In Blair's book and many many literatures, I see definition of a contact metric manifold which involves a relation \begin{equation} d\kappa \left( {X,Y} \right) = g\left( {X,\Phi Y} \right) \end{equation} as well as other standard relation like ${\Phi ^2} = - 1 + R \otimes \kappa $, $\Phi R = 0$, etc.

On the other hand, in Tanno's papers and a few others literatures, the definitions involves \begin{equation} d\kappa \left( {X,Y} \right) = 2 g\left( {X,\Phi Y} \right) \end{equation} differing from Blair's by a factor of $2$.

So I try to figure out which is "more correct", and then comes one important equation that really confuses me, in Blair's book (Lemma 6.2): \begin{equation} {\nabla _X}R = - \Phi X - \Phi hX, \;\;\;\;h \propto \mathcal{L}_R \Phi\;\;\;\;\;\;\;\;(*) \end{equation}

Now, let us consider the contact metric structure to be K-contact, and therefore $h = 0$, ${\mathcal{L}_R}g = 0$, and also equation $(*)$ should still hold, hence (in components) \begin{equation} {\nabla _X}R = - \Phi X \Leftrightarrow {\nabla _m}{R^n} = - {\Phi ^n}_m \Rightarrow {\nabla _m}{R_n} = - {g_{nk}}{\Phi ^k}_m\;\;\;\;\;\;\;\;(**) \end{equation}

On the other hand, since $\nabla$ is Levi-civita, we have \begin{equation} {d\kappa } \left( {X,Y} \right) = \left( {{\nabla _m}{R_n} - {\nabla _n}{R_m}} \right){X^m}{Y^n} = g\left( {X,\Phi Y} \right) = {g_{mk}}{\Phi ^k}_n{X^m}{Y^n} \end{equation} and therefore ${\nabla _m}{R_n} - {\nabla _n}{R_m} = {g_{mk}}{\Phi ^k}_n$. Now comes the final point: using the Killing vector equation ${\nabla _m}{R_n} = - {\nabla _n}{R_m}$, one obtains \begin{equation} - 2{\nabla _m}{R_n} = 2{\nabla _n}{R_m} = - {g_{mk}}{\Phi ^k}_n \end{equation} which directly contradict the equation $(**)$

So my question is

Is Blair's definition or Tanno's "more correct"? Is Blair's definition incompatible with the Lemma 6.2? Note that Tannos definition has the correct factor of 2, so $(*)$ is compatible with the definition.


The discrepancy comes from which coefficient convention is being used for the covariant derivative d. Blair uses the convention that for any 1-form $\eta$ $$d\eta(X,Y) = \frac{1}{2}\left( X\eta(Y) - Y\eta(X) - \eta[X,Y]\right)$$ whereas Tanno is most likely using $$d\eta(X,Y) = X\eta(Y) - Y\eta(X) - \eta[X,Y].$$

The difference between these is conceptually nil, but computationally important. And, in fact, it is derived from the wedge product convention for 1-forms: $$\eta \wedge \zeta = \frac{1}{2}(\eta \otimes \zeta - \zeta \otimes \eta)$$ versus $$\eta \wedge \zeta = \eta \otimes \zeta - \zeta \otimes \eta.$$

Neither is more correct than the other. It's just which convention one chooses.

| cite | improve this answer | |
  • $\begingroup$ I agree with Brendan on the two definitions of $d$: I found at the bottom of page 62, Blair wrote down $d\eta$ formula. I feel that a clear list of convention should be provided at the very beginning of the book. Thanks! $\endgroup$ – Lelouch May 4 '14 at 3:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.