A Model Structure on the Category of A∞-Categories with Strict Morphisms Authors: Mattia Ornaghi Published: 2026 ABSTRACT AbstractWe prove that the category of (strictly unital) A∞-categories, linear over a commutative ring R, with strict A∞-morphisms has a cofibrantly generated model structure. In this model structure every object is fibrant and the cofibrant objects have cofibrant morphisms. As a consequence we prove that the semi-free A∞-categories (resp. resolutions) are cofibrant objects (resp. resolution) in this model structure.Keywords: A-infty-categories, DG-categories, Model Categories Introduction and Statement of Results We fix a commutative ring R, an A∞-category is a R-linear DG-category associative up to homotopy. In a few words, an A∞-category A is a graded category equipped with multilinear maps for every integer n≥1 and sequence of objects x0,,...,xn∈A, satisfying axioms (see [13, Definition 1.1.1 (1.1)] or [16, §2.1]). Considering mA1 as the differential, we can associate to A the (graded) category H(A) whose hom spaces are given by: Taking two A∞-categories A and B, we call A∞-functor a family of multilinear maps {Fn}n≥0 of the form for every integer n≥1 and sequence of objects x0,...,xn∈A, satisfying axioms (see [13, Definition 1.2.1 (1.2)]). There are several notions of unit in the framework of A∞-categories, in this paper we consider the strictly unital A∞-categories with strictly unital A∞-functors (see [13, Definitions 1.1.4]). From now on, we denote by A∞Cat the category of A∞-categories with the A∞-functors. Note that A∞Cat is not complete (since it does not admit equalizers [2, Lemma 1.28]). We say that an A∞-functor F is a quasi-equivalence if it induces an equivalence of (graded) categories: and an equivalence H0(F):H0(A)→H0(B). We denote by Ho(A∞Cat) the homotopy category, i.e. the (Gabriel–Zisman) localization of A∞Cat with respect to the class of quasi-equivalences. A∞-categories were introduced in the early 1990s in the context of Homological Mirror Symmetry since the Fukaya category of a symplectic manifold, comes naturally equipped with a structure of this kind. Note that the pioneers of this field, such as Kontsevich, Fukaya, Oh, Ono, Ohta, Seidel, Soibelman, etc, assumed by definition a flatness hypothesis on the hom-spaces of an A∞-category. For example, in [6, Definition 1.1] and [7, §3.2.1] an A∞-category A is such that the hom-spaces A(x,y) are graded free R-modules, or the base commutative ring R is assumed to be a field, see [11] or [17]. On the other hand, the definition of A∞-category makes sense without any restriction on the hom-spaces. For this reason, more recently, many people such as Ganatra, Pardon, Shende, Oh, Tanaka started to used the term cofibrant A∞-category to indicate an A∞-category whose hom-spaces are h-projective DG-modules (see [8, Definition 2.6], [21, 22, Definition 1.2]). Note that, despite the name, this has nothing to do with a model structure on A∞Cat, it is well known that A∞Cat has no model structure [2, §1.5]. The term cofibrant is inherited by the DG-categories. Indeed a cofibrant DG-category (in the model structure of Example 2.2), has cofibrant hom-spaces. In particular it is a h-projective DG-category. The first goal of this paper is to give a precise definition of cofibrant A∞-category, namely we provided a model structure on the category of A∞-categories (taking a subset of A∞-functors). In this model structure, if an A∞-category is cofibrant then it has cofibrant hom-spaces, in particular it is h-projective (see Theorem A). Despite the lack of a model structure, we can describe the hom-spaces of the homotopy category of A∞Cat as follows: Here Ahps denotes a h-projective with splits unit A∞-category which is quasi-equivalent to A and ≈ denotes the weakly equivalence relation [14, Theorem A]. In order to prove that every A∞-category A has a resolution of the form Ahps, the notion of semi-free A∞-category was introduced in [14], where it was proven that the semi-free resolution of A is an h-projective A∞-category with splits units. Note that, in the case of DG-categories, the semi-free resolutions correspond to the cofibrant resolutions in Tabuada model structure (see Example 2.2). For this reason in loc. cit. it was left as an open question if the semi-free A∞-categories are a kind of cofibrant resolutions in an appropriate category. We recall that an A∞-functor F={Fn}n≥0 is called strict if Fn≥2=0, in particular F1 preserves the underlying graded quivers, id est: for every n≥1 and a1,...,an∈A. We denote by A∞Catstrict the category of A∞-categories with the strict A∞-functors. In this note we prove the following: Theorem A There is a cofibrantly generated model structure on A∞Catstrict whose weak equivalences are the quasi-equivalences. In such a model structure the fibrations are the isofibrations (see Definition 3.2) strict A∞-functors which are surjective on the morphisms. Every category is a fibrant object and every cofibrant object A is such that A(a1,a2) is a cofibrant object in Ch(R). Despite the category A∞Cat is not complete, A∞Catstrict is complete and cocomplete (see [14, Theorem 4.5]). As a consequence of Theorem A, we have that the semi-free A∞-categories are cofibrant (cf. [14, Lemma D]). Moreover, consider the category I so defined: Definition 1.1 I is the DG-category with two objects and two closed morphisms of degree zero such that j01·j10=Id2 and j10·j01=Id1. It is not difficult to prove that, given an A∞-category A, the path object of A described in [14, 2.5] is exactly the same path object of A in A∞Catstrict, with the model structure of Theorem A. Cf. also [20, §3] and [20, Remark 2.7]. We conclude by saying that, given an A∞-category A, we can take the cofibrant resolution Acof (with respect to the model structure of Theorem A) which is cofibrant in the sense of Ganatra, Pardon, Shende et al. State of Art By the work of Lefevré-Hasegawa [12] the category Alg∞ of (non unital) A∞-algebras linear over a field, has a model structure whose weak equivalences are the quasi-isomorphisms. In this model structure a morphism f is a fibration (resp. cofibration) if f1 is surjective (resp. injective). Note that Alg∞ has no equalizers (see [2]) and coproducts. To see that Alg∞ has no coproduct we use the fact that (the bar-cobar functor) U:Alg∞→DG-Alg is right adjoint to the inclusion (see [3]). Since the tensor product is the coproduct of two DG-algebras, we must have U(A)⊗U(B)≃U(A⊗B), for any A∞-algebras. Using this fact, it is easy to see that Alg∞ has no coproduct (it has to do with the fact that there is no a good notion of tensor product of A∞-algebras [15]). This is no longer true if we consider the category of A∞-categories, namely, it has coproducts (which is the disjoint union) but it does not have equalizers [2, Lemma 1.28]. On the other hand, if we consider the category of A∞-algebras with strict A∞-morphisms we have a model structure by [9, 2.2.1. Theorem]. In this model structure the fibrations are the morphisms f such that f1 is surjective. Moreover, in [4] we proved that the category of A∞-categories, linear over a field, is a fibrant category. The fibrations are the A∞-functors F which are isofibrations and such that F1 is degree-wise surjective. We conclude by saying that we do not known if Alg∞ and A∞Cat have coequalizers. This is an interesting question since, if it was true (at least in some cases i.e. along the cofibrations), one could try to prove that the category A∞Cat (linear over a commutative ring) has a structure of cofibrant category. Model Structures and Recognition Theorem We give two examples of model structures cofibrantly generated. It will be crucial the following result which goes under the name of Recognition Theorem and corresponds to [10, Theorem 2.1.19]: Theorem 2.1 (Recognition Theorem) Let C be a complete and cocomplete category with W a subcategory of C, and I and J sets of maps of C. There is a cofibrantly generated model structure on C with I as the set of generating cofibrations, J as the set of generating trivial cofibrations, and W as the subcategory of weak equivalences, if and only if the following conditions are satisfied: The subcategory W has the two out of three property and is closed under retracts. The subcategory W has the two out of three property and is closed under retracts. The domains of I are small relative to I-cell. The domains of I are small relative to I-cell. The domains of J are small relative to J-cell. The domains of J are small relative to J-cell. J-cell ⊂ W∩I-cof. J-cell ⊂ W∩I-cof. I-inj ⊂ W∩J-inj. I-inj ⊂ W∩J-inj. Either W∩I-cof⊂J-cof or W∩J-inj⊂I-inj. Either W∩I-cof⊂J-cof or W∩J-inj⊂I-inj. We recall that f∈I-inj if f has the right lifting property with respect to any morphism in I. We say that f∈I-cof if f has the left lifting property for any I-injective morphism. Example 2.1 We denote by Ch(R) the category of unbounded chain complexes over a commutative ring R. The category Ch(R) has a model structure whose weak-equivalences are the quasi-isomorphisms (see [10, Definition 2.3.3 and Theorem 2.3.11]). In order to define such a model structure we define the chain complexes Sn-1 and Dn. Fixed an integer n, where R is in degree -n+1, and where the only nonzero components are in degree -n and -n+1. The set of generating cofibrations I is given by the maps in:Sn-1→Dn, for each n∈Z. The set of generation trivial cofibrations J consists of jn:0→Dn, for each n∈Z. In this model structure the fibrations are the maps f such that fn is surjective for all n∈Z and the cofibrant objects are the h-projective degreewise projective modules (see [1, Theorem 9.6.1 (ii′) iff (v′)]). Before continuing, we define by K the Kontsevich category: Definition 2.1 K is the DG-category with two objects generated by the following morphisms: with the relations: d(r12)=r2·f-f·r1. d(r12)=r2·f-f·r1. d(r1)=g·f-Id1. d(r1)=g·f-Id1. d(r2)=f·g-Id2. d(r2)=f·g-Id2. Note that K is a semi-free resolution of the DG-category I, where I is defined in (1.1); see [5, 3.7.6. Remark]. To prove that K is semi-free, we consider the filtration: Here I0 is the discrete category with two objects: 1 and 2. I0 is the discrete category with two objects: 1 and 2. I1 is the DG-category freely generated by two closed generators f and g I1 is the DG-category freely generated by two closed generators f and g I2 is the DG-category freely generated on I1 by the generators r1 and r2, of degree 1, such that d(r1)=g∗f-Id1 and d(r2)=f∗g-Id2. I2 is the DG-category freely generated on I1 by the generators r1 and r2, of degree 1, such that d(r1)=g∗f-Id1 and d(r2)=f∗g-Id2. I3 is the DG-category freely generated on I2 adding the generator r12 and taking the quotient by the DG-ideal (d(r12)-r2∗f+f∗r1). I3 is the DG-category freely generated on I2 adding the generator r12 and taking the quotient by the DG-ideal (d(r12)-r2∗f+f∗r1). We have the DG-functors Ψ0:I0→I, Ψ1:I1→I and Ψ2:I2↠I, Ψ:K↠I. In particular, Ψ2 and Ψ are surjective on the morphisms and Ψ is a quasi-equivalence. In particular Ψ(j01)=f, Ψ(j10)=g and Ψ(r1)=Ψ(r2)=Ψ(r12)=0. Remark 2.1 Note that we added r12 in I3, since is a closed morphism which vanish in I via Ψ. It means that, if we want to make Ψ a quasi-equivalence then (6) must be a coboundary in cohomology. On the other hand, also is a closed morphism whose image via Ψ2 is 0. Nevertheless we do not need to add a new generator in I3, since (7) is the differential of the following morphism: Example 2.2 In [18], Tabuada proved that the category of DG-categories has a cofibrantly generated model structure whose weak-equivalences are the quasi-equivalences. The set of generating cofibrations I consists of the DG-functors Q and S(n) described as follows. First, we denote by A the DG-category: Q is the (only) DG-functor from the empty set (which is the initial category of DG-cats and A∞Cat). Fixed an integer n, C(n) and P(n) are the two DG-categories having two objects C1, C2 and P1, P2 and whose hom-spaces are given by and S(n) is the DG-functor Now we define the trivial cofibrations. First, we denote by B the DG-category which has two objects B1 and B2 and no non trivial morphisms. Fixed an integer n, the DG-functor R(n):B→P(n) is defined as follows: The functor F:A→K is the DG-functor sending the object A of A to the object 1 of K. The trivial cofibrations are generated by J:={F,R(n)}. In this model structure the fibrations are the isofibrations (see Definition 3.2) which are degreewise surjective [19, Proposition 1.13]. Moreover, every object is fibrant and the cofibrant objects are such that the hom-spaces are cofibrant in Ch(R), see [23, Proposition 2.3 (3)]. Proof of Theorem A In this section we prove the main result using the Recognition Theorem 2.1. First, we want to find an appropriate “K” (defined in Definition 2.1) in the case of A∞-categories. We recall that, given an A∞-category A, we can take a semi-free resolution of A. In [14, §5], we give a procedure to find such a resolution. We define Definition 3.1 The category KA∞ is the A∞ semi-free resolution of the category I (see Definition 1.1) defined as follows: 0.I0 is the discrete (strictly unital) category with two objects: 1 and 2. I0 is the discrete (strictly unital) category with two objects: 1 and 2. 1.I1 is the A∞-category freely generated by the closed morphisms j12 and j21. I1 is the A∞-category freely generated by the closed morphisms j12 and j21. 2.I2 is the A∞-category freely generated on I1 adding the generators r1 and r2, such that (i).(ii). I2 is the A∞-category freely generated on I1 adding the generators r1 and r2, such that (i).(ii). (i). . (ii). . Note that: As we already noted, this is not the same in the case of DG-categories, since This is why we need to add the free generator r12 in K which is the coboundary of the cocycle r1∗f-f∗r2. Nevertheless KA∞ and K are weakly equivalent since both are h-projective with splits unit and quasi-equivalent to I (see [14, Theorem 5.2] and [3, Definition 3.4]). It is important to say that, since KA∞ is h-projective, I(n,m)=(R,0) is a free R-module (for any n,m={1,2}) and they are quasi-equivalent, then (KA∞,mA∞1) is homotopy equivalent to the complex (R, 0). In particular, since KA∞ has split unit, it implies that the short exact sequence of graded complexes splits. Theorem 3.1 Let F′ to be the strict A∞-functor from A to KA∞ such that F′(A):=1. The category A∞Catstrict has a model structure cofibrantly generated by J′:={F′,R(n)} (defined in (11)) and I (defined in Example 2.2). First we give two definitions and a Lemma. Definition 3.2 An A∞-functor (resp. a DG-functor) F:A→B is an isofibration if, given a∈A and an isomorphism g:F(a)→b in H0(B), there exists an isomorphism f:a→a′ in H0(A), such that F1(f)=g. Definition 3.3 An A∞-functor (resp. a DG-functor) F:A→B is surjective if, F0 is surjective (as a map of sets) and F1 is a surjective quasi-isomorphism of complexes. We denote by Surj the set of strict surjective A∞-functors. The following Lemma which will be useful also to characterize the fibrations of the model structure provided by Theorem 3.1. Lemma 3.2 A strict A∞-functor F:C→D has right lifting property with respect to R(n) if and only if F1 is surjective on the morphisms. S(n) if and only if F1 is a quasi-isomorphism surjective on the morphisms. F if and only if F1 is an isofibration. Q if and only if F0 is surjective (as a map of sets). Proof Suppose that F is surjective on the morphisms, consider the diagram below: It is easy to find a T~ since P(n) is surjective on the morphisms and P(n) is uniquely determined by the image of T(R[1]⊕R). Suppose now that F has the right lifting property with respect to R(n). For every morphism g∈D we take the functor If there exists T~:P(n)→C such that Taking T~(1,0)∈C, we have F(T~(1,0))=T((1,0))=g, so F is surjective on morphisms. Now we prove F1′). We note that a diagram of the form: corresponds to the datum Moreover, the existence of a lift T~, corresponds to the datum Suppose that every diagram of the form (12) has a lift, we show that F1 is a surjective quasi-isomorphism. First, we prove that [F1] is surjective. Let g~∈D be a closed morphism, the pair (0,g~) gives rise to a diagram of the form (12). Since it admits a lift then, by the characterization (14), there is a closed morphism f~ such that F1(f~)=g~. Now we prove that F1 is surjective. Given a morphism g∈D, we can take the datum (0,mD1(g)), we just shown that there exists a closed morphism f~∈C such that F1(f~)=mD1(g). Using the characterization (13), the pair (f~,g) gives rise to a new diagram of the form (12). Since it has a lift, then it exists f∈C such that mC1(f)=f~ and F1(f)=g, and we are done. We prove that [F1] is injective. Suppose that f is a closed morphism in C such that F1(f) vanishes in cohomology. It means that there exists g such that F1(f)=mD1(g). Then (f, g) forms a diagram of the form (12). Since it is liftable, there exists f~ such that mC1(f~)=f so f is zero in cohomology. To conclude, we need to prove that, if F1 is a surjective quasi-isomorphism, then every diagram of the form has a lift. Using the characterization (13), we have (f,g~) such that mC1(f)=0andF1(f)=mD1(g~). We consider the short exact sequence of chain complexes: Since F1 is a quasi-isomorphism then the cohomology of Ker(F1) is trivial. Moreover F1 is surjective, so we can take a morphism h∈C(x,y) such that F1(h)=g~. We note that the morphism mC1(h)-f∈C is closed and F1(mC1(h)-f)=mD1(g~)-F1(f)=0, so mC1(h)-f∈Ker(F1). To conclude, since the cohomology of Ker(F1) is trivial, it exists j∈Ker(F1) such that mC1(j)=mC1(h)-f. Thanks to the characterization (14), the morphism f~=h-j corresponds to a lift of (15) and we are done. To prove F2), note that every strict A∞-functor KA∞→D is (uniquely) determined by the image of the generators. Suppose that F has right lifting property with respect to F, so for every commutative diagram of the form: there exists a lift T~. Suppose there exists g:F0(c)→d∈H0(D). It means that there exist g-1, h, t such that: and Then we can take the functor T:KA∞→D so defined: Since F has a right lifting property, there exists T~:KA∞→C. It implies that there exists an isomorphism T~(j12)=:f and an object T~(2)=:c′ in H0(C) such that F1(f)=T(j12)=g, and we are done. The proof of viceversa is similar. Now we prove F3). We consider the commutative diagram: since every functor from A is uniquely determined by the image of the object A. It is easy to see that if there exists a lifting T~ for every T then F must be surjective on the objects. On the other hand, if F0 is surjective then we can find a T~ for every T. □ Proof of Theorem 3.1 We want to use Theorem 2.1, so we must verify the six conditions of the theorem. 1. 2. 3.Straightforward. Straightforward. 4.First we note that I-inj = W∩ J-inj ⊂ J-inj. It implies that J-cof ⊂ I-cof and J-cell ⊂ J-cof ⊂ I-cof. So it remains to prove that, for every A∞-category M and ⋆∈J′, the strict A∞-functor inc, fitting the pushout diagram 17 is a quasi-equivalence. We consider the push-out diagram: 18 First we note that I-inj = W∩ J-inj ⊂ J-inj. It implies that J-cof ⊂ I-cof and J-cell ⊂ J-cof ⊂ I-cof. So it remains to prove that, for every A∞-category M and ⋆∈J′, the strict A∞-functor inc, fitting the pushout diagram 17 is a quasi-equivalence. We consider the push-out diagram: 18 We denote N(A)∈M by z. The push-out P is obtained by taking the disjoint union of M and KA∞ and gluing the object z of M with the object 1 of KA∞. See the following picture: We have: where Here K¯A∞(z,z) is the chain complex: Note that we take the quotient complex (20) because we “glue” the identity of the object z∈M with the identity of the object 1∈KA∞. Note that the chain complex (KA∞(1,1),mA∞1) is homotopy equivalent to (R, 0) and K¯A∞(z,z) is contractible. So is a quasi-isomorphism. It is also clear that H0(inc) is essentially surjective since N′(2) is quasi-isomorphic to N′(1)=N(A)=z. On the other hand, we consider the push-out diagram: The category P is given by the disjoint union of M and P(n), and by gluing the object T(B1) with the object P1, and the object T(B2) with the object P2. See Fig. 1. The push-out P We have: where Since Dn is contractile then inc is a quasi-equivalence and we are done. 5. 6.We claim that Surj = I-inj = J-inj ∩W. To prove of Surj = I-inj we can use Lemma 3.2 item F1′) and F3). Note that it is the same of [19, Lemme 1.11], since the set of generating cofibrations I is the same of Example 2.2. Now we want to prove J-inj ∩W = Surj. If f∈J-inj∩W then f has right lifting property with respect to R(n) so, by Lemma 3.2 item F1), it is surjective on the morphisms. For the item F3) of the same Lemma it is an isofibration. Since it is a quasi-equivalence it is surjective on the objects. On the other hand if f∈Surj, then R(n) has the right lifting property with respect to f (see item F1) of Lemma 3.2). Moreover F has the right lifting property with respect to f since KA∞ is semi-free (see [14, Lemma 6.6]) and we are done. We claim that Surj = I-inj = J-inj ∩W. To prove of Surj = I-inj we can use Lemma 3.2 item F1′) and F3). Note that it is the same of [19, Lemme 1.11], since the set of generating cofibrations I is the same of Example 2.2. Now we want to prove J-inj ∩W = Surj. If f∈J-inj∩W then f has right lifting property with respect to R(n) so, by Lemma 3.2 item F1), it is surjective on the morphisms. For the item F3) of the same Lemma it is an isofibration. Since it is a quasi-equivalence it is surjective on the objects. On the other hand if f∈Surj, then R(n) has the right lifting property with respect to f (see item F1) of Lemma 3.2). Moreover F has the right lifting property with respect to f since KA∞ is semi-free (see [14, Lemma 6.6]) and we are done. □ Corollary 3.3 The fibrations are the isofibrations F such that F1 are degreewise surjective. Every A∞-category is fibrant. Proof It follows directly from Lemma 3.2, every A∞-category is fibrant since the terminal object of A∞Catstrict is the category with one object and one (trivial) morphism A (defined Example 2.2). □ Theorem 3.4 If A is a cofibrant A∞-category then A(x,y) is a cofibrant object in Ch(R). Proof The proof is the same as in the case of DG-categories (see [23, Proposition 2.3 (3)]) since they have the same set of generating cofibrations I of Example 2.2. □ As in the case of DG-categories not all the h-projective A∞-categories are cofibrant object in this model structure. For example the category C(1)⊗C(1) is not a cofibrant object. It is not hard to prove that the map ∅→C(1)⊗C(1) does not have the right lifting property with respect to the trivial fibrations (see [24, Exercise 14 4.]). Acknowledgements I thank the anonymous reviewers for their careful reading of my manuscript and their comments, which greatly contributed to improving the readability of this article. Author Contributions M.O. wrote the manuscript and prepared the figures. Funding Open access funding provided by Università degli Studi di Milano within the CRUI-CARE Agreement. The author was supported by the research project FARE 2018 HighCaSt (grant number R18YA3ESPJ) and by ERC Advanced Grant-101095900-TriCatApp. Data Availability No datasets were generated or analysed during the current study. Declarations Ethical Approval Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement No datasets were generated or analysed during the current study.