0
$\begingroup$

sorry for the bad image quality, I had to improvise

Definiton of a hypercomplex number [it's in the tag] :

A hypercomplex number is an element of a finite-dimensional algebra over the real numbers that is unital and distributive (but not necessarily associative).

The defining properties of this system are as follows:

$\ ij=-j,$ $\ ji=-i,$ $\ i^2=j^2=-1 $

One thing to note is this system is neither commutative nor anti-commutative

$\ ji=-i, $ $\ ij=-j,$ $\ -ij=j$

$\ -i\neq-j\neq j $

$\therefore ji\neq ij\neq-ji$

the question has already been stated in the title, if there is anything missing than I will edit it accordingly

$\endgroup$

1 Answer 1

2
$\begingroup$

If this is assumed to be associative then $$ ij = jiij = 1 $$ and so $$ -j = iij = i. $$ But $$ 1 = -jj = ij = -j $$ but also $$ 1 = -ii = ji = -i. $$ Thus $1 = -1$ and your system is inconsistent.


Because the algebra is not associative it can't be the trivial algebra.

Note that the algebra is spanned by $1,i,j$.

Suppose now that $j = \alpha i$ for some scalar $\alpha$. Then $$ -i = (\alpha i)i = -\alpha^2 $$ and $i = \alpha^2$, making the algebra trivial. This is impossible, so $i$ and $j$ are linearly independent.

Suppose now that $1 = \alpha i + \beta j$ for some scalars $\alpha,\beta$. Then $$ 1 = (\alpha i + \beta j)^2 = -\alpha^2 - \beta^2 - \alpha\beta(i + j). $$ This implies that one of $\alpha, \beta$ must be nonzero. Thus $$ 1 = \frac{-\alpha\beta}{1+\alpha^2+\beta^2}(i+j). $$ Because $i,j$ are linearly independent, this implies $$ \alpha = \beta = \frac{-\alpha^2}{1+2\alpha^2} \implies 2\alpha^3 + \alpha^2 + \alpha = 0 $$ and because $\alpha$ cannot be zero we get $$ 2\alpha^2 + \alpha + 1 = 0. $$ The discriminant of this quadratic equation is $$ 1 - 4(2)(1) < 0 $$ so there are no real solutions, which contradicts the assumption that $1$ is a linear combination of $i,j$.

Thus $1,i,j$ is a basis for the algebra, and we can define multiplication by $$ (\alpha + \beta i + \gamma j)(\alpha' + \beta' i + \gamma' j) = \alpha\alpha' - \beta\beta' - \gamma\gamma' + (\alpha\beta' + \beta\alpha' - \gamma\beta')i + (\alpha\gamma' + \gamma\alpha' - \beta\gamma')j.$$ This multiplication satisfies the requisite identities; this proves that the algebra is consistent and unique.

$\endgroup$
6
  • 2
    $\begingroup$ Simpler: $\ [\color{#c00}{-j} = ij = \color{#0af}{\bf 1}]^2 \Rightarrow\, \color{#c00}{\bf {-}\!1} = \color{#0af}{\bf 1}\ \ \ $ $\endgroup$ Commented Jun 16, 2024 at 4:10
  • $\begingroup$ On what basis do we assume this? Is there a deeper reason it must be associative, or that this cannot count as simply just proof of a specific property by contradiction? If not then isn't it in the definition of a hyper-complex number that it is not necessarily associative? $\endgroup$ Commented Jun 16, 2024 at 8:47
  • $\begingroup$ @AnonymousPoster It can be whatever you want it to be, "hypercomplex" is not formally defined as far as I know. But you really need to prove your system is consistent even then, it's not something you can just assume. I suspect Hurwitz's theorem is applicable here, in which case your system is inconsistent. $\endgroup$ Commented Jun 16, 2024 at 18:33
  • $\begingroup$ You still haven't answered why you assumed it's associative but what you said recently is a more useful answer, is it ok if I add that in to your previous post? Also for clarification while I can do mathematics with quaternions and duals relatively well (I was gifted with a pretty ok 4d visualization at a weirdly young age so I can just graph it, really), I often struggle with the other hypercomplex systems, which why I'm asking for help online understanding this subject. $\endgroup$ Commented Jun 17, 2024 at 7:43
  • $\begingroup$ @AnonymousPoster No, I don't think you should add anything to the answer body about Hurwitz's theorem unless you prove it applies. In fact, it doesn't: I've added a proof that your algebra is uniquely defined. I've also come to the realization that consistency is trivial: writing $S = \{\pm,\pm j,\pm i\}$ and assuming all these elements are distinct, as long as the binary operation $S\times S \to S$ defined by your rules is consistent with $1$ being the identity and $\pm1$ being scalars then it automatically defines a 3D algebra. Your rules do do this... $\endgroup$ Commented Jun 17, 2024 at 15:56

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.