$$ \left\{ \begin{aligned} c_1 & = a_2b_3-b_2a_3 \\ c_2 & = a_3b_1-b_3a_1 \\ c_3 & = a_1b_2-b_1a_2 \end{aligned} \right. $$ $c_1,c_2,c_3\in \mathbb{Z}$ is given,prove that $\exists a_1,a_2,a_3,b_1,b_2,b_3\in\mathbb{Z}$ meet the equations set.

Apparently,the question equal to how to decompose the integer vector into the cross product (vector product) of two integer vector.And the real question I want to ask just is it.

  • 2
    $\begingroup$ This can be rewritten into vector cross product notation: $$ [a_1, a_2, a_3]\times [b_1, b_2, b_3] = [c_1, c_2, c_3] $$ $\endgroup$ – Arthur May 22 '15 at 8:12
  • $\begingroup$ Hint: the left hand side is a vector product. $\endgroup$ – Martigan May 22 '15 at 8:12
  • $\begingroup$ @Arthur yes,the real question I want to ask is how to decompose a integer vector into cross product (vector product) of two integer vector. $\endgroup$ – lanse2pty May 22 '15 at 8:21
  • $\begingroup$ @Martigan I know it ,but I don't know how to prove it . $\endgroup$ – lanse2pty May 22 '15 at 8:22
  • $\begingroup$ No other conditions? say some hypothesis on the coprimeness of $c_{1}, c_{2}, c_{3}$? $\endgroup$ – Megadeth May 22 '15 at 8:54

This is only a proof of a special case of the problem.

The problem becomes simple if we have in addition the assumption that $(c_{1}, c_{2}) \mid c_{3}.$ For, taking $a_{2} := c_{1},$ $b_{3} := 1,$ $a_{3} := 0,$ $a_{1} := -c_{2}$ fulfills the first two equations and from the third we have $c_{1}b_{1} + c_{2}b_{2} + c_{3} = 0.$ Since $(c_{1}, c_{2}) \mid c_{3},$ it follows that there are infinitely many integers $b_{1}, b_{2}$ for which the third holds.

  • $\begingroup$ So thank you. Can you explain the mean of $(c_1,c_2)|c_3$, I really forgot it .Sorry. $\endgroup$ – lanse2pty May 22 '15 at 9:14
  • $\begingroup$ No problem. In number theory it is usual to denote the greatest common divisor of two integers $x, y$ by $(x, y)$. $\endgroup$ – Megadeth May 22 '15 at 9:18
  • $\begingroup$ Thank you. Why $(c_1,c_2)|c_3$ imply $c_1b_1+c_2b_2+c_3=0$ has integer solution?Besides, as the other answer (Adelafif's answer),it is easy to see for having integer solution ,$c_1,c_2,c_3$ not must be $(c_1,c_2)|c_3$. But really thank you . $\endgroup$ – lanse2pty May 22 '15 at 10:52
  • $\begingroup$ @lanse2pty The number $(c_1, c_2)$ is the smallest non-zero number that can be written as a linear combination of $c_1$ and $c_2$. All other numbers that can be written as a linear combination of $c_1$ and $c_2$ is a multiple of that. And any multiple of the gcd can be written as a linear combination. Since if $mc_1 + nc_2 = (c_1, c_2)$, then the multiple $r(c_1, c_2) = r(mc_1 + nc_2) = (rm)c_1 + (rn)c_2$. $\endgroup$ – Arthur May 22 '15 at 11:33
  • $\begingroup$ Seemingly, I learned it ,but have forgot,thank you. $\endgroup$ – lanse2pty May 22 '15 at 11:41

In fact the result holds for any dimension.

Take a primitive vector $c$ (coordinates coprime).

Then there is $L\in GL(Z)$ such that $Lc$ be the unit vector of the canonical basis.

You take the cross product of all the lines of $L$ not corresponding to the $0$ of the vector and you get your initial vector $c$.

At the very basis of it is Euclid. You can transform your vector with only integer transformations to get to a unit vector.

  • $\begingroup$ The $c$ must be coordinates coprime? $\endgroup$ – lanse2pty May 22 '15 at 11:47

First let us assume that $c_1, c_2, c_3$ are coprime, that is $(c_1, c_2, c_3) = 1$.

Suppose first that there exist integer numbers $u_1, u_2, u_3$ and $v_1, v_2, v_3$ such that $$ \begin{vmatrix} c_1 & c_2 & c_3 \\ u_1 & u_2 & u_3 \\ v_1 & v_2 & v_3 \end{vmatrix} = 1 $$

In that case we can take $a = c \times u$ and $b = c \times v$. Then $a \times b = (c \times u) \times (c \times v) = (c \cdot (u \times v))c$ (this is one of triple product properties, see http://en.wikipedia.org/wiki/Triple_product for details). Now $c \cdot (u \times v)$ is equal to the determinant above and hence is equal to one. So $a \times b = c$.

Now let us go back and prove the assumption. Let $d = (c_2, c_3)$ and pick $u_2, u_3$ such that $c_2 u_3 - c_3 u_2 = d$. Also pick $u_1 = 0$. Now we have

$$ \begin{vmatrix} c_1 & c_2 & c_3 \\ u_1 & u_2 & u_3 \\ v_1 & v_2 & v_3 \end{vmatrix} = v_1 d - v_2 (c_1 u_3 - c_3 u_1) + v_3 (c_1 u_2 - c_2 u_1) = v_1 d + c_1(v_3 u_2 - v_2 u_3). $$

Now $(c_1, d) = 1$ and $(u_2, u_3) = 1$. Let us pick $v_2$ and $v_3$ such that $v_3 u_2 - v_2 u_3 \equiv c_1^{-1} \pmod d$. In that case the right hand side is congruent to $1$ modulo $d$. Finally let us pick $v_1$ such that it is equal to $1$.


The given integral vector v has at least one rational vector w such that w ,v are perpendicular. We take the cross product u of v and w and this is rational. Now we find a suitable common multiple of u and w to eliminate the denominators. Thus we get new integral vectors u',w' perpendicular to v.

  • 2
    $\begingroup$ The issue is that you have a vector which is proportionnal, not equal... The OP wants to prove that there exists integer vectors such as the direct cross product gives with no factor the initial integer vector. $\endgroup$ – Martigan May 22 '15 at 9:52
  • $\begingroup$ Although your answer is a little irrelevant ,but I think ,maybe there is a way to prove or deny the question in your answer.Thank you . $\endgroup$ – lanse2pty May 22 '15 at 10:56

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.