Number of matrices with unit determinant and fixed sum of elements

Question

Question. Let $\mathcal{M}_3$ be the set of $3\times 3$ matrices with non-negative integer entries and unit determinant. What is the number of $M\in \mathcal{M}_3$ with fixed sum of entries? What is the answer if we allow only matrices with strictly positive entries?

The rate of growth is also interesting. Below $n$ stands for the sum of entries.

Origin. For the similar question in the case of $2\times 2$ matrices we get $\phi(n)$ and $\phi(n) - 2$ respectively, see this recent post: Decomposition of a natural number as sum of positive integers.

Computation of the first elements. If we denote the answer sequence for non-negative case by $a_n$ then the first elements are

$n$	1	2	3	4	5	6	7	8	9	10	11	12	13	14
$a_n$	0	0	3	18	54	126	261	432	783	1134	1899	2286	3960	4680

$a = [0, 0, 0, 3, 18, 54, 126, 261, 432, 783, 1134, 1899, 2286, 3960, 4680, \ldots]$.

And if $b_n$ is the answer for the strictly positive case then $b_n = 0$ if $n \le 10$ and further we have

$n$	11	12	13	14	15	16	17	18	19	20	21	22	23	24
$b_n$	9	18	72	108	234	360	747	756	1818	1782	3222	3672	6615	5850

$b = [0,\ldots ,0, 9, 18, 72, 108, 234, 360, 747, 756, 1818, 1782, 3222, 3672, 6615, \ldots]$.

Unfortunately, both sequences do not seem to be in OEIS.

Thoughts on the growth. Evidently, $a_n\ge b_n$, and $a_n\le Cn^8$ because the number of partitions of $n$ into $9$ terms is $O(n^8)$. If we consider only upper-triangular matrices with ones on the diagonal, we will get the bound $a_n\ge Cn^2$.

Thus, the answer is of polynomial order. Note that this is different to the $2\times 2$ case where the number-theoretic properties played the main role.

Answer 1

(A comment rather than an answer.)

Here is a plot of $a_n/n^5$ (red) and $b_n/n^5$ (blue). It might not go far enough to show the asymptotic behaviour, but a possibility is that $a_n$ and $b_n$ are asymptotically the same and of order $\Theta(n^5)$.

Here is a plot of $a_n/b_n$ with odd $n$ in red and even $n$ in blue.

Values of $a_n$: 3, 18, 54, 126, 261, 432, 783, 1134, 1899, 2286, 3960, 4680, 6876, 8262, 12654, 12618, 20799, 20934, 30024, 32760, 48141, 43632, 68976, 68094, 91161, 93042, 138006, 112194, 187227, 170982, 224892, 226728, 310824, 265770, 418410, 372384, 484920, 455400, 677160, 520596, 839727, 726300, 905580, 900864, 1267065, 984474, 1528875, 1275318, 1680426, 1573398, 2227860, 1699722, 2558106, 2197980, 2829744, 2632266, 3709305, 2675448, 4336128, 3607416, 4446072, 4205142, 5623002, 4314492, 6752907, 5547510, 6989796, 6022962, 8947773, 6542532, 10176480, 8324190, 9964224, 9450396, 12778866, 9518256, 14843745, 11591730, 15006681, 13557816, 18773721, 13365792, 20262222, 17017884, 21061980, 18806256, 26303922, 18207054, 28574676, 23444388, 28962558, 26087202, 34559550, 25770906, 39755196, 31209228, 38935332, 33723702

Values of $b_n$: 9, 18, 72, 108, 234, 360, 747, 756, 1818, 1782, 3222, 3672, 6615, 5850, 11394, 11034, 16623, 17028, 30204, 22248, 45792, 39204, 56853, 57906, 87984, 72036, 128160, 108990, 154890, 141444, 236412, 167346, 306909, 253674, 334980, 332100, 503361, 369648, 636408, 505800, 707481, 646290, 988488, 712944, 1166760, 966708, 1306692, 1190376, 1797597, 1220004, 2150172, 1725192, 2204820, 2063214, 2893518, 2125296, 3564243, 2823642, 3708594, 3112686, 4920642, 3420018, 5671881, 4493988, 5531238, 5193000, 7316118, 5233482, 8656461, 6531660, 8762139, 7774002, 11231289, 7679628, 12234762, 10021086, 12747240, 11188980, 16320924, 10835082, 17876448, 14286942, 18103320, 16096968, 22026474, 15879564, 25652511, 19605996, 25073073, 21351708, 31695534, 21705012, 35093133, 27260136, 32296518, 30280140, 42823683, 29473992, 47349288

Answer 2

$a_n$ is odd if and only if $n>1$ is a power of a prime $\equiv3\pmod4$. To see this, note that adding the constraints $M=M^T$ and $WMW=M$ for $W=\left(\begin{smallmatrix}&&1\\&1&\\1&&\end{smallmatrix}\right)$ does not change the parity of the number of solutions. With the added constraints you can work out the number exactly (it's always $0$, $2^{\omega(n)}$ or $2^{\omega(n)-1}$), and the only case that gives an odd number is $n=p^k$ with $p\equiv3\pmod4$.

Edit: $a_n-b_n$ is odd iff $n>1$ is of the form $2x^2\pm1$.