1

Before reading this question it is essential that you understand power associativity

https://en.wikipedia.org/wiki/Power_associativity

In particular a commutative algebra does not necc imply a power-associative algebra.

Now Consider a commutative 3D algebra $T$ where the nonreal units $x,y$ satisfy

$$x^2 = A_1 + A_2 x + A_3 y $$ $$xy = B_1 + B_2 x + B_3 y$$ $$y^2 = C_1 + C_2 x + C_3 y$$

where all the parameters $A_1,A_2,..$ are real ofcourse.

So for instance we get

$$(a + b x + cy)^2 = a^2 + b^2 A_1 + 2bc B_1 + c^2 C_1 + ( 2ab + b^2 A_2 + 2bc B_2 + c^2 C_2) x + (2ac + b^2 A_3 + 2bc B_3 + c^2 C_3 ) y$$

And ofcourse there are some restrictions for the parameters $A_1,A_2,...$ ; the $x,y$ are not solvable in complex numbers ( such as $x = 1 +i, y = 2 $ ) and $x \neq y$ and more generally $x,y$ and the reals are linearly independant , because otherwise it is not really a 3D algebra. This implies that there are no numbers $i$ in the algebra that satisfy $i^2 = -1$.

I know that if $T$ is associative it is power-associative ofcourse, but that is trivial.

Now it is a known theorem that if an algebra is commutative and if for every element $p$ of that algebra $ p^4 = p^2 p^2 = p p^3 $ then the algebra is power-associative.

See also the link from the top.

This is a powerful tool.

Can we easily classify the condition for power-associative ?

In principle we just need to compute $p^2 p^2 - p p^3 $ and see if it is always $0$.

But that is alot of work.

Maybe there is an easier way.

Or maybe we can simplify things ?

addendum

For example

$$x^2 = 1 + 2 x + 3 y $$ $$xy = 4 + 5 x + 6 y$$ $$y^2 = 7 + 8 x + 9 y$$

Is not power-associative.

We have for instance $x^2 x^2 \neq x x^3 $


Update :

It is relatively easy to show that One requires

$$B_2,A_3,B_3,C_3 \neq 0$$

for the dimension to be actually $3$, aka the linear independance of $x,y$ I mentioned above.

I am also aware that this algebra is not closed under square roots.

Just informing.

edit 1

(removed due to mistake)

EDIT 2

It turns out that by substition $x/q - v = x' , y/s - w = y'$,

we can make $A_2 = C_3 = 0$ , $B_1 = 1$.

This greatly simplifies the system !

So in essense we are studying

$$x^2 = A_1 + A_3 y $$ $$xy = 1 + B_2 x + B_3 y$$ $$y^2 = C_1 + C_2 x$$


Remark : Im thinking about the connection to nilpotenty ... ($v^n = 0$), assuming one exists.


mick
  • 17,886
  • 1
    How do you define $p^n$? Isn't it by $p^{n+1}=p^np$? And, so, if your algebra is commutative, isn't your condition $p^4=pp^3$ automatically satified? why mention it then? – Anne Bauval Dec 20 '23 at 12:39
  • @AnneBauval no $p^2 p^2 = p^3 p$ is not automatically satisfied in general. It requires more than commutative and less than non-associative. See https://en.wikipedia.org/wiki/Power_associativity

    Albert's theorem : an algebra is power-associative if and only if it satisfies [ ... is the associator (Albert 1948).

    – mick Dec 20 '23 at 23:39
  • @AnneBauval the uniqueness of defining $p^n$ is exactly the thing " power-associative " . Like do you mean $(pp)(pp) $ or $p (ppp) $. They might be different ! – mick Dec 20 '23 at 23:42
  • 4
    mick: I know the definition and these Wikipedia pages, but you did not answer my 2 questions: when writing "if an algebra is commutative and if [...] $ p^4 = p^2 p^2 = p p^3 $ then the algebra is power-associative", your $p^2p^2$ is clearly defined, but $p^4$ and $p^3$ are (precisely!) ambiguous, so what is your definition for them? And (since your algebra is commutative) isn't the condition $p^4=pp^3$ automatically satisfied? (assuming that $p^n$ is defined inductively by $p^{n+1}=p^np$) – Anne Bauval Dec 21 '23 at 10:57
  • @AnneBauval no Anne that is not " automatically satisfied. Again $(pp) (pp) $ might be different from $p ( p * p * p )$ , there IS NO clear definition of $p^4$ unless those 2 are in fact identical , THAT IS THE WHOLE POINT of power-associativity. – mick Jan 05 '24 at 23:00
  • @AnneBauval Did you vote to close ? Being confused is not a reason to vote to close, I remind you and others. – mick Jan 05 '24 at 23:02
  • 2
    Please don't shout. And re-read my previous comment, which shows I am absolutely not "confused". And answer my 2 questions. I didn't vote to close but I shall now, for the same reason as the first voter: lack of details and clarity. – Anne Bauval Jan 06 '24 at 06:58
  • Anne Bauval I am not shouting, I am putting emphasis. – mick Jan 06 '24 at 18:43
  • (at everyone) I added an example where it is not power-associative. @AnneBauval – mick Jan 06 '24 at 18:44
  • @AnneBauval did you even look at my comment with link about power associative ? Albert's theorem shows me correct – mick Jan 06 '24 at 18:46

2 Answers2

1

Edit: I'm still quite confused by the wording, and this is not a complete answer but to correct my earlier mistakes while keep the meaningful parts.

Due to commutativity, for any $t\in T$, $t(tt)=(tt)t$, hence $t^3$ can be defined without ambiguity. The condition for power-associative by Albert requires $t^2t^2=t^3t$, in which case, $t^4$ can be defined without ambiguity. (We should not take $p^4=p^2p^2=p^3p$ as the condition, rather the condition is just $p^2p^2=p^3p$ and then use it to define $p^4$.)

I don't know an easy method to verify the condtion but here is a pretty easy example of power-associative but not associative algebra: $x^2=y^2=0, xy=yx=1$. One can check by hands this is indeed power-associative, or by Sage:

Build T

class T:
    def __init__(self, a, b, c):
        try:
            self.a = a.simplify_full()
            self.b = b.simplify_full()
            self.c = c.simplify_full()
        except:
            self.a, self.b, self.c = a, b, c
    def __str__(self):
        return str([self.a, self.b, self.c])
    def __eq__(self, B):
        return self.a==B.a and self.b==B.b and self.c == B.c
    def __mul__(self, B):
        return T(self.a*B.a + self.b*B.c + self.c*B.b, self.a*B.b + self.b*B.a, self.a*B.c + self.c*B.a)

Test

one = T(1, 0, 0)
x = T(0, 1, 0)
y = T(0, 0, 1)
zero = T(0, 0, 0)
assert(one*one==one and one*x==x and one*y==y)
assert(x*one==x and x*x==zero and x*y==one)
assert(y*one==y and y*y==zero and y*x==one)

Verification

var('a b c')
A = T(a, b, c)
A_sq = A*A
A_cub = A_sq*A
A4 = A_sq*A_sq
A4_ = A_cub*A
print(A4)
print(A4_)
assert(A4_ == A_cub*A)
Just a user
  • 22,048
  • I am sorry but most of your "answer " or even " comment " is a mix of confusion and being wrong. First of all , I can give a counterexample to your Proposition, what also implies your proof is wrong. The definition of $p^4$ is EXACTLY the issue ; if $p$ is power-associative then there is no debate : $p^4 $ is BOTH equal to $p^2 p^2 $ and $p p^3$. or if you want $ (pp) (pp) = p (p * p * p) = p^4 $.

    AS FOR YOUR EXAMPLE $x^2 = y^2 = 0 , xy = 0$ we get $(xy)^2 \neq x^2 y^2 $ because $0 \neq 1$.

    – mick Jan 04 '24 at 23:25
  • @mick you're right that I have some confusions earlier, sorry about it. But my example holds true. there is no need for $(xy)^2=x^2y^2$ for the algebra to be power-associative. I'm still very confused what those numerical examples in the post are trying to solve... if you mean for $p^2p^2=p^3p$ to be true, surely there are many different(non-isomorphic) solutions. – Just a user Jan 08 '24 at 05:24
  • That is an improvement yes. But I was thinking $xy = 1$, so $1/x = y$. But then $0 = y^2 = (1/x)^2 = ?? 1/(x^2) ?? = 1/0 = \infty $. Lets think : $(1/x)^2 = (x^{-1})^2 = (x^2)^{-1}$ ? And then we see the core thing : $(xy)^2 = x^2 y^2$ might not be required, but on the other hand $y$ is a negative power of $x$ in this case. It is quite strange $(xy)^2 = 1^2 = 1.$ But $x^2 y^2 = x^2 x^{-2} = 0.$ It is quite strange to have a number $x^2 x^{-2} = 0$ rather than $x^2 x^{-2} = 1$ or $x^{2 -2} = 1$ – mick Jan 08 '24 at 19:08
  • Also $0/y = y$ or $0/y = 0 x = 0$ ... I have already thought about these numbers, but I need to think about it more... – mick Jan 08 '24 at 19:17
  • Im upvoting anyways. – mick Jan 08 '24 at 20:41
  • it feels a bit like gluing two 2 dimensional algebra's together. but ok I guess ? – mick Jan 08 '24 at 21:28
  • For those that wonder about when this is not associative : $x(xy) = x , (xx)y = 0$. – mick Jan 08 '24 at 22:49
  • @mick To think in terms of inverse is not a good idea when the ring is not associative. For example, $(x+ay)y=1$ for any $a\in\mathbb R$, so which one of them is $1/y$? Also it's not a good idea either to think in terms of numbers or infinity in the context of abstract algebra. – Just a user Jan 09 '24 at 02:13
  • 1
    I am aware that zero divisors conflict with unique inverses. Even with associative. – mick Jan 09 '24 at 11:52
1

After alot of algebra it turns out that if $x^2 \neq y^2$ we get that power-associative implies that the algebra is completely associative.

The shortcut to showing this is that if $x^2 \neq y^2$ and we can show $x$ and $y$ are power-associative, we get the same equations required for associative.

Hence power-associative implies that the algebra is completely associative.

The special case $x^2 = y^2 = 0$ has been given by the answer of "Just a user".

The other cases $x^2 = y^2 \neq 0$ reduces to lower dimensions such as complex numbers or $x= y$, so it is not really 3D.

mick
  • 17,886