δ \delta -biderivations of Virasoro related algebras Chengkang Xu Abstract . We determine all δ \delta -biderivations for the Witt algebra, the Virasoro algebra, the W W -algebras W ⁡ ( a , b ) W(a,b) and their universal central extensions W ~ ​ ( a , b ) \widetilde{W}(a,b) , and then give some applications. 2020 Mathematics Subject Classification 17B10, 17B65, 17B68 1. Introduction The study of biderivations arises from associative ring theory. The notion of a (skew-symmetric) biderivation is a valid way to study commuting linear maps of associative rings (see [ 1 ] ). In [ 22 ] , the authors gave the definition of a biderivation for Lie algebra, and studied skew-symmetric biderivations of parabolic subalgebras of finite dimensional simple Lie algebras. Since then, the studies on biderivations of Lie algebras have attracted many attentions. Skew-symmetric biderivations were determined for the Schrödinger-Virasoro algebra [ 23 ] , generalized Witt algebras [ 5 ] , the W -algebra W(a,b) [ 9 ] , the affine-Virasoro algebra of type A_{1} [ 24 ] , Kac-Moody algebras [ 6 ] , a loop Schrödinger-Witt algebra [ 16 ] , the Witt algebra and W(2,2) [ 17 ] , a block Lie algebra \mathcal{B}(q) [ 15 ] , the twisted Heisenberg-Virasoro algebra [ 20 ] , the mirror Heisenberg-Virasoro algebra [ 14 ] , and so on. It is also worth to mention that Br \check{\text{e}} sar and Zhao in [ 4 ] introduced a new approach to study skew-symmetric biderivations of Lie algebras using the notion of centroid. Commuting linear maps were also determined for the aforementioned Lie algebras separately. Symmetric biderivations are related to commutative post-Lie algebras. There are also many studies on symmetric biderivations for Lie algebras. For example, finite-dimensional complex simple Lie algebras [ 18 ] , the Witt algebra and the W -algebra W(2,2) [ 17 ] , the block Lie algebra \mathcal{B}(q) [ 15 ] , the twisted Heisenberg-Virasoro algebra [ 20 ] , the mirror Heisenberg-Virasoro algebra [ 14 ] , the W -algebras W(a,b) [ 19 ] , higher rank Witt algebras [ 21 ] . Among the aforementioned Lie algebras, commutative post-Lie algebra structures were determined for the Witt algebra, W(2,2) , the twisted Heisenberg-Virasoro algebra, the W -algebras W(a,b) and higher rank Witt algebras. As a generalization of biderivation, the notion of \delta -biderivation was introduced in [ 25 ] , where \delta is an arbitrary complex number. This notion also generalizes the notion of \delta -derivation, which was first studied by Filippov in [ 7 ] . Especially, the \frac{1}{2} -derivation has deep connection with transposed Poisson algebra, which was introduced in [ 2 ] as a dual notion of the Poisson algebra. The authors in [ 25 ] also determined all \frac{1}{2} -derivations of the twisted Heisenberg-Virasoro algebra, the Schrödinger-Virasoro algebra, the extended Schrödinger-Virasoro algebra and the twisted Schrödinger-Virasoro algebra, and used them to determine transposed Poisson algebra structures on these algebras. However, they did not go deeper into computing \delta -biderivations of Lie algebras. Transposed Poisson algebra structures over some Lie algebras were determined recently, see [ 10 , 11 , 13 , 26 ] and references therein. The studies on transposed Poisson algebra structures over Lie algebras in the aforementioned references are all by the means of \frac{1}{2} -derivations. But the connection between \frac{1}{2} -derivations and transposed Poisson algebra structures on Lie algebras is not direct, but actually through \frac{1}{2} -biderivations. Therefore, we introduce a notion, called transposed \delta -Poisson algebra structure, for Lie algebras, as a generalization of transposed Poisson algebra structure in the last section, and also as an application of \delta -biderivations. In this paper, we determine all \delta -biderivations for the Witt algebra, the Virasoro algebra, the W -algebras W(a,b) and their universal central extensions \widetilde{W}(a,b) . Let us now briefly describe the organization of this paper. In Section 2, we recall the definitions of these algebras, and the notions of \delta -derivation and \delta -biderivation. In Section 3, we present some basic properties of \delta -biderivations. Section 4 is dedicated to determine \delta -biderivations of the Witt algebra and the Virasoro algebra, and Section 5 is to determine \delta -biderivations of the algebras W(a,b) and \widetilde{W}(a,b) for any complex numbers a,b . In the last section we provide some applications of \delta -biderivations to determine commuting linear maps, commutative post-Lie algebra structures and transposed \delta -Poisson algebra structures on the algebras mentioned above. Throughout this paper, the symbols \mathbb{Z} and \mathbb{C} are to denote the sets of integers and complex numbers respectively. All vector spaces are assumed to be over \mathbb{C} , and for any Lie algebra \mathfrak{g} we denote by \mathfrak{g}^{\prime} the derived subalgebra of \mathfrak{g} . Let G be an additive group and V be a G -graded space (or Lie algebra) such that V=\bigoplus_{g\in G}V_{g} . We call V_{g} a homogeneous subspace of degree g , and call nonzero elements in V_{g} are of degree g . 2. Preliminaries In this section, we recall the Lie algebras we study in this paper, the notions of \delta -derivation and \delta -biderivation, and some basic results. Definition 2.1 . The Witt algebra is a complex Lie algebra \mathcal{W} with a basis \left\{L_{n}\mid n\in\mathbb{Z}\right\} subjecting to the commutation relation [L_{m},L_{n}]=(m-n)L_{m+n}\quad\text{ for any }m,n\in\mathbb{Z}. The Virasoro algebra \mathcal{V} is the universal central extension of \mathcal{W} . Equivalently, \mathcal{V} has a basis \left\{L_{n},C_{0}\mid n\in\mathbb{Z}\right\} with commutation relation [L_{m},L_{n}]=(m-n)L_{m+n}+\frac{1}{12}(m^{3}-m)C_{0}\delta_{m+n,0}\quad\text{ for any }m,n\in\mathbb{Z}. The W -algebras are tensor products of the Witt algebra and its modules of intermediate series. Definition 2.2 . Let a,b\in\mathbb{C} , the W -algebra W(a,b) is a Lie algebra with a \mathbb{C} -basis \{L_{n},I_{n}\mid n\in\mathbb{Z}\} satisfying Lie brackets [L_{i},L_{j}]=(i-j)L_{i+j},\quad[L_{i},I_{j}]=-(a+j+bi)I_{i+j},\quad[I_{i},I_{j}]=0\quad\text{for }i,j\in\mathbb{Z}. Since W(a,b)\cong W(a+k,b) for any k\in\mathbb{Z} , we always assume 0\leq a<1 in this paper. Note that W(a,b) is perfect if and only if (a,b)\neq(0,1) . The universal central extension \widetilde{W}(a,b) of W(a,b) was determined in [ 8 ] . Theorem 2.3 . The algebra \widetilde{W}(a,b) satisfies the following Lie brackets \displaystyle[L_{m},L_{n}]=(m-n)L_{m+n}+\frac{m^{3}-m}{12}C_{0}\delta_{m+n,0}, \displaystyle[L_{m},I_{n}]=-(a+n+bm)I_{m+n}+\delta_{m+n,0}\delta_{a,0}\left((m^{2}+m)C_{1}^{0}\delta_{b,0}+(mC_{1}^{1}+C_{2}^{1})\delta_{b,1}+\frac{m^{3}-m}{12}C_{1}^{-1}\delta_{b,-1}\right), \displaystyle[I_{m},I_{n}]=mC_{2}^{0}\delta_{m+n,0}\delta_{a,0}\delta_{b,0}+(2m+1)C_{2}^{\frac{1}{2},0}\delta_{m+n+1,0}\delta_{a,\frac{1}{2}}\delta_{b,0}, where C_{0},C_{1}^{0},C_{1}^{1},C_{2}^{1},C_{1}^{-1},C_{2}^{0},C_{2}^{\frac{1}{2},0} are central elements. Denote by \mathfrak{C} the center of \widetilde{W}(a,b) . Note that \widetilde{W}(a,b) is perfect unless (a,b)=(0,1) . Definition 2.4 . Let \delta\in\mathbb{C} and let \mathcal{L} be a Lie algebra. A linear map f:\mathcal{L}\longrightarrow\mathcal{L} is called a \delta -derivation if \delta[f(x),y]+\delta[x,f(y)]=f([x,y])\quad\text{ for any }x,y\in\mathcal{L}. Clearly, when \delta=1 , the notion of \delta -derivation is the ordinary derivation. Denote by \mathfrak{Der}_{\delta}(\mathcal{L}) the space of \delta -derivations of \mathcal{L} for a fixed \delta . For simplicity we write \mathfrak{Der}(\mathcal{L})=\mathfrak{Der}_{1}(\mathcal{L}) . All \delta -derivations of the Witt algebra \mathcal{W} and the Virasoro algebra \mathcal{V} were determined in [ 12 ] . Theorem 2.5 ( [ 12 ] ) . (1) \mathfrak{Der}(\mathcal{W})=\mathrm{span}_{\mathbb{C}}\left\{\mathrm{ad}L_{j}\mid j\in\mathbb{Z}\right\} . (2) \mathcal{D}er_{\frac{1}{2}}(\mathcal{W})=\mathrm{span}_{\mathbb{C}}\left\{\varphi_{j}\mid j\in\mathbb{Z}\right\} , where \varphi_{j}(L_{i})=L_{i+j} . (3) \mathfrak{Der}_{\delta}(\mathcal{W})=0 if \delta\neq\frac{1}{2},1 . Now we recall the notion of \delta -biderivation from [ 25 ] . Definition 2.6 . Let \delta\in\mathbb{C} and let \mathcal{L} be a Lie algebra. A bilinear map f:\mathcal{L}\times\mathcal{L}\longrightarrow\mathcal{L} is called a \delta -biderivation of \mathcal{L} if \displaystyle\delta[x,f(y,z)]+\delta[f(x,z),y]=f([x,y],z), (2.1) \displaystyle\delta[f(x,y),z]+\delta[y,f(x,z)]=f(x,[y,z]) (2.2) for any x,y,z\in\mathcal{L} . When \delta=1 , we simply call f a biderivation . A \delta -biderivation of \mathcal{L} is called central if f(x,y)\in Z(\mathcal{L}) the center of \mathcal{L} , called symmetric if f(x,y)=f(y,x) , and called skew-symmetric if f(x,y)=-f(y,x) for all x,y\in\mathcal{L} . Denote by \mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right) the space of \delta -biderivations of \mathcal{L} for a fixed \delta . Let f be a \delta -biderivation of \mathcal{L} , then for any x\in\mathcal{L} , the linear maps f(x,\cdot) and f(\cdot,x) are \delta -derivations of \mathcal{L} . 3. Basic properties of \delta -biderivations In this section we give some basic properties of \delta -biderivations over an arbitrary Lie algebra. Proposition 3.1 . Let \mathcal{L} be a Lie algebra with center Z(\mathcal{L}) , and f a \delta -biderivation of \mathcal{L} . (1) If \delta\neq 0 , then f(\alpha,y)=f(y,\alpha)\in Z(\mathcal{L}) for any \alpha\in Z(\mathcal{L}),y\in\mathcal{L} . (2) f(\alpha,y)=f(y,\alpha)=0 for any \alpha\in Z(\mathcal{L}),y\in[\mathcal{L},\mathcal{L}] . (3) If \mathcal{L} is perfect, then f(\alpha,y)=f(y,\alpha)=0 for any \alpha\in Z(\mathcal{L}),y\in\mathcal{L} . Proof. (1) For any x\in\mathcal{L} , \phi_{x}=f(x,\cdot) is a \delta -derivation of \mathcal{L} . Then for any y\in\mathcal{L} , we have 0=\phi_{x}([\alpha,y])=\delta[\phi_{x}(\alpha),y]+\delta[\alpha,\phi_{x}(y)]=\delta[\phi_{x}(\alpha),y], which gives f(x,\alpha)=\phi_{x}(\alpha)\in Z(\mathcal{L}) if \delta\neq 0 . Similarly, one gets f(\alpha,x)\in Z(\mathcal{L}) . (2) Since f is bilinear, we may assume y=[z,z^{\prime}] for some z,z^{\prime}\in\mathcal{L}. Then we have f(y,\alpha)=f([z,z^{\prime}],\alpha)=\delta[z,f(z^{\prime},\alpha)]+\delta[f(z,\alpha),z^{\prime}]=0. Similarly, one gets f(\alpha,y)=0 . (3) follows from (2). ∎ Proposition 3.2 . Any central \delta -biderivation of a perfect Lie algebra is a zero map. Proof. Let \mathcal{L} be a perfect Lie algebra and f a central \delta -biderivation of \mathcal{L} . For any x,y\in\mathcal{L} , write y=[z,z^{\prime}] for some z,z^{\prime}\in\mathcal{L}. By ( 2.2 ), we have f(x,y)=f(x,[z,z^{\prime}])=\delta[z,f(x,z^{\prime})]+\delta[f(x,z^{\prime}),z]=0 . So f=0 . ∎ Proposition 3.3 . Suppose G is an abelian group and \mathcal{L} is a finitely generated G -graded Lie algebra such that \mathcal{L}=\bigoplus_{g\in G}\mathcal{L}_{g} . Then \mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right) is also G -graded, \mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right)=\bigoplus_{g\in G}\mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right)_{g}, where \mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right)_{g}=\{\phi\in\mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right)\mid\phi(x,y)\in\mathcal{L}_{g+a+b}\text{ for any }a,b\in G,x\in\mathcal{L}_{a},y\in\mathcal{L}_{b}\} . We call elements in \mathfrak{BD}^{[\delta]}\left(\mathcal{L}\right)_{g} are of degree g . Proof. The proof is similar to that of Lemma 2.1 in [ 14 ] , and we omit the details. ∎ Proposition 3.4 . Let \phi be a \delta -biderivation of a Lie algebra \mathcal{L} with center Z(\mathcal{L}) . Define \overline{\phi}(\overline{x},\overline{y})=\overline{\phi(x,y)}=\phi(x,y)+Z(\mathcal{L}), for all x,y\in\mathcal{L} , where \overline{x}=x+Z(\mathcal{L})\in\mathcal{L}/Z(\mathcal{L}) . Then \overline{\phi} is a \delta -biderivation of \mathcal{L}/Z(\mathcal{L}) . Proof. The verification is easy by using Proposition 3.1 , and we omit the details. ∎ 4. \delta -biderivations of the Witt and Virasoro algebras In this section we determine all \delta -biderivations of the Witt algebra \mathcal{W} and the Virasoro algebra \mathcal{V} in two subsections. 4.1. \delta -biderivations of the Witt algebra Clearly the bilinear map \pi:\mathcal{W}\times\mathcal{W}\longrightarrow\mathcal{W} defined by \pi(L_{i},L_{j})=[L_{i},L_{j}]\quad\text{ for all }i,j\in\mathbb{Z}, is a (1-)biderivation, and for any n\in\mathbb{Z} the bilinear map \theta_{n}:\mathcal{W}\times\mathcal{W}\longrightarrow\mathcal{W} defined by \theta_{n}(L_{i},L_{j})=L_{n+i+j}\quad\text{ for all }i,j\in\mathbb{Z}, is a \frac{1}{2} -biderivation. Theorem 4.1 . The space of all \delta -biderivations of the Witt algebra \mathcal{W} is \mathfrak{BD}^{[\delta]}\left(\mathcal{W}\right)=\begin{cases}\mathbb{C}\,\pi&\text{ if }\delta=1;\\ \mathrm{span}_{\mathbb{C}}\left\{\theta_{n}\mid n\in\mathbb{Z}\right\}&\text{ if }\delta=\frac{1}{2};\\ 0&\text{ otherwise}.\end{cases} Proof. Let f be a \delta -biderivation of \mathcal{W} . Then for any i\in\mathbb{Z} , the linear map f(L_{i},\cdot) and f(\cdot,L_{i}) are \delta -derivations of \mathcal{W} . Moreover, we may assume f\in\mathfrak{BD}^{[\delta]}\left(\mathcal{W}\right)_{n}=\{g\in\mathfrak{BD}^{[\delta]}\left(\mathcal{W}\right)\mid g(L_{i},L_{j})\in\mathbb{C}L_{n+i+j}\text{ for any }i,j\in\mathbb{Z}\} for some n\in\mathbb{Z} by Proposition 3.3 . If \delta\neq\frac{1}{2},1 , then by Theorem 2.5 (3) we have f(L_{i},\cdot)=0 , i.e., f(L_{i},L_{j})=0 for any i,j\in\mathbb{Z} . Hence f=0 . If \delta=1 , then for any i\in\mathbb{Z} the linear map f(L_{i},\cdot) is a derivation of degree n+i , hence f(L_{i},\cdot)=\lambda\mathrm{ad}L_{n+i} for some \lambda\in\mathbb{C} by Theorem 2.5 (1). So we have f(L_{i},L_{j})=\lambda[L_{n+i},L_{j}]=\lambda(n+i-j)L_{n+i+j} . Similarly, since f(\cdot,L_{j}) is a derivation of degree n+j , we have f(L_{i},L_{j})=\mu(n-i+j)L_{n+i+j} for some \mu\in\mathbb{C} . So \lambda(n+i-j)=\mu(n-i+j)\quad\text{ for any }i,j\in\mathbb{Z}. Let j=n+i and we get n\mu=0 . If n\neq 0 , then \mu=0 , and further we have \lambda=0 . So f=0 . If n=0 , then it is clear that \mu=-\lambda and hence f=\lambda\pi . This proves that \mathcal{BD}^{[1]}(\mathcal{W})=\mathbb{C}\,\pi. If \delta=\frac{1}{2} , then for any i,j\in\mathbb{Z} we have by Theorem 2.5 (2) that f(L_{i},\cdot)=a_{i}\varphi_{n+i},\quad f(\cdot,L_{j})=b_{j}\varphi_{n+j}, where a_{i},b_{j}\in\mathbb{C} . Hence a_{i}L_{n+i+j}=f(L_{i},L_{j})=b_{j}L_{n+i+j} , which further implies that a_{i}=b_{j} for all i,j\in\mathbb{Z} . So f is a multiple of \theta_{n} . This proves the case \delta=\frac{1}{2} . ∎ 4.2. \delta -biderivations of the Virasoro algebra In this subsection we determine all \delta -biderivations of the Virasoro algebra \mathcal{V} . It is easy to check that the bilinear map \widetilde{\pi}:\mathcal{V}\times\mathcal{V}\longrightarrow\mathcal{V} defined by \widetilde{\pi}(L_{i},L_{j})=[L_{i},L_{j}]=(i-j)L_{i+j}+\frac{1}{12}(i^{3}-i)C_{0}\delta_{i+j,0}\quad\text{ for all }i,j\in\mathbb{Z}, is a (1-)biderivation of degree 0 . Theorem 4.2 . The space of all \delta -biderivations of the Virasoro algebra \mathcal{V} is \mathfrak{BD}^{[\delta]}\left(\mathcal{V}\right)=\begin{cases}\mathbb{C}\,\widetilde{\pi}&\text{ if }\delta=1;\\ 0&\text{ otherwise}.\end{cases} Proof. Let f be a \delta -biderivation of \mathcal{V} and \overline{f} the \delta -biderivation of the quotient algebra \mathcal{V}/\mathbb{C}C_{0} defined by Proposition 3.4 . Since \mathcal{V} is perfect, we have f(C_{0},\mathcal{V})=f(\mathcal{V},C_{0})=0 by Proposition 3.1 (3). Note that \mathcal{V}/\mathbb{C}C_{0}\cong\mathcal{W} . If \delta\neq\frac{1}{2},1 , then \overline{f}=0 by Theorem 4.1 , i.e., f is a central \delta -biderivation of \mathcal{V} . So f=0 by Proposition 3.2 . If \delta=1 , then \overline{f}=\lambda\,\pi for some \lambda\in\mathbb{C} by Theorem 4.1 . So f_{1}=f-\lambda\widetilde{\pi} is a central biderivation of \mathcal{V} . Then f_{1}=0 by Proposition 3.2 , and hence f=\lambda\,\widetilde{\pi} . For the case \delta=\frac{1}{2} , we may further assume that f is of degree n\in\mathbb{Z} by Proposition 3.3 . Hence \overline{f}=a_{n}\theta_{n} for some a_{n}\in\mathbb{C} by Theorem 4.1 . So for i,j\in\mathbb{Z} we may write f(L_{i},L_{j})=a_{n}L_{n+i+j}+C_{i,j}C_{0}\quad\text{for some }C_{i,j}\in\mathbb{C}. For any i\in\mathbb{Z}\setminus\{0\} , apply x=L_{i},y=L_{-i},z=L_{-n} to ( 2.1 ) and we get a_{n}(i^{2}-1)-24C_{0,-n}=0, which implies a_{n}=0 . So f is a central \frac{1}{2} -biderivation of \mathcal{V} . Hence f=0 by Proposition 3.2 . ∎ Remark 4.3 . (1) It is well known that a nontrivial derivation of a perfect Lie algebra \mathcal{L} may be uniquely extended to a nontrivial derivation of the universal central extension of \mathcal{L} (see [ 3 ] ). Such a extension may not exist for \frac{1}{2} -biderivation. From Theorem 4.1 and Theorem 4.2 we see that the \frac{1}{2} -biderivation \theta_{n} of \mathcal{W} has no nontrivial extension to the Virsaoro algebra. (2) One could also prove Theorem 4.2 in the same way as in the proof of Theorem 4.1 using \delta -derivations of \mathcal{V} , which were given in [ 12 ] . 5. \delta -biderivations of the W -algebras In this section we determine all \delta -biderivations of the W -algebras W(a,b) and \widetilde{W}(a,b) for arbitrary \delta,a,b\in\mathbb{C} . All \delta -derivations of W(a,b) were given in [ 12 ] . Proposition 5.1 . (1) The space of (1-)derivations of W(a,b) is \mathfrak{Der}(W(a,b))=\begin{cases}\mathrm{Inn}(W(a,b))\oplus\mathbb{C}D_{1}\oplus\mathbb{C}D_{2}^{0}\oplus\mathbb{C}D_{3}&\text{if }(a,b)=(0,0),\\ \mathrm{Inn}(W(a,b))\oplus\mathbb{C}D_{1}\oplus\mathbb{C}D_{2}^{k}&\text{if }(a,b)=(0,k),k=1,2,\\ \mathrm{Inn}(W(a,b))\oplus\mathbb{C}D_{1}&\text{otherwise},\end{cases} where the derivations D_{1},D_{2}^{k},k=0,1,2 , and D_{3} are defined by \displaystyle D_{1}(L_{i})=0,\quad D_{1}(I_{i})=I_{i}; \displaystyle D_{2}^{k}(L_{i})=i^{k+1}I_{i},\quad D_{2}^{k}(I_{i})=0; \displaystyle D_{3}(L_{i})=I_{i},\quad D_{3}(I_{i})=0. (2) The space of \frac{1}{2} -derivations of W(a,b) is \mathfrak{Der}_{\frac{1}{2}}(W(a,b))=\begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\varphi_{j},\psi_{j}\mid j\in\mathbb{Z}\right\}&\text{if }b=-1,\\ \mathbb{C}\,\varphi_{0}&\text{if }b\neq-1,\end{cases} where the \frac{1}{2} -derivations \varphi_{j} and \psi_{j} are defined by \varphi_{j}(L_{i})=L_{i+j},\ \varphi_{j}(I_{i})=I_{i+j};\quad\quad\psi_{j}(L_{i})=I_{i+j},\ \psi_{j}(I_{i})=0. (3) For \delta\neq\frac{1}{2},1 , the space of \delta -derivations of W(a,b) is 0. 5.1. \delta -biderivations of W(a,b) In this subsection we determine the space \mathfrak{BD}^{[\delta]}\left(W(a,b)\right) of all \delta -biderivations of the W -algebra W(a,b) . Theorem 5.2 . (1) If \delta\neq\frac{1}{2},1 , then \mathfrak{BD}^{[\delta]}\left(W(a,b)\right)=0 . (2) For \delta=1 , we have \mathfrak{BD}^{[1]}(W(a,b))=\begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\pi,\Psi_{n}^{(a,0)}\mid n\in\mathbb{Z}\right\}&\text{ if }b=0;\\ \mathrm{span}_{\mathbb{C}}\left\{\pi,\Psi_{n}^{(a,1)}\mid n\in\mathbb{Z}\right\}&\text{ if }b=1;\\ \mathrm{span}_{\mathbb{C}}\left\{\pi,\Theta^{(0,-1)}\right\}&\text{ if }a=0,b=-1;\\ \mathbb{C}\,\pi&\text{ otherwise},\end{cases} where \pi(x,y)=[x,y],x,y\in W(a,b) , and the biderivations \Psi_{n}^{(a,0)},\Psi_{n}^{(a,1)},\Theta^{(0,-1)} are defined by \displaystyle\Psi_{n}^{(a,0)}(L_{i},L_{j})=I_{i+j+n},\quad\Psi_{n}^{(a,0)}(L_{i},I_{j})=\Psi_{n}^{(a,0)}(I_{i},L_{j})=\Psi_{n}^{(a,0)}(I_{i},I_{j})=0; \displaystyle\Psi_{n}^{(a,1)}(L_{i},L_{j})=(a+i+j+n)I_{i+j+n},\quad\Psi_{n}^{(a,1)}(L_{i},I_{j})=\Psi_{n}^{(a,1)}(I_{i},L_{j})=\Psi_{n}^{(a,1)}(I_{i},I_{j})=0; \displaystyle\Theta^{(0,-1)}(L_{i},L_{j})=(i-j)I_{i+j},\quad\Theta^{(0,-1)}(L_{i},I_{j})=\Theta^{(0,-1)}(I_{i},L_{j})=\Theta^{(0,-1)}(I_{i},I_{j})=0. (3) For \delta=\frac{1}{2} , we have \mathfrak{BD}^{[\frac{1}{2}]}(W(a,b))=\begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\Phi_{n},\Psi_{n}\mid n\in\mathbb{Z}\right\}&\text{ if }b=-1;\\ 0&\text{ otherwise},\end{cases} where \Phi_{n},\Psi_{n} are the bilinear maps defined by \displaystyle\Phi_{n}(L_{i},L_{j})=L_{n+i+j},\quad\Phi_{n}(L_{i},I_{j})=\Phi_{n}(I_{i},L_{j})=I_{n+i+j},\quad\Phi_{n}(I_{i},I_{j})=0; \displaystyle\Psi_{n}(L_{i},L_{j})=I_{n+i+j},\quad\Psi_{n}(L_{i},I_{j})=\Psi_{n}(I_{i},L_{j})=\Psi_{n}(I_{i},I_{j})=0 Proof. Let f be a \delta -biderivation of W(a,b) . Since for any x\in W(a,b) , f(x,\cdot) is a \delta -derivation of W(a,b) , statement (1) follows from Proposition 5.1 (3). Statement (2) is Theorem 3.2 in [ 19 ] . (3) If b\neq-1 , then by Proposition 5.1 (2), for any x,y\in W(a,b) , we have f(x,\cdot)=\lambda\varphi_{0} and f(\cdot,y)=\mu\varphi_{0} for some \lambda,\mu\in\mathbb{C} . So f(x,y)=\lambda y=\mu x , which forces \lambda=\mu=0 and f=0 . Now let b=-1 and we may assume that f is of some fixed degree n\in\mathbb{Z} by Proposition 3.3 . Then by Proposition 5.1 (2) we may write for i\in\mathbb{Z} that \displaystyle f(L_{i},\cdot)=a_{i,n}\varphi_{n+i}+b_{i,n}\psi_{n+i},\quad f(\cdot,L_{i})=a^{\prime}_{i,n}\varphi_{n+i}+b^{\prime}_{i,n}\psi_{n+i}, \displaystyle f(I_{i},\cdot)=c_{i,n}\varphi_{n+i}+d_{i,n}\psi_{n+i},\quad f(\cdot,I_{i})=c^{\prime}_{i,n}\varphi_{n+i}+d^{\prime}_{i,n}\psi_{n+i}, where a_{i,n},b_{i,n},c_{i,n},d_{i,n},a^{\prime}_{i,n},b^{\prime}_{i,n},c^{\prime}_{i,n},d^{\prime}_{i,n}\in\mathbb{C} . Then we have f(L_{i},L_{j})=a_{i,n}L_{n+i+j}+b_{i,n}I_{n+i+j}=a^{\prime}_{j,n}L_{n+i+j}+b^{\prime}_{j,n}I_{n+i+j}, which implies that a^{\prime}_{j,n}=a_{i,n},b^{\prime}_{j,n}=b_{i,n} for any i,j\in\mathbb{Z} . Set a_{n}=a_{i,n},b_{n}=b_{i,n} and we have f(L_{i},L_{j})=a_{n}L_{n+i+j}+b_{n}I_{n+i+j}. (5.1) Write f(L_{i},I_{j}) in two ways, f(L_{i},I_{j})=a_{n}I_{n+i+j}=c^{\prime}_{j,n}L_{n+i+j}+d^{\prime}_{j,n}I_{n+i+j}, (5.2) which gives c^{\prime}_{j,n}=0,d^{\prime}_{j,n}=a_{n} for any j\in\mathbb{Z} . Similarly, write f(I_{i},L_{j}) in two ways, and we get c_{j,n}=0,d_{j,n}=a_{n} for any j\in\mathbb{Z} . Hence f(I_{i},L_{j})=a_{n}I_{n+i+j}. (5.3) Then f(\cdot,I_{i})=a_{n}\psi_{n+i} and f(I_{j},I_{i})=0 . Together with ( 5.1 ), ( 5.2 ) and ( 5.3 ), we see that f=a_{n}\Phi_{n}+b_{n}\Psi_{n} . This complete the proof. ∎ 5.2. \delta -biderivations of \widetilde{W}(a,b) In this subsection we determine all \delta -biderivations of the algebra \widetilde{W}(a,b) . The algebra \widetilde{W}(0,1) has three independent central \delta -biderivations \mathcal{A},\mathcal{B},\mathcal{C} for any \delta\in\mathbb{C} , which are defined by \mathcal{A}(I_{0},I_{0})=C_{0},\quad\mathcal{B}(I_{0},I_{0})=C_{1}^{1},\quad\mathcal{C}(I_{0},I_{0})=C_{2}^{1}, and all other values being zero. Theorem 5.3 . (1) If \delta\neq\frac{1}{2} or 1, then \mathfrak{BD}^{[\delta]}\left(\widetilde{W}(a,b)\right)=\begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\mathcal{A},\mathcal{B},\mathcal{C}\right\}&\text{if }(a,b)=(0,1);\\ \quad\quad 0&\text{otherwise}.\end{cases} (2) If \delta=1 , then \mathfrak{BD}^{[1]}\left(\widetilde{W}(a,b)\right)=\begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\widetilde{\pi},\widetilde{\Theta}^{(0,-1)}\right\}&\text{if }a=0,b=-1;\\ \mathrm{span}_{\mathbb{C}}\left\{\widetilde{\pi},\mathcal{A},\mathcal{B},\mathcal{C}\right\}&\text{if }a=0,b=1;\\ \mathbb{C}\,\widetilde{\pi}&\text{otherwise},\end{cases} where \widetilde{\pi}(x,y)=[x,y],x,y\in\widetilde{W}(a,b) and \widetilde{\Theta}^{(0,-1)} is the bilinear map on \widetilde{W}(0,-1) with only nonvanishing values \widetilde{\Theta}^{(0,-1)}(L_{i},L_{j})=(i-j)I_{i+j}+\frac{i^{3}-i}{12}C_{1}^{-1}\delta_{i+j,0}\delta_{a,0}. (3) If \delta=\frac{1}{2} , then \mathfrak{BD}^{[\frac{1}{2}]}\left(\widetilde{W}(a,b)\right)=\begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\widetilde{\Psi}_{n}\mid n\in\mathbb{Z}\right\}&\text{ if }0-a-p-1 , then we have by Theorem 5.2 (2) \sum_{n=q}^{p}\mu_{n}(a+n+k+1)I_{n+k}=f(L_{1},L_{k})=[L_{1},L_{0}]\circ L_{k}=f\left(L_{1},f(L_{0},L_{k})\right)-f\left(L_{0},f(L_{1},L_{k})\right)=0, which implies all \mu_{n}=0 . So f=0 . Hence any commutative post-Lie algebra structures on W(a,1) is trivial. If \mathcal{L}=\widetilde{W}(0,1) , then we have f(x,y)\in\mathfrak{C} for any x,y\in\widetilde{W}(0,1) , and one may easily check that for any a,b,c\in\mathbb{C} , \Lambda_{a,b,c} actually defines a commutative post-Lie algebra structure on \widetilde{W}(0,1) . This proves the theorem. ∎ 6.3. Transposed \delta -Poisson algebra structure In this subsection we determine all transposed \delta -Poisson algebra structures on the algebra \mathcal{L} . In this subsection we always assume that \delta is a nonzero complex number. Let us first recall the definition of transposed \delta -Poisson algebra. Definition 6.4 . A transposed \delta -Poisson algebra is a vector space \mathcal{P} equipped with two bilinear operations \circ and [\cdot,\cdot] such that (\mathcal{P},\circ) is a commutative associative algebra, (\mathcal{P},[\cdot,\cdot]) is a Lie algebra and these two operations satisfy the following compatibility condition \delta z\circ[x,y]=[z\circ x,y]+[x,z\circ y]\quad\text{ for any }x,y,z\in\mathcal{P}. A transposed \delta -Poisson algebra structure on a Lie algebra \mathfrak{g} is a bilinear operation on \mathfrak{g} making \mathfrak{g} a transposed \delta -Poisson algebra. A transposed \delta -Poisson algebra structure \circ on \mathfrak{g} is called trivial if x\circ y=0 for any x,y\in\mathfrak{g} . Let (\mathfrak{g},\circ,[\cdot,\cdot]) be a transposed \delta -Poisson algebra. It is easy to check that the bilinear map f(x,y)=x\circ y on \mathfrak{g} is a symmetric \frac{1}{\delta} -biderivation. This fact enables us to classify all transposed \delta -Poisson algebra structures on the algebra \mathcal{L} . Theorem 6.5 . (1) Any transposed 2 -Poisson algebra structure on \mathcal{W} is of the form L_{i}\circ L_{j}=\sum_{n\in\mathbb{Z}}\lambda_{n}L_{n+i+j} for some \lambda_{n}\in\mathbb{C} . For \delta\neq 2,0 , any transposed \delta -Poisson algebra structure on \mathcal{W} is trivial. (2) Any transposed 1 -Poisson algebra structure on W(a,0) is of the form \displaystyle L_{i}\circ L_{j}=\sum_{n\in\mathbb{Z}}\lambda_{n}I_{n+i+j},\quad L_{i}\circ I_{j}=I_{i}\circ I_{j}=0 for some \lambda_{n}\in\mathbb{C} . For \delta\neq 1,0 , any transposed \delta -Poisson algebra structure on W(a,0) is trivial. (3) Any transposed 1 -Poisson algebra structure on W(a,1) is of the form \displaystyle L_{i}\circ L_{j}=\sum_{n\in\mathbb{Z}}(a+n+i+j)\lambda_{n}I_{n+i+j},\quad L_{i}\circ I_{j}=I_{i}\circ I_{j}=0 for some \lambda_{n}\in\mathbb{C} . For \delta\neq 1,0 , any transposed \delta -Poisson algebra structure on W(a,1) is trivial. (4) Any transposed 2 -Poisson algebra structure on W(a,-1) is of the form \displaystyle L_{i}\circ L_{j}=\sum_{n\in\mathbb{Z}}\lambda_{n}L_{n+i+j}+\sum_{n\in\mathbb{Z}}\mu_{n}I_{n+i+j},\quad L_{i}\circ I_{j}=\sum_{n\in\mathbb{Z}}\lambda_{n}I_{n+i+j},\quad I_{i}\circ I_{j}=0 for some \lambda_{n},\mu_{n}\in\mathbb{C} . For \delta\neq 2,0 , any transposed \delta -Poisson algebra structure on W(a,-1) is trivial. (5) Let a\notin\mathbb{Z} . Any transposed 2 -Poisson algebra structure on \widetilde{W}(a,-1) , is of the form L_{i}\circ L_{j}=\sum_{n\in\mathbb{Z}}\mu_{n}I_{n+i+j},\quad L_{i}\circ I_{j}=I_{i}\circ I_{j}=\mathfrak{C}\circ\widetilde{W}(a,-1)=0 for some \mu_{n}\in\mathbb{C} . For \delta\neq 2,0 , any transposed \delta -Poisson algebra structure on \widetilde{W}(a,-1) is trivial. (6) Any transposed 2 -Poisson algebra structure on \widetilde{W}(0,1) is of the form (\lambda I_{0}+x)\circ(\mu I_{0}+y)=\lambda\mu\left(aC_{0}+bC_{1}^{1}+cC_{2}^{1}\right),\quad\lambda,\mu\in\mathbb{C},x,y\in\widetilde{W}(0,1)^{\prime}, for some a,b,c\in\mathbb{C} . For \delta\neq 2,0 , any transposed \delta -Poisson algebra structure on \widetilde{W}(0,1) is trivial. (7) Let \mathcal{L} be any of the Virasoro algebra \mathcal{V} , the W -algebras W(a,b) or \widetilde{W}(a,b) except the ones listed in (2), (3), (4), (5) and (6). Then any transposed \delta -Poisson algebra structure on \mathcal{L} is trivial. Proof. By Theorem 4.1 , Theorem 4.2 , Theorem 5.2 and Theorem 5.3 , the space of all symmetric \frac{1}{\delta} -biderivations of \mathcal{L} equals to \begin{cases}\mathrm{span}_{\mathbb{C}}\left\{\theta_{n}\mid n\in\mathbb{Z}\right\}&\quad\text{if }\mathcal{L}=\mathcal{W},\delta=2;\\ \mathrm{span}_{\mathbb{C}}\left\{\Psi_{n}^{(a,0)}\mid n\in\mathbb{Z}\right\}&\quad\text{if }\mathcal{L}=W(a,0),\delta=1;\\ \mathrm{span}_{\mathbb{C}}\left\{\Psi_{n}^{(a,1)}\mid n\in\mathbb{Z}\right\}&\quad\text{if }\mathcal{L}=W(a,1),\delta=1;\\ \mathrm{span}_{\mathbb{C}}\left\{\Phi_{n},\Psi_{n}\mid n\in\mathbb{Z}\right\}&\quad\text{if }\mathcal{L}=W(a,-1),\delta=2;\\ \mathrm{span}_{\mathbb{C}}\left\{\widetilde{\Psi}_{n}\mid n\in\mathbb{Z}\right\}&\quad\text{if }\mathcal{L}=\widetilde{W}(a,-1)\text{ and }a\notin\mathbb{Z},\delta=2;\\ \mathrm{span}_{\mathbb{C}}\left\{\mathcal{A},\mathcal{B},\mathcal{C}\right\}&\quad\text{if }\mathcal{L}=\widetilde{W}(0,1),\delta=2;\\ 0&\quad\text{otherwise.}\end{cases} Then the theorem follows. ∎ Acknowledgments: C. Xu is supported by the National Natural Science Foundation of China (No. 12261077).