Does this symmetric rank-3 tensor vanish?

81 Views Asked by At
  • Suppose we have a rank-3 tensor $T$ on some vector space $\mathbb{V}$. We can view $T$ as a map: $$T: \mathbb{V} \times \mathbb{V} \times \mathbb{V} \to \mathbb{R},$$ which maps triples of vectors in $\mathbb{V}$ to the real numbers. By the definition of tensors, $T$ is linear: $$ T(\alpha \vec{v}_1 + \beta \vec{v}_2, \vec{u}, \vec{w}) = \alpha T(\vec{v}_1 , \vec{u}, \vec{w}) + \beta T(\vec{v}_2 , \vec{u}, \vec{w}), $$ and similarly for the other two arguments.

  • Suppose, further, that $T$ is completely symmetric in its inputs: $$ T(\vec{v}_1, \vec{v}_2, \vec{v}_3) = T(\vec{v}_1, \vec{v}_3, \vec{v}_2) = T(\vec{v}_2, \vec{v}_1, \vec{v}_3) = \dots $$ for any $\{\vec{v}_1, \vec{v}_2, \vec{v}_3\} \in \mathbb{V}$, and for all six permutations of the inputs.

  • Finally, suppose that we know that $T(\vec{v}, \vec{v}, \vec{v}) = 0$ for all $\vec{v} \in \mathbb{V}$.

My question:

Does it follow that $T(\vec{v}_1, \vec{v}_2, \vec{v}_3) = 0$ for all $\{\vec{v}_1, \vec{v}_2, \vec{v}_3\}$?

The analogous statement is true for rank-2 tensors, so it would seem strange to me if it wasn't true here as well. On the other hand, the argument I'm familiar with for the rank-2 case [i.e.: plug $\vec{v}_1 + \vec{v}_2$ into both slots and use the linearity properties to argue that $T(\vec{v}_1, \vec{v}_2) = 0$] doesn't carry over to this case, so I'm not sure how to prove it.

1

There are 1 best solutions below

0
On BEST ANSWER

Expanding by linearity and removing all terms we know are zero, and using symmetry to gather the rest of the terms, we get $$\begin{align} 0 = &\,\frac13T(\vec v_1 + \vec v_2 + \vec v_3,\vec v_1 + \vec v_2 + \vec v_3, \vec v_1 + \vec v_2 + \vec v_3)\\ =&\, T(\vec v_1, \vec v_1, \vec v_2)+ T(\vec v_1, \vec v_2, \vec v_2) \\ &+ T(\vec v_1, \vec v_1, \vec v_3) + T(\vec v_1, \vec v_3, \vec v_3) \\ &+ T(\vec v_2, \vec v_2, \vec v_3) + T(\vec v_2, \vec v_3, \vec v_3)\\& + 2T(\vec v_1, \vec v_2, \vec v_3)\end{align} $$ (where the $\frac13$ is just for readability). In order to get any further, we need to study the value of $T(\vec v, \vec v, \vec u) + T(\vec v ,\vec u , \vec u)$, which appears in three different variations above. Applying $T$ to three copies of $\vec v + \vec u$, we get $$ 0 = \frac13T(\vec v + \vec u, \vec v + \vec u, \vec v + \vec u)= T(\vec v, \vec v, \vec u) + T(\vec v, \vec u, \vec u) $$ and therefore we conclude that $T$ vanishes.