8
$\begingroup$

For two non-collinear vectors, $a$ and $b$, the quantity $$|a|^2 |b|^2 - (a \cdot b)(b \cdot a) = |a \times b|^2$$ is the square of the area of the parallelogram spanned by these two vectors. For three non-coplanar vectors, $a$, $b$ and $c$, we can form a similar expression $$B = |a|^2 |b|^2 |c|^2 - (a \cdot b) (b \cdot c) (c \cdot a)$$ which is not equal to the square of the volume of the parallelepiped spanned by these three vectors.

Does $B$ have any geometrical (or other) meaning?

$\endgroup$
  • 2
    $\begingroup$ So when you write $a^2$, do you mean $a\cdot a$? $\endgroup$ – Arnaud D. Sep 24 '18 at 14:52
  • $\begingroup$ @ArnaudD. Yes, exactly! $\endgroup$ – Fizikus Sep 24 '18 at 14:53
  • 1
    $\begingroup$ I took the liberty of replacing $x^2$ with $|x|^2$ to avoid further confusion about notation. $\endgroup$ – Blue Sep 24 '18 at 15:11
  • $\begingroup$ Is $B$ a definition or is the right hand side of $B$ derived from some other starting point? $\endgroup$ – DWade64 Sep 24 '18 at 22:52
  • 1
    $\begingroup$ @DWade64: $B$ is just the definition. $\endgroup$ – Fizikus Sep 24 '18 at 23:26
2
$\begingroup$

I don't know if this hint is conclusive, but this is what I got:

\begin{align*} B & = \lVert a\rVert^{2}\lVert b\rVert^{2}\lVert c\rVert^{2} - \langle a,b\rangle\langle b,c\rangle\langle c,a\rangle\\ & = \lVert a\rVert^{2}\lVert b\rVert^{2}\lVert c\rVert^{2} - \lVert a\rVert^{2}\lVert b\rVert^{2}\lVert c\rVert^{2}\cos(\theta_{1})\cos(\theta_{2})\cos(\theta_{3})\\ & = \lVert a\rVert^{2}\lVert b\rVert^{2}\lVert c\rVert^{2}[1 - \cos(\theta_{1})\cos(\theta_{2})\cos(\theta_{3})] \end{align*}

$\endgroup$
  • 1
    $\begingroup$ We can see that this quantity vanishes iff $a, b, c$ are parallel, so as in the case with two vectors, $B$ is a measure of failure of the three vectors to be parallel. $\endgroup$ – Travis Willse Sep 24 '18 at 15:18
2
$\begingroup$

I don't know a geometric interpretation for this quantity, but there's a natural interpretation of this quantity that makes it more transparently analogous to the other candidate you mentioned, namely the square of volume $V$ of the parallelepiped spanned by ${\bf a}, {\bf b}, {\bf c}$, at least for $n = 3$.

Recall that that squared area can be written as $$\begin{align*}V^2 &= \operatorname{det}\pmatrix{{\bf a} \cdot {\bf a} & {\bf a} \cdot {\bf b} & {\bf a} \cdot {\bf c} \\ {\bf b} \cdot {\bf a} & {\bf b} \cdot {\bf b} & {\bf b} \cdot {\bf c} \\ {\bf c} \cdot {\bf a} & {\bf c} \cdot {\bf b} & {\bf c} \cdot {\bf c}} \\ &= ({\bf a} \cdot {\bf a})({\bf b} \cdot {\bf b})({\bf c} \cdot {\bf c}) - \mathfrak{S}[({\bf a} \cdot {\bf a})({\bf b} \cdot {\bf c})^2] + 2 ({\bf b} \cdot {\bf c}) ({\bf c} \cdot {\bf a})({\bf a} \cdot {\bf b}),\end{align*}$$ where $\mathfrak{S}[\cdot]$ denotes the cyclic sum of $\cdot$ in ${\bf a}, {\bf b}, {\bf c}$.

On the other hand, recall that the determinant of an $n \times n$ matrix $M = (m_{ij})$ can be written as $$\det M = \sum_{\sigma} (\operatorname{sign} \sigma) m_{1 \sigma(1)} \cdots m_{n \sigma(n)},$$ where the sum is over all permutations $\sigma$ of $n$. Here, $\operatorname{sign} \sigma$ is $+1$ if the permutation is even and $-1$ if it is odd.

Now, if you know a little representation theory, you know $\operatorname{sign}$ is a representation of the group $S_n$ of permutations of $n$ elements. By replacing $\operatorname{sign}$ in the definition of the determinant with some other representation we get a generalization of the determinant (called an immanant) and---perhaps you see where this is going---can replace $\det$ in the first display equation above to get a new invariant for the (unordered) triple $({\bf a}, {\bf b}, {\bf c})$.

Taking $n = 3$ and the two-dimensional representation $\lambda$ gives the immanant $$\operatorname{imm}_{\lambda}(M) = 2 m_{11} m_{12} m_{13} - m_{23} m_{31} m_{12} - m_{32} m_{13} m_{21} .$$ Replacing $\det$ in the first display equation with $\operatorname{imm}_{\lambda}$ gives exactly twice the quantity in question: $$\boxed{\tfrac{1}{2} \operatorname{imm}_{\lambda} \pmatrix{{\bf a} \cdot {\bf a} & {\bf a} \cdot {\bf b} & {\bf a} \cdot {\bf c} \\ {\bf b} \cdot {\bf a} & {\bf b} \cdot {\bf b} & {\bf b} \cdot {\bf c} \\ {\bf c} \cdot {\bf a} & {\bf c} \cdot {\bf b} & {\bf c} \cdot {\bf c}} = ({\bf a} \cdot {\bf a})({\bf b} \cdot {\bf b})({\bf c} \cdot {\bf c}) - ({\bf b} \cdot {\bf c})({\bf c} \cdot {\bf a})({\bf a} \cdot {\bf b})} .$$

Remark Up to isomorphism there are only three irreducible representations of $S_3$, and the one remaining is the trivial representation. This gives rise to the permanent, $\operatorname{per}$, which is defined just by removing from the definition of $\det$ the sign of the permutation, so $$\begin{multline*}\operatorname{per}\pmatrix{{\bf a} \cdot {\bf a} & {\bf a} \cdot {\bf b} & {\bf a} \cdot {\bf c} \\ {\bf b} \cdot {\bf a} & {\bf b} \cdot {\bf b} & {\bf b} \cdot {\bf c} \\ {\bf c} \cdot {\bf a} & {\bf c} \cdot {\bf b} & {\bf c} \cdot {\bf c}} \\ = ({\bf a} \cdot {\bf a})({\bf b} \cdot {\bf b})({\bf c} \cdot {\bf c}) + \mathfrak{S}[({\bf a} \cdot {\bf a})({\bf b} \cdot {\bf c})^2] + 2 ({\bf b} \cdot {\bf c}) ({\bf c} \cdot {\bf a})({\bf a} \cdot {\bf b}),\end{multline*}$$ where $\mathfrak{S}[\cdot]$ denotes the cyclic sum of $\cdot$ in ${\bf a}, {\bf b}, {\bf c}$. It would be interesting to have a geometric interpretation of this quantity, too.

$\endgroup$
  • $\begingroup$ Very nice, I wasn't aware of these generalizations of the determinant. Do they have some practical applications? $\endgroup$ – Fizikus Sep 24 '18 at 20:06
  • 1
    $\begingroup$ See this thread on MO: mathoverflow.net/questions/66284/… . This post on Terry Tao's blog about the permanent in a particular case is an interesting read, too: terrytao.wordpress.com/2008/04/16/… I don't know offhand of any applications for the Immanent $\endgroup$ – Travis Willse Sep 24 '18 at 23:27
0
$\begingroup$

It occurred to me that if one seeks a geometrical interpretation of $B$, then one has to write it in terms of other geometrical quantities, i.e., length squared $|a|^2$, area squared $|a \times b|^2$ and volume squared $|a \cdot (b \times c)|^2$. Doing this, we find $$B = \frac{1}{2} \left[ |a|^2 |b \times c|^2 + |b|^2 |c \times a|^2 + |c|^2 |a \times b|^2 - |a \cdot (b \times c)|^2\right].$$ Now, $B$ vanishes only if all three vectors are collinear so it cannot represent any geometrical volume. On the other hand, the dimensionality of $B$ eliminates the possibility of describing some area. Therefore, I conclude that $B$ does not represent any simple geometrical quantity.

$\endgroup$

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.