Admissible weak factorization systems on extriangulated categories

Since g:X\rightarrow Y is a deflation, there is an \mathbb{E}-triangle \xymatrix{K\ar[r]^{h}&X\ar[r]^{g}&Y\ar@{-->}[r]^{\rho}&}. For any U\in \mathcal{U} and take \delta\in \mathbb{E}(U,K). Then there exists a conflation K\xrightarrow{i} E\xrightarrow{} U which realizes \delta. Thus we have the following commutative diagram by [NP, Proposition 3.15]:

Consider the following commutative diagram

Since \widetilde{i} is a inflation with \mathrm{Cone}(\widetilde{i})=U\in \mathcal{U}, there exists t:W\rightarrow X such that t \widetilde{i}=1_X and \widetilde{g}=gt. By \rm (ET 3)^{op}, there is a morphism of \mathbb{E}-triangles

Therefore, there is a morphism of \mathbb{E}-triangles

which implies that ji is an isomorphism. So one can get that i is a split inflation, and \delta=0. Thus K\in \mathcal{U}^{\perp}.

The other part of the lemma is dual.

Assume that (\mathcal{U},\mathcal{V}) is a cotorsion pair on \mathcal{C}. We shall prove that every morphism h:X\rightarrow Y factorizes as h=gf with f\in \mathrm{Infl}~\mathcal{U} and g\in \mathrm{Defl}~\mathcal{V}. Suppose first that h is an inflation. Then there is an \mathbb{E}-triangle \xymatrix{X\ar[r]^{h}&Y\ar[r]^{y}&C\ar@{-->}[r]^{\delta}&}. Since (\mathcal{U},\mathcal{V}) is a cotorsion pair, there is an \mathbb{E}-triangle \xymatrix{V^{C}\ar[r]^{}&U^{C}\ar[r]^{}&C\ar@{-->}[r]^{}&} with U^C\in\mathcal{U} and V^C\in\mathcal{V}. By [NP, Proposition 3.15], we obtain a commutative diagram of conflations

The corresponding factorization of h then appears in the leftmost square. A dual argument applies if h is a deflation. If h is an arbitrary morphism, we can factor it as

Note that X\xrightarrow{\tiny\begin{pmatrix}1_X\\0\end{pmatrix}}X\oplus Y is a split inflation. It follows that we can factor it as \tiny\begin{pmatrix}1_X\\0\end{pmatrix}=g_{1}f_{1}

with f_{1}\in \mathrm{Infl}~\mathcal{U} and g_{1}\in \mathrm{Defl}~\mathcal{V}. As (h,1_{Y}) is a deflation by [NP, Corollary 3.16], (h,1_{Y})\circ g_{1} is a deflation. Hence we can factor it as g_{2}f_{2} with f_{2}\in\mathrm{Infl}~\mathcal{U} and g_{2}\in \mathrm{Defl}~\mathcal{V} by above argument.

It follows that h=(h, 1_Y){\tiny\begin{pmatrix}1_X\\0\end{pmatrix}}=(h, 1_Y)g_1f_1=g_{2}(f_{2}f_{1}). Note that f_{2} f_{1} is an inflation and there is a conflation \mathrm{Cone}(f_{1})\rightarrow \mathrm{Cone}(f_{2}f_{1})\rightarrow \mathrm{Cone}(f_{2}) by {\rm (\mathrm{ET4})^{op}}; also see [HZZ1, Lemma 2.11]. Since \mathcal{U} is closed under extensions, it follows that f_{2}f_{1}\in \mathrm{Infl}~\mathcal{U}. Next, we shall prove that \mathrm{Infl}~\mathcal{U}={^{\Box}\mathrm{Defl}~\mathcal{V}}. If f\in \mathrm{Infl}~\mathcal{U} and g\in \mathrm{Defl}~\mathcal{V}, then f\Box g by [HZZ1, Lemma 2.13], which implies \mathrm{Infl}~\mathcal{U}\subseteq{^{\Box}\mathrm{Defl}~\mathcal{V}}. If h\in{^{\Box}\mathrm{Defl}~\mathcal{V}}, then h\Box g for any g\in\mathrm{Defl}~\mathcal{V}. By above argument, we have h=g_1f_1 with f_{1}\in\mathrm{Infl}~\mathcal{U} and g_{1}\in \mathrm{Defl}~\mathcal{V}. Since h\Box g_1, there exists a morphism t: B\rightarrow C which makes the following diagram commutative

Since f_1=th is an inflation, one can get that h is an inflation by Condition (WIC). It follows from Lemma 3.5 that {\rm Cone}(h)\in {^\perp{\mathcal{V}}}, hence h\in \mathrm{Infl}~\mathcal{U} as \mathcal{U}={^\perp\mathcal{V}}, and {^{\Box}\mathrm{Defl}~\mathcal{V}}\subseteq\mathrm{Infl}~\mathcal{U}. So we have \mathrm{Infl}~\mathcal{U}={^{\Box}\mathrm{Defl}~\mathcal{V}}, we can prove \mathrm{Defl}~\mathcal{V}=\mathrm{Infl}~\mathcal{U}^\Box by a dual argument. Therefore, (\mathrm{Infl}~\mathcal{U},\mathrm{Defl}~\mathcal{V}) is a weak factorization system. It is easy to check that (\mathrm{Infl}~\mathcal{U},\mathrm{Defl}~\mathcal{V}) is an admissible weak factorization system.

Conversely, let (\mathcal{L},\mathcal{R}) be an admissible weak factorization system and put \mathcal{U}=\mathrm{Cone}(\mathcal{L}) and \mathcal{V}=\mathrm{CoCone}(\mathcal{R}). Since \mathcal{L} and \mathcal{R} are closed under retracts by Remark 3.3, both \mathcal{U} and \mathcal{V} are closed under direct summands. By Lemma 3.5, we have \mathbb{E}(\mathcal{U},\mathcal{V})=0 for each \mathcal{U}\in\mathcal{U} and V\in\mathcal{V}. In order to prove the existence of approximation sequences, consider X\in \mathcal{C} and the following two factorizations

and

with respect to (\mathcal{L},\mathcal{R}). Thus there are conflations X\rightarrow V_{X}\rightarrow U_{X} and V^{X}\rightarrow U^{X}\rightarrow X with V_{X}, V^{X}\in \mathcal{V} and U_{X}, U^{X}\in \mathcal{U}. Hence (\mathcal{U},\mathcal{V}) is a cotorsion pair on \mathcal{C}.

(1)\Rightarrow (2) Let g:B\rightarrow C and f:A\rightarrow B be morphisms in \mathcal{C} such that gf and g are in \mathrm{Infl} (\mathcal{X}). Since \mathcal{C} satisfies Condition (WIC), f is an inflation. By [NP, Lemma 3.14], we have the following commutative diagram

Since \mathcal{X} is closed under cocones of deflations by the assumption, and both \mathrm{Cone}(gf) and \mathrm{Cone}(g) are in \mathcal{X}, it follows that \mathrm{Cone}(f)\in \mathcal{X} as well, so f\in \mathrm{Infl}~\mathcal{X}.

(2)\Rightarrow(1) Let \xymatrix{X_{1}\ar[r]^{h}&X_{2}\ar[r]^{}&X_{3}\ar@{-->}[r]^{}&} be an \mathbb{E}-triangle with X_{2} and X_{3}\in \mathcal{X}. Let l be the zero morphism from 0 to X_{1}. Then hl\in \mathrm{Infl}~(\mathcal{X}), and so is l as h is in \mathrm{Infl}~(\mathcal{X}). Thus X_{1}\in \mathcal{X}.

The proof is dual to that of Lemma 3.12.

We only prove the “only if " part as the “if " part is a dual statement. Since \mathrm{Infl}~\mathcal{U} satisfies the left cancellation property, \mathcal{U} is closed under cocones of deflation by Lemma 3.12. Since (\mathcal{U},\mathcal{V}) is a cotorsion pair by Theorem 3.11, it follows from [HZZ1, Proposition 2.18] that \mathcal{V} is closed under cones of inflations. Thus \mathrm{Defl}~\mathcal{V} satisfies the right cancellation property by Lemma 3.13.

The equivalence follows from Theorem 3.11 and Lemmas 3.12, 3.13 and 3.14.

Since (\mathcal{L},\widetilde{\mathcal{R}}) is an admissible weak factorization system, we have h=h_{2} h_{1} with h_{1}\in \mathcal{L} and h_{2}\in \widetilde{\mathcal{R}}. Then g=fh_{2}h_{1} with fh_{2}\in \widetilde{\mathcal{R}} by Remark 3.3. If the conclusion holds for h_1, that is, h_1\in \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}}, then h_1\in \mathcal{L}\bigcap \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}}. It follows from [DLLiang, Lemma 1.12] that h_1\in\widetilde{\mathcal{L}}, so h=h_2h_1\in \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}}. So we only deal with the special case that h\in \mathcal{L}. Let h:X\rightarrow Z and f:Z\rightarrow Y. Since g=f h\in \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}}, by definition, there is a morphism \widetilde{l}:X\rightarrow Z' in \widetilde{\mathcal{L}} and a morphism \widetilde{r}:Z'\rightarrow Y in \widetilde{\mathcal{R}} such that the following diagram commutes:

Since (\mathcal{L},\widetilde{\mathcal{R}}) is an admissible weak factorization system, f and \widetilde{r} are deflations by definition. Thus we have the following commutative diagram:

Then one has that \widetilde{f_{1}} and \widetilde{f_{2}} are in \widetilde{\mathcal{R}}. Note that

is a weak pullback by the dual of [NP, Lemma 3.13]. It follows that there is a morphism \alpha:X\rightarrow P such that \widetilde{f_{1}}\alpha=\widetilde{l} and \widetilde{f_{2}}\alpha=h. Since (\widetilde{\mathcal{L}},\mathcal{R}) is an admissible weak factorization system, there is a morphism \widetilde{l'}:X\rightarrow P' in \widetilde{\mathcal{L}} and a morphism f':P'\rightarrow P in \mathcal{R}

with \alpha=f' \widetilde{l'}. Since \widetilde{l}=\widetilde{f_{1}}\alpha=\widetilde{f_{1}}f'\widetilde{l'} in \widetilde{\mathcal{L}}, we have that \widetilde{f_{1}}f' belongs to \widetilde{\mathcal{R}} by condition (CP3). It follows from condition (\mathrm{CP2}) that f'\in \widetilde{\mathcal{R}}. Consequently, h=\widetilde{f_{2}}f'\widetilde{l'} is in \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}}.

(1)\Rightarrow (2) By Theorem 3.11, (\mathrm{Infl}~\mathcal{U},\mathrm{Defl}~\widetilde{\mathcal{V}}) and (\mathrm{Infl}~\widetilde{\mathcal{U}},\mathrm{Defl}~\mathcal{V}) are two admissible weak factorization systems. Now we show that they are compatible. By [Z, Theorem 3.9] there is a thick class \mathcal{W} of \mathcal{C} such that \widetilde{\mathcal{U}}=\mathcal{U}\bigcap\mathcal{W} and \widetilde{\mathcal{V}}=\mathcal{V}\bigcap\mathcal{W}. The (CP1) holds clearly. Suppose that f:X\rightarrow Y and g:Y\rightarrow Z are in \mathrm{Defl}~(\mathcal{V}). Then g f is a deflation and then there exists a conflation

in \mathcal{C} by [HZZ1, Lemma 2.11]. Thus the condition (CP2) holds as \mathcal{W} is a thick class.

To prove (CP3), let h:A\rightarrow B be in \mathrm{Infl}~\widetilde{\mathcal{U}} and f:B\rightarrow C be in \mathrm{Defl}~\mathcal{V} such that g=f h is in \mathrm{Infl}~\widetilde{\mathcal{U}}. Then h and g are inflations with \mathrm{Cone}(h) and \mathrm{Cone}(g) in \widetilde{\mathcal{U}}. Thus by [HZZ1, Lemma 2.12], there is a conflation

Since \mathcal{W} is a thick class, it follows that \mathrm{CoCone}(f) belongs to \mathcal{W}. Thus \mathrm{CoCone}(f) is in \widetilde{\mathcal{V}}.

(2)\Rightarrow(1) By Theorem 3.11, (\mathcal{U},\widetilde{\mathcal{V}}) and (\widetilde{\mathcal{U}},\mathcal{V}) are two cotorsion pairs in \mathcal{C}. If X\in \widetilde{\mathcal{U}}, then 0\rightarrow X is in \mathrm{Infl}~\widetilde{\mathcal{U}} and hence in \mathrm{Infl}~\mathcal{U}, so X\in \mathcal{U} which yields that \widetilde{\mathcal{U}}\subseteq\mathcal{U}. Next we claim that \widetilde{\mathcal{U}}\bigcap\mathcal{V}\subseteq\mathcal{U}\bigcap \widetilde{\mathcal{V}}. Take M\in \widetilde{\mathcal{U}}\bigcap\mathcal{V}. Then 0\rightarrow M is in \mathrm{Infl}~\widetilde{\mathcal{U}} and M\rightarrow 0 is in \mathrm{Defl}~\mathcal{V}. Since (\mathrm{Infl}~\mathcal{U},\mathrm{Defl}~\widetilde{\mathcal{V}}) and (\mathrm{Infl}~\widetilde{\mathcal{U}},\mathrm{Defl}~\mathcal{V}) are compatible weak factorization systems, it follows from the Condition (CP3) that M\rightarrow 0 is in \mathrm{Defl}~\widetilde{\mathcal{V}} as 0\rightarrow 0 is in \mathrm{Infl}~\widetilde{\mathcal{U}}. Hence M\in \widetilde{\mathcal{V}}, so M\in \mathcal{U}\bigcap\widetilde{\mathcal{V}}.

We put \widetilde{\mathcal{L}}=\mathrm{Infl}~\widetilde{\mathcal{U}} and \widetilde{\mathcal{R}}=\mathrm{Defl}~~\widetilde{\mathcal{V}}. To prove \mathcal{U}\bigcap\widetilde{\mathcal{V}}\subseteq\widetilde{\mathcal{U}}\bigcap\mathcal{V}, let N be an object in \mathcal{U}\bigcap\widetilde{\mathcal{V}}. Then 0\rightarrow N is in \mathrm{Infl}~\mathcal{U} and N\rightarrow 0 is in \mathrm{Defl}~\widetilde{\mathcal{V}}. Since 0\rightarrow 0 is in \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}}, it follows from Lemma 3.20 that 0\rightarrow N is in \mathcal{W}_{\widetilde{\mathcal{L}},\widetilde{\mathcal{R}}} \bigcap\mathrm{Infl}~\mathcal{U}. Thus, by [DLLiang, Lemma 1.12], 0\rightarrow N is in \mathrm{Infl}~\widetilde{\mathcal{U}}. So N\in \widetilde{\mathcal{U}}. On the other hand, N\in \widetilde{\mathcal{V}}=\mathcal{U}^{\perp}\subseteq\widetilde{\mathcal{U}}^{\perp}=\mathcal{V}. Therefore \mathcal{U}\bigcap\widetilde{\mathcal{V}}\subseteq\widetilde{\mathcal{U}}\bigcap \mathcal{V}. We have proven that \mathcal{U}\bigcap\widetilde{\mathcal{V}}=\widetilde{\mathcal{U}}\bigcap \mathcal{V}, so (\widetilde{\mathcal{U}},\mathcal{V}) and (\mathcal{U},\widetilde{\mathcal{V}}) are compatible cotorsion pairs.