wehaveamapofoperadsE′ →Endop(I). This ∞ ′ means that the unit interval I has a structure of E -colagebra [2, p.4]. Moreover, ∞ we have a commutative diagram in Op given by k As(cid:127)s_ ∆ //Ass⊗Ass φ⊗f //P⊗E′∞ φ id⊗r ∼ (cid:15)(cid:15) (cid:15)(cid:15)(cid:15)(cid:15) P id //P⊗Com=P where the diagonal map ∆ : Ass → Ass⊗Ass is induced by the diagonals Σ → n Σ ×Σ . Hence,themapP⊗E′ →Padmitsasection. Itimpliesby[2,Proposition n n ∞ 4.1],thatPisadmissibleandΣ-cofibrant. SinceallobjectsinP−dgAlg arefibrant k and φ∗ is a right Quillen adjoint, it preserves fibrations and weak equivalences. Since P is an admissible operad, we have a Quillen adjunction dgAlgk oo φ! //P−dgAlgk, φ∗ where the functor φ∗ is identified to the forgetful functor. Moreover, the model structure on P − dgAlg is the transfered model structure from the cofibrantly k generated model structure dgAlg via the adjunction φ,φ∗. Suppose that f :A→ k ! C is a cofibrationin P−dgAlg suchthat A is cofibrantin dgMod . We factorthis k k map as a cofibration followed by a trivial fibration A(cid:31)(cid:127) i //P p ////B ∼ 4 in the category dgAlg . By [7, Lemma 4.1.16], we have an induced map of endo- k morphism operads (of diagrmas): End →End {A→P→B} {A→B} which is a trivial fibration. Moreover,we have the following commutative diagram in Op k As(cid:127)s_ //End99{A→P→B} ∼ (cid:15)(cid:15) (cid:15)(cid:15)(cid:15)(cid:15) P //End {A→B} Since Op is a model category, it implies that we have a lifting map of operads k P→End , hence i and p are maps of P−dgAlg . Therefore, we consider {A→P→B} k the following commutative square in the category P−dgAlg k A(cid:127)_ i //>>P r f ∼ p (cid:15)(cid:15) (cid:15)(cid:15)(cid:15)(cid:15) B id // B the lifting map r exists since P−dgAlg is a model category, we conclude that k p◦r=idandr◦f =i,whichmeansthatf isaretractofi,hencef isacofibration in dgAlg . (cid:3) k Remark 1.2. Withthesamenotationasin1.1,ifAisacofibrantobjectinP−dgAlg k thenAisacofibrantobjectindgMod . Indeedk →AisacofibrationinP−dgAlg , k k by the previous lemma k → A is a cofibration in dgAlg . Therefore, k → A is a k cofibration in dgMod . k 2. Suspension in CGDA and DGA ′ We denote the the operad E of the previous section by E , and k=Q. ∞ ∞ 2.1. E -DGA. We have a map of operads Ass → Com, which we factor as cofi- ∞ bration followed by a trivial fibration. Ass(cid:31)(cid:127) //E ∼ ////Com ∞ As a consequence, we have the following Quillen adjunctions dgAlgk oo Ab∞// E∞dgAlgk oo str // dgCAlgk U U′ These adjunctions have the following properties: • The functors U′ and U ◦U′ and are the forgetful functors, they are fully faithful cf 2.3 and 2.2. • Thefunctorsstr, U′ formaQuillenequivalencesincek =Qcf[6,Corollary 1.5]. The functor str is the strictification functor. • The functors Ab , U form a Quillen pair. ∞ • The composition str ◦Ab is the abelianization functor Ab : dgAlg → ∞ k dgCAlg . k • The functors str and Ab are idenpotent functors. cf 2.3 and 2.2. 5 The model categories dgCAlg∗ and dgAlg∗ and E dgAlg∗ are pointed model cate- k k ∞ k gories. It is natural to introduce the suspension functors in these categories. Definition 2.1. Let C be any pointed model category, we denote the point by 1, and let A∈C, a suspension ΣA is defined as hocolim(1←A→1). Proposition 2.2. Anymapf :A→S inE dgAlg ,whereS isindgCAlg factors ∞ k k in a unique way as A → str(A) → S and the forgetful functor U′ : dgCAlg → k E dgAlg is fully faithful. Moreover, the unit of the adjunction ν : A → str(A) ∞ k A is a fibration. Proof. Suppose thatwehaveamaph:R→S inE dgAlg suchthatRandS are ∞ k objectsindgCAlg . BydefinitionoftheoperadE themaphhastobeassociative, k ∞ therefore h is a morphism in dgCAlg since R and S are commutative differential k graded algebras. The forgetful functor U′ : dgCAlg → E dgAlg is fully faithful, k ∞ k thisimpliesthatstr(S)=S foranyS ∈dgCAlg . Wehaveacommutativediagram k ′ induced by the unit ν of the adjunction (U , str) : A f //S νA νS=id str((cid:15)(cid:15)A) str(f//)str(S)(cid:15)(cid:15) =S. We conclude that f = str(f) ◦ ν . The surjectivity of the ν follows from the A A universal property of str(A). Hence, ν is a fibration in E dgAlg . (cid:3) A ∞ k Proposition2.3. Anymapf :A→S indgAlg ,whereS isindgCAlg factorsina k k uniquewayasA→Ab(A)→S andtheforgetfulfunctorU◦U′ :dgCAlg →dgAlg k k isfullyfaithful. Moreover,theunitoftheadjunctionν :A→Ab(A)isafibration. A Proof. The proof is the same as in 2.2. (cid:3) Proposition 2.4. Suppose that we have a trivial cofibration k →k in E dgAlg . ∞ k Then the universal map π : Ab(k) → str(k) is a trivial fibration and admits a section in the category dgCAlg . k Proof. We consider the following commutative diagram in E dgAlg : ∞ k k ∼ //k id (cid:15)(cid:15) (cid:15)(cid:15) k =str(k) ∼ //str(k). The map k → str(k) is an equivalence since str is left Quillen functor, the same thing holds for the abelianization functor i.e., k → Ab(k) is a trivial fibration, since k → k is a trivial cofibration in dgAlg 1.1 and Ab is a left Quillen functor. k On another hand the map k → str(k), which is a trivial fibration in E dgAlg ∞ k and hence in dgAlg , can be factored (cf 2.3) as k → Ab(k) → str(k), where k Ab(k)→str(k)isatrivialfibrationbetweencofibrantobjectindgCAlg . Itfollows k that we have a retract l:str(k)→Ab(k). (cid:3) Definition 2.5. The suspension functor in the pointed model categoriesdgCAlg∗, k dgAlg∗ and E dgAlg∗ are denoted by B, Σ and B . k ∞ k ∞ 6 Lemma 2.6. Suppose that A is a cofibrant object in E dgAlg∗, and i : A → k a ∞ k cofibration, then str(B A) is a retract of Ab(ΣA) in the category dgCAlg . ∞ k Proof. First of all if a map f is associative, commutative resp. E -map we put ∞ an index f , f resp. f , notice that by definition of the operad E any E -map a c ∞ ∞ ∞ is a strictly associative map. Suppose that A is a cofibrant object in E dgAlg . ∞ k Consider the following commutative square: A(cid:127)(cid:31)_(cid:127) i∞ // k(cid:127)_ i∞ ha f∞ k(cid:15)(cid:15) (cid:31)(cid:127) ha //ΣA(cid:15)(cid:15) ∃! f∞ ua ** "" (cid:22)(cid:22) B A ∞ where ΣA is the (homotopy 1.1) pushout in dgAlg and B A is the (homotopy) k ∞ pushout in E dgAlg . By proposition 2.2 and proposition 2.3 we have a following ∞ k commutative square in dgAlg : k ΣA ua //B A ∞ (cid:15)(cid:15) (cid:15)(cid:15) Ab(ΣA) xc // str[B A]=B[str(A)]. ∞ By 2.4 we have an inclusion of commutative differential graded algebras l : c str(k)→Ab(k) and after strictification we obtain on another (homotopy) pushout square in dgCAlg given by k str(A)(cid:31)(cid:127) ic //str(k)(cid:31)(cid:127) lc //Ab(k) (cid:127)_ (cid:127)_ (cid:127)_ ic fc str((cid:15)(cid:15) k)(cid:31)(cid:127) fc //B[str(cid:15)(cid:15)(A)] hc (cid:127)_ ∃! Alcb((cid:15)(cid:15) k)(cid:31)(cid:127) hc uc //Ab&& (Σ(cid:15)(cid:15)(A)). In order to prove that B[str(A)] is a retract of Ab(Σ(A)) it is sufficient to prove that x ◦h ◦l =f . c c c c 7 By proposition 2.2 and proposition 2.3, the composition E -maps ∞ k f∞// B A // str[B A] ∞ ∞ can be factored in a unique way as k //Ab(k) π // str(k) αc //str[B A]=B[str(A)]. ∞ By unicity, α = f . On another hand, using the first poushout in E dgAlg , the c c ∞ k previous composition k →str[B A] is factored as ∞ k ha //ΣA // Ab(ΣA) xc //str[B A]. ∞ We summarize the previous remarks in the following commutative diagram: k pr // Ab(k) π //str(k) fc //str[B A] ∞ id hc id (cid:15)(cid:15) (cid:15)(cid:15) (cid:15)(cid:15) k //Ab(ΣA) xc // str[B A] ∞ by definition of h , the doted map h makes the left square commutative. Since a c the whole square is commutative and the map pr is surjective we conclude that x ◦h = f ◦π. Since the map l : Str(k) →Ab(k) is a retract of π (Cf. 2.4) i.e., c c c c π◦l = id, we conclude that x ◦h ◦l = f . Finally, by unicity of the pushout, c c c c c we deduce that the following composition B[str(A)] uc // Ab(ΣA) xc //B[str(A)] is identity. (cid:3) 3. Main result and applications Theorem 3.1. For any R and S in dgCAlg∗, the induced map by the forgetful k functor ΩMapdgCAlg∗(R,S)→ΩMapdgAlg∗(R,S), k k has a retract, in particular πiMapdgCAlg∗(R,S)→πiMapdgAlg∗(R,S) k k is injective ∀ i>0. Proof. Suppose that R is (cofibrant) object in E dgAlg and S any object in ∞ k dgCAlg . By adjunction, we have that k ΩMapdgCAlg∗(str(R),S) ∼ MapdgCAlg∗(B[str(R)],S) (3.1) k k ∼ MapdgCAlg∗(str[B∞R],S) (3.2) k ∼ MapE∞dgAlg∗k(B∞R,S) (3.3) ∼ ΩMapE∞dgAlg∗k(R,S). (3.4) By Lemma 2.6, we have a retract MapdgCAlg∗(B[str(R)],S)→MapdgCAlg∗(Ab(ΣR),S)→MapdgCAlg∗(B[str(R)],S). k k k Again by adjunction: MapdgCAlg∗(Ab(ΣR),S)∼MapdgAlg∗(ΣR,S)∼ΩMapdgAlg∗(R,S). k k k 8 We conclude that ΩMapE∞dgAlg∗k(R,S) U // ΩMapdgAlg∗k(R,S) // ΩMapE∞dgAlg∗k(R,S) is a retract. Hence, the forgetful functor U induces a injective map on homotopy groups i.e., πiMapdgCAlg∗k(str(R),S)≃πiMapE∞dgAlg∗k(R,S)→πiMapdgAlg∗k(R,S) is injective ∀ i>0. (cid:3) 3.1. Rational homotopy theory. We give an application of our theorem 3.1 in the context of rational homotopy theory. Let X be a simply connected rational spacesuchthatπ X isfinite dimensionalQ-vectorspaceforeachi>0. LetC∗(X) i be the differential graded Q-algebra cochain associated to X which is a connective E∞dgAlgk. By Sullivan theorem πiX ≃ πiMapdgCAlg∗(C∗(X),Q). By 3.1, we have k that πiX is a sub Q-vector space of πiMapdgAlg∗(R,S). On another hand [1], since k C∗(X) is connective, we have that for any i>1 πiMapdgAlg∗(C∗(X),Q)≃HH−1+i(C∗(X),Q), k where HH∗ is the Hochschild cohomology. Since we have assumed finiteness condi- tion on X, we have that HH−1+i(C∗(X),Q)≃HH (C∗(X),Q). i−1 The functor C∗(−,Q):Topop →E dgAlg commutes with finite homotopy limits, ∞ k where Top is the category of simply connected spaces. Hence, HH (C∗(X),Q)=Hi−1[C∗(X)⊗L Q]≃Hi−1(ΩX,Q). −1+i C∗(X×X) We conclude that π X is a sub Q-vector space of Hi−1(ΩX,Q). i More generally by Block-Lazarev result [3] on rational homotopy theory and [1], we have an injective map of Q-vector spaces AQ−i(C∗(X),C∗(Y))→HH−i+1(C∗(X),C∗(Y)), where the C∗(X)-(bi)modules structure on C∗(Y) is given by C∗(X) → Q → C∗(Y), and AQ∗ is the Andr´e-Quillen cohomology. We also assume that X and Y are simply connected and i>1. More generally, π Map (R,S)=AQ−i(R,S)→HH−i+1(R,S)=π Map (R,S) i E∞dgAlgk i dgAlgk is an injective map of Q-vector spaces for all i > 1 and any augmented E - ∞ differential graded connective Q-algebras R and S, where the action of S on R is given by S →Q→R. Acknowledgement : I’m grateful to Benoˆıt Fresse for his nice explanation of Lemma 1.1, the key point of the proof is due to him. 9 References [1] Ilias Amrani. The mapping space of unbounded differential graded algebras. arXiv preprint arXiv:1303.6895, 2013. [2] C.BergerandI.Moerdijk.Axiomatichomotopytheoryforoperads.CommentariiMathematici Helvetici,78(4):805–831, 2003. [3] Jonathan Block and Andrej Lazarev. Andr´e–Quillen cohomology and rational homotopy of functionspaces.Advances in Mathematics,193(1):18–39, 2005. [4] Benoit Fresse. The bar complex of an E-infinity algebra. Advances in Mathematics, 223(6):2049–2096, 2010. [5] Kathryn Hess. Rational homotopy theory: a brief introduction. Contemporary Mathematics, 436:175, 2007. [6] IgorKrizandJPeterMay.Operads, algebras, modulesandmotives.Soci´et´emath´ematiquede France,1995. [7] Charles Waldo Rezk. Spaces of algebra structures and cohomology of operads. PhD thesis, Massachusetts InstituteofTechnology, 1996. [8] MarkusSpitzweck.Operads,algebrasandmodulesingeneralmodelcategories.arXivpreprint math/0101102, 2001. DepartmentofMathematics,MasarykUniversity,Kotlarska 2,Brno,Czech Repub- lic. E-mail address: [email protected] E-mail address: [email protected]