4

I. First condition

The question of what integers of form $ax^2+bxy+cy^2$ are closed under multiplication is quite interesting. We give a simple proof that it is the case if $\color{blue}{a=1}$,

$$(p^2 + bpq + cq^2)(r^2 + brs + cs^2) = z_1^2+bz_1z_2+cz_2^2$$

\begin{align} z_1 &= p r - c q s\\ z_2 &= p s + q r + b q s\end{align}

So it remains to consider $a\neq1$. Vladimir Arnold showed that if $ax^2+bxy+cy^2 =1$ for some integer $(x,y)$, then there is multiplicative closure. But it is also true that,

$$(2p^2+pq+3q^2)(2r^2+rs+3s^2)=2x^2+xy+3y^2$$

and $2x^2+xy+3y^2\neq1$ for any integer $(x,y)$. We need a second condition.


II. Second condition

"If there are integers $(r_1, r_2, r_3, r_4)$ such that,

$$ar_1^2+br_1r_2+cr_2^2=ac\\ r_3=(ar_1+br_2)/c\\ r_4=(br_1+cr_2)/a$$

then the quadratic form $ax^2+bxy+cy^2$ has multiplicative closure."

Proof: This is given by the identity,

$$(au^2+buv+cv^2)(ax^2+bxy+cy^2)\left(1+\frac{ar_1^2+br_1r_2+cr_2^2-ac}{ac}\right) = az_1^2+bz_1z_2+cz_2^2$$

where,

\begin{align}z_1 &= u(-r_2x+r_1y)+v(r_1x+r_4y)\\ z_2 &= u(r_3x+r_2y)-v(-r_2x+r_1y)\\ r_3 &= (ar_1+br_2)/c\\ r_4 &= (br_1+cr_2)/a \end{align}

So all we need is to solve $ar_1^2+br_1r_2+cr_2^2-ac=0$ (which also yields integer $r_3,r_4$) and the closure follows.


III. Examples

The quadratic form $2x^2+xy+3y^2$ has discriminant $d=-23$, class number $h(d)=3$,

$$(2u^2+uv+3v^2)(2x^2+xy+3y^2) = 2z_1^2+z_1z_2+3z_2^2$$

\begin{align}z_1 &= u(-x+y)+v(x+2y)\\ z_2 &= u(x+y)-v(-x+y)\end{align}

since $(r_1, r_2, r_3, r_4) = (1,1,1,2)$.

For class number $h(d) = 3, 6, 9$, there are (16, 51, 34) negative fundamental discriminants $d$, respectively. In the data below, for brevity only the first five $d$ for each will be given. The format is $d;\; (a,b,c);\;(r_1, r_2, r_3, r_4)$.

A. Class number 3

\begin{align} -23;&\;(2,1,3);\;(1,1,1,2)\\ -31;&\;(2,1,4);\;(2,0,1,1)\\ -59;&\;(3,1,5);\;(2,-1,1,-1)\\ -83;&\;(3,1,7);\;(2,1,1,3)\\ -107;&\;(3,1,9);\;(3,0,1,1)\\ \end{align}

B. Class number 6

\begin{align} -87;&\;(4,3,6);\;(0,2,1,3)\\ -104;&\;(3,2,9);\;(3,0,1,2)\\ -116;&\;(5,2,6);\;(2,1,2,2)\\ -152;&\;(6,4,7);\;(1,2,2,3)\\ -212;&\;(6,2,9);\;(3,0,2,1)\qquad \end{align}

C. Class number 9

\begin{align} -199;\;&(7, 5, 8);\;(3, -1, 2, 1)\\ -367;\;&(8, 7, 13);\;(-1, 3, 1, 4)\\ -419;\;&(5, 1, 21);\;(4, 1, 1, 5)\\ -491;\;&(5, 3, 25);\;(5, 0, 1, 3)\\ -563;\;&(11, 3, 13);\;(2, -3, 1, -3) \end{align}

and so on.


IV. Question

Q: Is it true that quadratic forms $ax^2+bxy+cy^2$ with $a\neq1$ and negative discriminants with class number $h(d)=3m$, then ALL have multiplicative closure?

Update: As pointed out by Jagy in his answer, I should have focused on quadratic forms in the principal genus. For example, for $d=-23$ with class number 3, the 3 forms are,

$$\color{blue}{(1, 1, 6), (2, -1, 3), (2, 1, 3)}$$

but for $d=-87$ (not prime) with class number 6, the 6 forms are,

$$\color{blue}{(1, 1, 22)}, (2, -1, 11), (2, 1, 11), (3, 3, 8), \color{blue}{(4, -3, 6), (4, 3, 6)}$$

and for $d=-199$ (prime) with class number 9, the 9 forms are,

$$\color{blue}{(1, 1, 50)}, (2, -1, 25), (2, 1, 25), (4, -3, 13), (4, 3, 13), (5, -1, 10), (5, 1, 10), \color{blue}{(7, -5, 8), (7, 5, 8)}$$

and only those in blue have multiplicative closure. Thus, for class number $h(d)=3m>3$, then not all quadratic forms per discriminant $d$ will qualify, but it seems we can find at least three.

2 Answers2

2

Well, no. Any discriminant with more than one genus possesses forms without such closure. That is, the "duplicate" of any form lies in the principal genus, that being different from the given form.

Discriminant $-87,$ which is composite $3 \cdot 29$ has two genera. The principal genus has

$$ < 1, 1, 22> \; , \; < 4, -3, 6> \; , \; < 4, 3, 6> $$ while the other genus is $$ < 3, 3, 8> \; , \; < 2, -1, 11> \; , \; < 2, 1, 11> $$

The duplicate of $<2,\pm 1, 11> $ will be $<4,\pm 3, 6>, $ not sure about the $\pm$ signs. Also, the duplicate of $<3, 3, 8> $ will be $<1,1, 22>, $

More interesting are cases where the class number is odd, so just one genus, and class number a multiple of $3.$ For example, discriminant $-199$ has class number $9,$ while $-439$ has class number $15$

The thing that is true: any form of order $3$ in the class group has your multiplication property. However, in discriminant $-199,$ the cube of $<2,1,25>$ is actually $<7,-5,8>.$ This latter form $<7,-5,8>$ really does have order $3$ Meanwhile, the square of $<2,1,25>$ is $<4,-3,13>.$

Will Jagy
  • 146,052
  • I think I was not precise in my question. Reading more closely the article I wrote years ago, I observed that for $h(d)=3m$, then at least three forms $(1,b_1,c_1), (a,b,c), (a,-b,c)$ have multiplicative closure. For $h(d)=3$ that's all the forms. For $h(d)=6$ and first ten $d$, these obey the 2nd condition, \begin{align} -87;&;(4,3,6)\ -104;&;(3,2,9)\ -116;&;(5,2,6)\ -152;&;(6,4,7)\ -212;&;(6,2,9)\ -244;&;(5,4,13)\ -247;&;(4,3,16)\ -339;&;(7,5,13)\ -411;&;(7,3,15)\ -424;&;(10,4,11)\end{align} Am I right in assuming these are in the principal genus for each $d$? – Tito Piezas III Jan 03 '25 at 19:39
  • 1
    @TitoPiezasIII yes, at least three, and these are in the principal genus, as $f^2 = \bar{f}$ forces $\bar{f}$ and then $f$ to lie in the principal genus. I'm looking up some things in Buell. – Will Jagy Jan 03 '25 at 19:45
  • Will, I updated the post to clarify the question. Did you find anything new in Buell? – Tito Piezas III Jan 04 '25 at 05:15
  • 1
    @TitoPiezasIII yes, confirmed that there may be more than three of your forms, if the class group is not cyclic. The class group is finite abelian and, most of the time, is cyclic. But not always. – Will Jagy Jan 04 '25 at 18:22
  • I should add that there are infinitely many of these. Take integers $a,b$ coprime. If $b^2 - 4 a^3 < 0$ then the form $<a, b, a^2 >$ has order three. If $b > a$ it is not reduced, of course... My memory from years ago is that we cannot expect to find all class three forms this way... – Will Jagy Jan 04 '25 at 19:01
  • @TitoPiezasIII Same idea, if $b^2 > 4a^3$ and so $d = b^2 - 4 a^3 > 0$ and is not a perfect square, the form $<a,b, a^2>$ still cubes to the identity. However, with indefinite firms, it is extra work to determine whether it is the principal form, extra work to find whether two forms are equal, a certain amount of annoyance to find whether the principal form also represents $-1$ (this come down to the negative Pell equation) – Will Jagy Jan 04 '25 at 21:14
2

I was trying to decide whether more than three of your forms would be possible. That means a form class group that is not cyclic. Buell published on this, finding prescribed class groups. On page 154 he lists several interesting examples. The first one, with group $3 \times 9,$ is discriminant $-3299,$ shows nine forms of order three.

Starting page 135 he mentions the strong tendency of class groups to be cyclic.

On page 156 he lists discriminants with the first occurrence of 3-rank three , -3321607, and 3-rank four, -653329427

$$ \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc $$

Each line below is a form, then its square, then its cube.. nine times the cube is <1, 1, 825>

 all  
    3299:  < 1, 1, 825>    Square      3299:  < 1, 1, 825> cube     3299:  < 1, 1, 825>
    3299:  < 3, -1, 275>    Square      3299:  < 9, -7, 93> cube     3299:  < 27, -7, 31>
    3299:  < 3, 1, 275>    Square      3299:  < 9, 7, 93> cube     3299:  < 27, 7, 31>
    3299:  < 5, -1, 165>    Square      3299:  < 25, -1, 33> cube     3299:  < 27, -7, 31>
    3299:  < 5, 1, 165>    Square      3299:  < 25, 1, 33> cube     3299:  < 27, 7, 31>
    3299:  < 9, -7, 93>    Square      3299:  < 17, -13, 51> cube     3299:  < 27, 7, 31>
    3299:  < 9, 7, 93>    Square      3299:  < 17, 13, 51> cube     3299:  < 27, -7, 31>
    3299:  < 11, -1, 75>    Square      3299:  < 11, 1, 75> cube     3299:  < 1, 1, 825>
    3299:  < 11, 1, 75>    Square      3299:  < 11, -1, 75> cube     3299:  < 1, 1, 825>
    3299:  < 13, -9, 65>    Square      3299:  < 5, -1, 165> cube     3299:  < 27, 7, 31>
    3299:  < 13, 9, 65>    Square      3299:  < 5, 1, 165> cube     3299:  < 27, -7, 31>
    3299:  < 15, -11, 57>    Square      3299:  < 15, 11, 57> cube     3299:  < 1, 1, 825>
    3299:  < 15, -1, 55>    Square      3299:  < 29, -23, 33> cube     3299:  < 27, 7, 31>
    3299:  < 15, 1, 55>    Square      3299:  < 29, 23, 33> cube     3299:  < 27, -7, 31>
    3299:  < 15, 11, 57>    Square      3299:  < 15, -11, 57> cube     3299:  < 1, 1, 825>
    3299:  < 17, -13, 51>    Square      3299:  < 3, 1, 275> cube     3299:  < 27, -7, 31>
    3299:  < 17, 13, 51>    Square      3299:  < 3, -1, 275> cube     3299:  < 27, 7, 31>
    3299:  < 19, -11, 45>    Square      3299:  < 15, -1, 55> cube     3299:  < 27, -7, 31>
    3299:  < 19, 11, 45>    Square      3299:  < 15, 1, 55> cube     3299:  < 27, 7, 31>
    3299:  < 23, -17, 39>    Square      3299:  < 23, 17, 39> cube     3299:  < 1, 1, 825>
    3299:  < 23, 17, 39>    Square      3299:  < 23, -17, 39> cube     3299:  < 1, 1, 825>
    3299:  < 25, -1, 33>    Square      3299:  < 13, 9, 65> cube     3299:  < 27, 7, 31>
    3299:  < 25, 1, 33>    Square      3299:  < 13, -9, 65> cube     3299:  < 27, -7, 31>
    3299:  < 27, -7, 31>    Square      3299:  < 27, 7, 31> cube     3299:  < 1, 1, 825>
    3299:  < 27, 7, 31>    Square      3299:  < 27, -7, 31> cube     3299:  < 1, 1, 825>
    3299:  < 29, -23, 33>    Square      3299:  < 19, 11, 45> cube     3299:  < 27, -7, 31>
    3299:  < 29, 23, 33>    Square      3299:  < 19, -11, 45> cube     3299:  < 27, 7, 31>

$$ \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc $$

$$ \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc $$

Will Jagy
  • 146,052
  • Mathematica says $d=-3299$ has class number $27$. I find it has $\large{9}$ quadratic forms that have multiplicative closure, namely $(1,1,85)$ plus four pairs of $(a,\pm b, c)$. Format below is still $d;,(a,b,c);,(r_1,r_2,r_3,r_4)$, \begin{align} -3299;;& (11, 1, 75);;; (7, -2, 1, -13)\ -3299;;& (15, 11, 57);; (6, -3, 1, -7)\ -3299;;& (23, 17, 39);; (2, -5, -1, -7)\ -3299;;& (27, 7, 31);;; (1, 5, 2, 6) \end{align} P.S. It's nice that its $r_i$ are still relatively small compared to the positive discriminant $d=2857$ from the previous post. – Tito Piezas III Jan 04 '25 at 19:26
  • WolframAlpha has $d=-3321607$ (prime) with class number $h(d)=27\times21=567$ and $d = -653329427$ (not prime) with $h(d)=27\times210 = 5670$. Interesting that the second $h(d)$ is a multiple of the first. – Tito Piezas III Jan 04 '25 at 19:39
  • I find $h(-4027) = 9$ has ALL $9$ of its reduced quadratic forms with multiplicative closure. There are also $h(d) = 18$ with $9$ such forms. Kindly see this new MO question. – Tito Piezas III Jan 07 '25 at 09:22