Consider two different charts \((U,\varphi)\) and \((V,\psi)\) on a manifold \(M\) with \(p\in U\cap V \neq \emptyset\). Assuming \((x^i)\) and \((\widetilde{x}^i)\) are the corresponding local coordinates, we obtain on \(T_pM\)

\begin{equation*} \frac{\partial }{\partial x^i}\bigg|_p=\frac{\partial \widetilde{x}^{j}}{\partial x^i}(p)\frac{\partial }{\partial \widetilde{x}^j}\bigg|_p, \end{equation*}

where \(\widetilde{x}^j=\psi\circ \varphi^{-1}\) in the numerator.

This implies that the components of a vector

\begin{equation*} v=v^i\frac{\partial }{\partial x^i}|_p \end{equation*}

are given by

\begin{equation*} \widetilde{v}^j=\frac{\partial \widetilde{x}^j}{\partial x^i}(p) v^i. \end{equation*}

Since the components of \(v\) transforms in the opposite way as the coordinate partial derivatives, tangent vectors are called contravariant vectors.

Remark
  • from set theoretic point of view a tangent vector should be called covariant vector [1, p.276].

References Link to heading

  1. J. Lee, Introduction to Smooth Manifolds. New York ; London: Springer, 2013.