# Difference Between Algebra of Infinitesimal Conformal Transformations & Conformal Algebra

in Blumenhagen Book on conformal field theory, It is mentioned that the algebra of infinitesimal conformal transformation is different from the conformal algebra and on page 11, conformal algebra is defined by a redefinition of generators of infinitesimal conformal transformation. I have three question about this :

1. How this is possible that by a redefinition of generators, one could obtain a sub-algebra of an algebra? in this case one obtain conformal algebra as a sub-algebra of algebra of generators of infinitesimal conformal transformations?

2. Does this is related to «special conformal transformation» which is not globally defined?

3. How are these related to topological properties of conformal group?

Any comment or reference would greatly be appreciated!

Let \begin{align} \overline{\mathbb{R}^{p,q}}~~:=&~~\left\{y\in \mathbb{R}^{p+1,q+1}\backslash\{0\}\mid \eta^{p+1,q+1}(y,y)=0\right\}/\mathbb{R}^{\times} \cr ~~\subseteq &~~ \mathbb{P}_{p+q+1}(\mathbb{R})~~\equiv~~(\mathbb{R}^{p+1,q+1}\backslash\{0\})/\mathbb{R}^{\times}, \cr &\mathbb{R}^{\times}~~\equiv~~\mathbb{R}\backslash\{0\}, \end{align}\tag{1} denote the conformal compactification of $$\mathbb{R}^{p,q}$$. Topologically, $$\overline{\mathbb{R}^{p,q}}~\cong~(\mathbb{S}^p\times \mathbb{S}^q)/\mathbb{Z}_2 .\tag{2}$$ The $$\mathbb{Z}_2$$-action in eq. (2) identifies points related via a simultaneous antipode-swap on the spatial and the temporal sphere. The embedding $$\imath: \mathbb{R}^{p,q}\hookrightarrow \overline{\mathbb{R}^{p,q}}$$ is given by $$\imath(x)~:=~\left(1-\eta^{p,q}(x,x): ~2x:~ 1+\eta^{p,q}(x,x)\right). \tag{3}$$ Let $$n:=p+q$$. [If $$n=1$$, then any transformation is automatically a conformal transformation, so let's assume $$n\geq 2$$. If $$p=0$$ or $$q=0$$ then the conformal compactification $$\overline{\mathbb{R}^{p,q}}~\cong~\mathbb{S}^n$$ is an $$n$$-sphere.]

1. On one hand, there is the (global) conformal group $${\rm Conf}(p,q)~\cong~O(p\!+\!1,q\!+\!1)/\{\pm {\bf 1} \}\tag{4}$$ consisting of the set globally defined conformal transformations on $$\overline{\mathbb{R}^{p,q}}$$. This is a $$\frac{(n+1)(n+2)}{2}$$ dimensional Lie group. The quotients in eqs. (2) & (4) are remnants from the projective space (1).

The connected component that contains the identity element is $${\rm Conf}_0(p,q)~\cong~\left\{\begin{array}{ll} SO^+(p\!+\!1,q\!+\!1)/\{\pm {\bf 1} \} &\text{if both p and q are odd},\cr SO^+(p\!+\!1,q\!+\!1) &\text{if p or q are even}.\end{array}\right.\tag{5}$$ The two cases in eq. (5) correspond to whether $$-{\bf 1}\in SO^+(p\!+\!1,q\!+\!1)$$ or not, respectively. The global conformal group $${\rm Conf}(p,q)$$ has 4 connected components if both $$p$$ and $$q$$ are odd, and 2 connected components if $$p$$ or $$q$$ are even. The (global) conformal algebra $${\rm conf}(p,q)~\cong~so(p\!+\!1,q\!+\!1)\tag{6}$$ is the corresponding $$\frac{(n+1)(n+2)}{2}$$ dimensional Lie algebra. Dimension-wise, the Lie algebra breaks down into $$n$$ translations, $$\frac{n(n-1)}{2}$$ rotations, $$1$$ dilatation, and $$n$$ special conformal transformations.

2. On the other hand, there is the local conformal groupoid $${\rm LocConf}(p,q)~=~\underbrace{{\rm LocConf}_+(p,q)}_{\text{orientation-preserving}} ~\cup~ \underbrace{{\rm LocConf}_-(p,q)}_{\text{orientation-reversing}} \tag{7}$$ consisting of locally defined conformal transformations. Let us denote the connected component that contains the identity element $${\rm LocConf}_0(p,q)~\subseteq~{\rm LocConf}_+(p,q). \tag{8}$$ The local conformal algebroid $${\rm locconf}(p,q)~=~{\rm LocConfKillVect}(\overline{\mathbb{R}^{p,q}})\tag{9}$$ consists of locally defined conformal Killing vector fields, i.e. generators of conformal transformations.

• For $$n\geq 3$$, (the pseudo-Riemannian generalization of) Liouville's rigidity theorem states that all local conformal transformations can be extended to global conformal transformations, cf. e.g. this & this Phys.SE posts. Thus the local conformal transformations are only interesting for $$n=2$$.

• For the 1+1D Minkowski plane we consider light-cone coordinates $$x^{\pm}\in \mathbb{S}$$, cf. e.g. this Phys.SE post. The locally defined orientation-preserving conformal transformations are products of 2 monotonically increasing (decreasing) diffeomorphisms on the circle $$\mathbb{S}^1$$ \begin{align} {\rm LocConf}_+(1,1)~=~& {\rm LocDiff}^+(\mathbb{S}^1)~\times~ {\rm LocDiff}^+(\mathbb{S}^1) \cr &~\cup~ {\rm LocDiff}^-(\mathbb{S}^1)~\times~ {\rm LocDiff}^-(\mathbb{S}^1),\cr {\rm LocConf}_0(1,1)~=~& {\rm LocDiff}^+(\mathbb{S}^1)~\times~ {\rm LocDiff}^+(\mathbb{S}^1).\end{align}\tag{10} A orientation-reversing transformation is just a orientation-preserving transformation composed with the map $$(x^+,x^-)\mapsto (x^-,x^+)$$. The corresponding local conformal algebra $${\rm locconf}(1,1)~=~{\rm Vect}(\mathbb{S}^1)\oplus {\rm Vect}(\mathbb{S}^1) \tag{11}$$ becomes two copies of the real Witt algebra, which is an infinite-dimensional Lie algebra.

• For the 2D Euclidean plane $$\mathbb{R}^{2}\cong \mathbb{C}$$, when we identify $$z=x+iy$$ and $$\bar{z}=x-iy$$, then the locally defined orientation-preserving (orientation-reversing) conformal transformations are non-constant holomorphic (anti-holomorphic) maps on the Riemann sphere $$\mathbb{S}^2=\mathbb{P}^1(\mathbb{C})$$ \begin{align} {\rm LocConf}_0(2,0)~=~&{\rm LocConf}_+(2,0)\cr ~=~&{\rm LocHol}(\mathbb{S}^2), \cr\cr {\rm LocConf}_-(2,0)~=~&\overline{{\rm LocHol}(\mathbb{S}^2)} ,\end{align}\tag{12} respectively. An anti-holomorphic map is just a holomorphic map composed with complex conjugation $$z\mapsto\bar{z}$$. The corresponding local conformal algebroid $${\rm locconf}(2,0)~=~{\rm LocHolVect}(\mathbb{S}^2)\tag{13}$$ consists of generators of locally defined holomorphic (sans anti-holomorphic!) maps on $$\mathbb{S}^2$$. It contains a complex Witt algebra.

References:

1. M. Schottenloher, Math Intro to CFT, Lecture Notes in Physics 759, 2008; Chapter 1 & 2.

2. R. Blumenhagen and E. Plauschinn, Intro to CFT, Lecture Notes in Physics 779, 2009; Section 2.1.

3. P. Ginsparg, Applied Conformal Field Theory, arXiv:hep-th/9108028; Chapter 1 & 2.

4. J. Slovak, Natural Operator on Conformal manifolds, Habilitation thesis 1993; p.46. A PS file is available here from the author's homepage. (Hat tip: Vit Tucek.)

5. A.N. Schellekens, CFT lecture notes, 2016.

• Notes for later: ${\rm Conf}(p,q)~\cong~SO(p\!+\!1,q\!+\!1)$ if $p$ xor $q$ are even. ${\rm Conf}(p,q)~\cong~O^+(p\!+\!1,q\!+\!1)$ if $p$ is even. ${\rm Conf}_0(2,0)~\cong~SO^+(3,1)~\cong~PSL(2,\mathbb{C})$. ${\rm Conf}_0(1,1)~\cong~SO^+(2,2)/\{\pm {\bf 1} \}~\cong~PSL(2,\mathbb{R})\times PSL(2,\mathbb{R})$. Jun 22, 2019 at 12:30
• I was looking at this great answer and I had some doubts that I thought it would be reasonable to ask here instead of a new question. In Schottenloher there is no reference to the local conformal grupoid or algebroid. In particular, what is the mechanism of passing from LocConf(1,1) to locconf(1,1) and similarly with Euclidean space? Jul 29, 2021 at 17:31