"Polyadic sigma matrices" by S. Duplij, arxiv: 2403.19361

arXiv:2403.19361v1
[math.GR]
28
Mar
2024
POLYADIC SIGMA MATRICES
STEVEN DUPLIJ
University of Münster, D-48149 Münster, Germany
ABSTRACT. We generalize σ-matrices to higher arities using the polyadization procedure proposed by the
author. We build the nonderived n-ary version of SU p2q using cyclic shift block matrices. We define a
new function, the polyadic trace, which has an additivity property analogous to the ordinary trace for block
diagonal matrices and which can be used to build the corresponding invariants. The elementary Σ-matrices
introduced here play a role similar to ordinary matrix units, and their sums are full Σ-matrices which can
be treated as a polyadic analog of σ-matrices. The presentation of n-ary SU p2q in terms of full Σ-matrices
is done using the Hadamard product. We then generalize the Pauli group in two ways: for the binary case
we introduce the extended phase shifted σ-matrices with multipliers in cyclic groups of order 4q (q ą 4),
and for the polyadic case we construct the correspondent finite n-ary semigroup of phase-shifted elementary
Σ-matrices of order 4q pn ´ 1q ` 1, and the finite n-ary group of phase-shifted full Σ-matrices of order
4q. Finally, we introduce the finite n-ary group of heterogeneous full Σhet
-matrices of order p4q pn ´ 1qq
4
.
Some examples of the lowest arities are presented.
CONTENTS
1. INTRODUCTION 2
2. PRELIMINARIES 2
3. POLYADIC SU p2q 3
4. POLYADIC ANALOG OF SIGMA MATRICES 6
4.1. Elementary Σ-matrices 6
4.2. Full Σ-matrices 7
5. TERNARY SUp2q AND Σ-MATRICES 9
6. n-ARY SEMIGROUPS AND GROUPS OF Σ-MATRICES 12
6.1. The Pauli group 12
6.2. Groups of phase-shifted sigma matrices 13
6.3. The n-ary semigroup of elementary Σ-matrices 14
6.4. The n-ary group of full Σ-matrices 15
7. HETEROGENEOUS FULL Σ-MATRICES 16
REFERENCES 19
E-mail address: douplii@uni-muenster.de; sduplij@gmail.com; http://www.uni-muenster.de/IT.StepanDouplii.
Date: of start February 12, 2024. Date: of completion March 27, 2024.
Total: 15 references.
2010 Mathematics Subject Classification. 20B05, 20H20, 20N10, 20N15, 20M10.
Key words and phrases. sigma matrix, Pauli matrix, Pauli group, arity, n-ary semigroup, polyadic group, ternary group,
n-ary group, querelement, finite semigroup, finite group, cyclic group, cyclic shift matrix, Hadamard product.

1. INTRODUCTION
The role of sigma matrices (σ-matrices, Pauli matrices) in mathematical physics is hard to overestimate.
For a review, see, e.g. FEYNMAN ET AL. [1965], SCHIFF [1968], LIBOFF [2002]. Here we will generalize
σ-matrices to higher arities by using the polyadization procedure proposed in DUPLIJ [2022].
First, we recall the connection between σ-matrices and SU p2q, and then construct an n-ary version
of them using cyclic block shift matrices along the lines of DUPLIJ [2022]. We define a new function,
a polyadic trace, which has an additivity property analogous to the ordinary trace for block diagonal
matrices, and can be used to build the corresponding invariants. The elementary Σ-matrices introduced
here play a role similar to ordinary matrix units, and their sums are full Σ-matrices which can be treated
as a polyadic analog of ordinary (binary) σ-matrices. The presentation of n-ary SU p2q in terms of full
Σ-matrices is done by means of the Hadamard product.
Second, we consider generalizations of the Pauli group (see, e.g. NIELSEN AND CHUANG [2000],
KIBLER [2009]): for the binary case we introduce extended phase shifted σ-matrices with multipliers
in cyclic groups of order 4q (q ą 4), for the polyadic case we construct the corresponding finite n-ary
semigroup of phase-shifted elementary Σ-matrices of order 4q pn ´ 1q ` 1, and a finite n-ary group of
phase-shifted full Σ-matrices of order 4q.
Finally, we introduce the finite n-ary group of heterogeneous full Σhet
-matrices of order p4q pn ´ 1qq4
,
and some illustrative examples in the lowest arities are presented.
2. PRELIMINARIES
First, to establish notation we briefly recall some well-known facts in language convenient for further
generalizations to higher arity. Consider the special unitary group SU p2q represented by 2 ˆ 2 complex
(3-parameter) matrices of the form
M “ Mpx0, Ý
Ñ
x q “
ˆ
x0 ` ix1 x2 ` ix3
´x2 ` x3i x0 ´ x1i
˙
, (2.1)
det M “ x2
0 ` Ý
Ñ
x 2
“ 1, x0, x1, x2, x3 P R, (2.2)
such that
M:
M “ MM:
“ I2, (2.3)
where p:q is the Hermitian (conjugate) transpose (here M:
“ M´1
), and Ý
Ñ
x “ px1, x2, x3q.
The binary product of the matrices (2.1) is closed and in terms of the parameters is
Mpx1
0, Ý
Ñ
x 1
qMpx2
0, Ý
Ñ
x 2
q “ Mpx0, Ý
Ñ
x q, (2.4)
x12
0 ` Ý
Ñ
x 12
“ x22
0 ` Ý
Ñ
x 22
“ 1, x1
0, x1
1, x1
2, x1
3, x2
0, x2
1, x2
2, x2
3 P R, (2.5)
where
x0 “ x1
0x2
0 ´ x1
1x2
1 ´ x1
2x2
2 ´ x1
3x2
3, x1 “ x1
1x2
0 ` x1
0x2
1 ` x1
2x2
3 ´ x1
3x2
2, (2.6)
x2 “ x1
2x2
0 ` x1
0x2
2 ` x1
3x2
1 ´ x1
1x2
3, x3 “ x1
3x2
0 ` x1
0x2
3 ` x1
1x2
2 ´ x1
2x2
1. (2.7)
The standard (binary) σ-matrices (Pauli matrices)
σ0 “
ˆ
1 0
0 1
˙
” I2 ” 1, σ1 “
ˆ
0 1
1 0
˙
, σ2 “
ˆ
0 ´i
i 0
˙
, σ3 “
ˆ
1 0
0 ´1
˙
(2.8)
are involutory (reflections)
σ2
j “ I2, σ´1
j “ σj, (2.9)
– 2 –

PRELIMINARIES
and they satisfy the (binary) commutation and anti-commutation relations
rσj, σks “ rσj, σksr2s
“ σjσk ´ σkσj “ 2iǫjklσl, j, k, l “ 1, 2, 3, (2.10)
tσj, σku “ tσj, σkur2s
“ σjσk ` σkσj “ 2δjkI2, (2.11)
where ǫjkl is the Levi-Civita (permutation) symbol in three dimensions, while r , sr2s
and t , ur2s
are
the ordinary (binary) commutator and anti-commutator (here we denote arity by bold numbers in square
brackets).
In terms of σ-matrices the group element of SU p2q has the following presentation (see (2.1))
M “ Mpx0, Ý
Ñ
x q “ x0σ0 ` iÝ
Ñ
x Ý
Ñ
σ . (2.12)
The binary SU p2q invariant Igen can be obtained using the ordinary trace in the standard way
Igen “
1
2
tr
`
M:
px1
0, Ý
Ñ
x 1
qMpx2
0, Ý
Ñ
x 2
q
˘
“ x1
0x2
0 ` Ý
Ñ
x 1Ý
Ñ
x 2
. (2.13)
The invariant in the form (2.13) will allow us to obtain its analog for higher arities.
3. POLYADIC SU p2q
Let us apply the matrix polyadization procedure to SU p2q (2.1) following DUPLIJ [2022]. Consider
the set of pn ´ 1q complex 2 ˆ 2 matrices
Mpkq
“ Mpx
pkq
0 , Ý
Ñ
x pkq
q “
˜
x
pkq
0 ` ix
pkq
3 ix
pkq
1 ` x
pkq
2
ix
pkq
1 ´ x
pkq
2 x
pkq
0 ´ ix
pkq
3
¸
“ x
pkq
0 σ0 ` iÝ
Ñ
x pkqÝ
Ñ
σ , (3.1)
Mpkq:
Mpkq
“ 1, det Mpkq
“
´
x
pkq
0
¯2
`
´
Ý
Ñ
x pkq
¯2
“ 1, k “ 1, . . . n ´ 1, (3.2)
such that each one represents the group SU p2q (see (2.1)). For clarity, we denote the 2 ˆ 2 identity and
zero matrices by 1 and 0.
Definition 3.1. The polyadic (n-ary) special linear 3 pn ´ 1q parameter group SUrns
p2q is isomorphic to
the set of 2 pn ´ 1q ˆ 2 pn ´ 1q cyclic shift block matrices of the form DUPLIJ [2022]
M “ M2pn´1qˆ2pn´1q “
¨
˚
˚
˚
˚
˚
˝
0 Mp1q
. . . 0 0
0 0 Mp2q
. . . 0
0 0
...
...
.
.
.
.
.
.
.
.
.
... 0 Mpn´2q
Mpn´1q
0 . . . 0 0
˛
‹
‹
‹
‹
‹
‚
, (3.3)
with respect to the n-ary multiplication (as the ordinary 2 pn ´ 1q ˆ 2 pn ´ 1q-matrix product p¨q)
µrns
»
–
n
hkkkkkkkkkikkkkkkkkkj
M1
, M2
, . . . , M3
ﬁ
ﬂ “
n
hkkkkkkkkkikkkkkkkkkj
M1
¨ M2
. . . ¨ M3
“ M, (3.4)
which is non-derived (i.e. not closed for fewer than n multipliers). The allowed number of M’s in any
polyadic product is ℓ pn ´ 1q ` 1, ℓ P N.
– 3 –

POLYADIC SU p2q
Sometimes, we omit µrns
, if the arity is clear from the context.
In terms of SU p2q blocks Mpkq
(3.3) the product (3.4) is given by the pn ´ 1q cycled products
n
hkkkkkkkkkkkkkkkkkkikkkkkkkkkkkkkkkkkkj
Mp1q1
Mp2q2
. . . Mpn´1q3
Mp1q4
“ Mp1q
, (3.5)
n
hkkkkkkkkkkkkkkkkikkkkkkkkkkkkkkkkj
Mp2q1
Mp3q2
. . . Mp1q3
Mp2q4
“ Mp2q
, (3.6)
.
.
. (3.7)
n
hkkkkkkkkkkkkkkkkkkkkkikkkkkkkkkkkkkkkkkkkkkj
Mpn´1q1
Mp1q2
. . . Mpn´2q3
Mpn´1q4
“ Mpn´1q
. (3.8)
The n-ary analog of the binary inverse is the querelement Ă
M DÖRNTE [1929] defined by
µrns
»
—
–
n
hkkkkkkkikkkkkkkj
M, M, . . . , Ă
M
fi
ffi
fl “ M, (3.9)
where Ă
M can be on any place. The manifest form of Ă
M can be obtained from (3.5)–(3.9) as (taking into
account that M´1
“ M:
, M P SU p2q)
Ă
M “
¨
˚
˚
˚
˚
˚
˚
˚
˝
0 Mpn´1q: . . . Mp3q:Mp2q: . . . 0 0
0 0 Mp1q: . . . Mp4q:Mp3q: . . . 0
0 0
...
...
.
.
.
.
.
.
.
.
.
... 0 Mpn´3q: . . . Mp1q:Mpn´1q:
Mpn´2q: . . . Mp2q:Mp1q: 0 . . . 0 0
˛
‹
‹
‹
‹
‹
‹
‹
‚
. (3.10)
Thus, we have
Definition 3.2. The set of matrices (3.3) together with the n-ary product (3.4) and querelement (3.10)
defines a polyadic (n-ary) special unitary group
SUrns
p2q “
A
tMu | µrns
, Ă
p q
E
. (3.11)
The existence of the polyadic identity in an n-ary group is not necessary, as only the querelement is
important in this context DÖRNTE [1929]. For the n-ary group SUrns
p2q the situation is the opposite: the
polyadic identity is not unique. Indeed, recall that the left El, middle Em (on n ´ 2 places) and right Er
polyadic identities are defined by
µrns
»
–
n
hkkkkkkkikkkkkkkj
El, M, . . . , M
fi
fl “ M, (3.12)
µrns
»
–
n
hkkkkkkkikkkkkkkj
M, Em, . . . , M
fi
fl “ M, (3.13)
µrns
»
–
n
hkkkkkkkikkkkkkkj
M, . . . , M, Er
fi
fl “ M, @M P SUrns
p2q , (3.14)
respectively.
– 4 –

POLYADIC SU p2q
Assertion 3.3. The n-ary group SUrns
p2q has sets of left and right identities of the form
El paq “
¨
˚
˚
˚
˚
˚
˝
0 ap1q
I2 . . . 0 0
0 0 ap2q
I2 . . . 0
0 0
...
...
.
.
.
.
.
.
.
.
.
... 0 apn´2q
I2
apn´1q
I2 0 . . . 0 0
˛
‹
‹
‹
‹
‹
‚
, (3.15)
ap1q
ap2q
. . . apn´1q
“ 1, apkq
P Rz t0u , (3.16)
Er pbq “
¨
˚
˚
˚
˚
˚
˝
0 bp1q
I2 . . . 0 0
0 0 bp2q
I2 . . . 0
0 0
...
...
.
.
.
.
.
.
.
.
.
... 0 bpn´2q
I2
bpn´1q
I2 0 . . . 0 0
˛
‹
‹
‹
‹
‹
‚
, (3.17)
bp1q
bp2q
. . . bpn´1q
“ 1, bpkq
P Rz t0u , (3.18)
where both apkq
and bpkq
, k “ 1, . . . , n ´ 1, are different sets of pn ´ 2q non-zero reals (satisfying the
additional conditions (3.16) and (3.18)), and there are no middle identities at all. They form two different
pn ´ 2q parameter subgroups of SUrns
p2q.
Remark 3.4. In the case when all apkq
“ bpkq
“ 1, both identities coincide
El p1q “ Er p1q “ Erns
“
¨
˚
˚
˚
˚
˚
˝
0 I2 . . . 0 0
0 0 I2 . . . 0
0 0
...
...
.
.
.
.
.
.
.
.
.
... 0 I2
I2 0 . . . 0 0
˛
‹
‹
‹
‹
‹
‚
, (3.19)
and when E is in one of the middle places in (3.13), it permutes the internal blocks of (3.3), and therefore
E is not a middle polyadic identity. Nevertheless, if all the blocks are equal Mpkq
“ M, then E is also a
middle identity, or simply the polyadic identity of SUrns
p2q.
For the cyclic shift block matrices of the form (3.3), the determinant is
det M “ p´1qn´1
det Mp1q
. . . det Mpn´1q
, (3.20)
which for Mpkq
P SU p2q becomes det M “ p´1qn´1
.
Consider a m pn ´ 1q ˆ m pn ´ 1q block-diagonal matrix A “ diag pA1 . . . An´1q, where Ak are
pn ´ 1q matrices of the size m ˆ m. The connection between trace and determinant of any matrix A
is
det eA
“ etr A
, (3.21)
which in case of the block-diagonal matrix becomes
det eA
“ etr A1
. . . etr An´1
“ etr A1`...`tr An´1
, (3.22)
and for the block-diagonal matrix A we have
det A “ det A1 . . . det An´1 (3.23)
tr A “ tr A1 ` . . . ` tr An´1 (3.24)
– 5 –

POLYADIC SU p2q Elementary Σ-matrices
Note that the ordinary trace of the cyclic shift block matrix (3.3) is identically zero tr M “ 0. Compar-
ing (3.23) with (3.20) we observe that both cyclic shift block matrix M and the block-diagonal matrix A
are monomial block matrices. By analogy with (3.24), we propose a new function
Definition 3.5. A polyadic (or n-ary) trace Trprns
: SUrns
p2q Ñ R is defined for the 2 pn ´ 1qˆ2 pn ´ 1q
cyclic shift block matrix (3.3) by
Trprns
M “ tr Mp1q
` . . . ` tr Mpn´1q
. (3.25)
Remark 3.6. In this case we see that the invariance properties of the polyadic trace remain the same as for
the ordinary trace for a block-diagonal matrix. This is true for any monomial block matrix, for which we
can use (3.24), (3.25).
Remark 3.7. In the binary case (n “ 2) we have M “ Mp1q
, and the polyadic trace coincides with the
ordinary trace
Trpr2s
M “ tr Mp1q
. (3.26)
Example 3.8. The polyadic identities (3.15)–(3.17) are traceless tr El paq “ tr Er pbq “ 0, while their
polyadic traces are different
Trprns
El paq “ 2
`
ap1q
` . . . ` apn´1q
˘
, Trprns
Er pbq “ 2
`
bp1q
` . . . ` bpn´1q
˘
. (3.27)
4. POLYADIC ANALOG OF SIGMA MATRICES
Next we present the SUrns
p2q matrices (3.3) in a form similar to (2.12) using (3.1). We insert (3.1) into
(3.3) and get
M “
¨
˚
˚
˚
˚
˚
˚
˚
˚
˝
0 x
p1q
0 σ0 ` iÝ
Ñ
x p1qÝ
Ñ
σ . . . 0 0
0 0 x
p2q
0 σ0 ` iÝ
Ñ
x p2qÝ
Ñ
σ . . . 0
0 0
...
...
.
.
.
.
.
.
.
.
.
... 0 x
pn´2q
0 σ0 ` iÝ
Ñ
x pn´2qÝ
Ñ
σ
x
pn´1q
0 σ0 ` iÝ
Ñ
x pn´2qÝ
Ñ
σ 0 . . . 0 0
˛
‹
‹
‹
‹
‹
‹
‹
‹
‚
. (4.1)
The ordinary trace of M is obviously zero, while the polyadic trace (3.25) is
Trprns
M “ 2
´
x
p1q
0 ` . . . ` x
pn´1q
0
¯
. (4.2)
4.1. Elementary Σ-matrices. To obtain an analog of the decomposition (2.12) for M (3.3), (4.1), we
should use cyclic shift block matrices constructed from σ-matrices.
Definition 4.1. The 4 pn ´ 1q polyadic elementary Σ-matrices of the size 2 pn ´ 1qˆ2 pn ´ 1q are defined
as the cyclic shift block matrices (3.3) containing only one non-zero 2ˆ2 block, in which a corresponding
σ-matrix is placed
Σ
pkq
j “
¨
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˝
0 0 p1q
0 . . . 0 . . . 0
0 0 0 p2q
. . . 0 . . . 0
0 0 0
...
.
.
. . . .
.
.
.
0 0 0 . . . σj
pkq
. . . 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.
.
0 0 0 . . . 0 . . . 0 pn´2q
0 pn´1q
0 0 . . . 0 . . . 0
˛
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‚
, (4.3)
– 6 –

Full Σ-matrices POLYADIC ANALOG OF SIGMA MATRICES
where σj is in kth block on the diagonal.
Theorem 4.2. The SUrns
p2q matrix can be represented as an expansion in the polyadic elementary Σ-
matrices by
M “ x
p1q
0 Σ
p1q
0 `iÝ
Ñ
x p1qÝ
Ñ
Σ
p1q
` x
p2q
0 Σ
p2q
0 ` iÝ
Ñ
x p2qÝ
Ñ
Σ
p2q
` . . . ` x
pn´1q
0 Σ
pn´1q
0 ` iÝ
Ñ
x pn´1qÝ
Ñ
Σ
pn´1q
, (4.4)
Ý
Ñ
Σ
pkq
“
´
Σ
pkq
1 , Σ
pkq
2 , Σ
pkq
3
¯
,
´
x
pkq
0
¯2
`
´
Ý
Ñ
x pkq
¯2
“ 1, k “ 1, . . . n ´ 1. (4.5)
Proof. The presentation (4.4) follows from the exact form (4.1), the definition of the Σ-matrix (4.3),
linearity and properties of the scalar multiplication.
Proposition 4.3. Each elementary Σ-matrix (4.3) is non-invertible, but the sum (4.4) gives an invertible
SUrns
p2q matrix.
Proof. If one block in the cycled block shift matrix has zero determinant, the whole matrix becomes non-
invertible, because of the product of determinants (3.20). Following the sphere S3 condition (4.5) for each
block
´
x
pkq
0
¯2
`
´
x
pkq
1
¯2
`
´
x
pkq
2
¯2
`
´
x
pkq
3
¯2
“ 1, at least one x
pkq
i ‰ 0, and so each block in (4.1) is
invertible (Mpkq
P SU p2q), then the whole matrix M is invertible.
Another form of the polyadic elementary Σ-matrices can be given in terms of a product of a vector-
column and a vector-row. Indeed, introduce two pn ´ 1q ˆ 1 vectors
Vpkq
“
¨
˚
˚
˚
˚
˚
˚
˝
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
0
.
.
.
I2
pkq
.
.
.
0
,
/
/
/
/
/
/
.
/
/
/
/
/
/
-
pn ´ 1q
˛
‹
‹
‹
‹
‹
‹
‚
, S
pkq
j “
¨
˚
˚
˚
˚
˚
˚
˝
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
ˇ
0
.
.
.
σj
pkq
.
.
.
0
,
/
/
/
/
/
/
.
/
/
/
/
/
/
-
pn ´ 1q
˛
‹
‹
‹
‹
‹
‹
‚
, j “ 0, 1, 2, 3, (4.6)
then we obtain matrices with one nonzero block (block matrix units)
Σ
pkq
j “ Vpkq
¨
´
S
pk`1q
j
¯T
, k “ 1, . . . , n ´ 2, Σ
pn´1q
j “ Vpn´1q
¨
´
S
p1q
j
¯T
, (4.7)
which coincide with (4.3).
Remark 4.4. As opposed to the binary case and σ-matrices, the elementary Σ-matrices (4.3) do not belong
to SUrns
p2q, because they are non-invertible, nilpotent and det Σ
pkq
j “ 0.
4.2. Full Σ-matrices. Note that from (3.19) and (4.3) it follows that the sum of elementary Σ-matrices
Σ
p1q
0 ` Σ
p2q
0 ` . . . ` Σ
pn´1q
0 “ E PSUrns
p2q , (4.8)
is invertible, because det E “ 1, and therefore, the sum is in SUrns
p2q. This allows us to make
Definition 4.5. The full Σ-matrices Σj are defined by the sums of elementary Σ-matrices
Σ
p1q
j ` Σ
p2q
j ` . . . ` Σ
pn´1q
j “ Σj, j “ 0, 1, 2, 3. (4.9)
– 7 –

POLYADIC ANALOG OF SIGMA MATRICES Full Σ-matrices
In manifest form the full Σ-matrices are
Σj “
¨
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˝
0 σj
p1q
0 . . . 0 . . . 0
0 0 σj
p2q
. . . 0 . . . 0
0 0 0
...
.
.
. . . .
.
.
.
0 0 0 . . . σj
pkq
. . . 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.
.
0 0 0 . . . 0 . . . σj
pn´2q
σj
pn´1q
0 0 . . . 0 . . . 0
˛
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‚
, j “ 0, 1, 2, 3. (4.10)
In this notation Σ0 “ E from (3.19).
Proposition 4.6. The full Σ-matrices are in the polyadic group SUrns
p2q.
Proof. This follows directly from (4.1) with all x
pkq
j “ 1, j “ 0, 1, 2, 3, k “ 1, . . . , n ´ 1.
Using the full Σ-matrices we could derive the restricted version of (4.4) as (cf. the binary case (2.12))
M “ x0Σ0 ` iÝ
Ñ
x
Ý
Ñ
Σ,
Ý
Ñ
Σ “ pΣ1, Σ2, Σ3q , (4.11)
xj “ x
p1q
j “ x
p2q
j “ . . . “ x
pn´1q
j , j “ 0, 1, 2, 3, (4.12)
x2
0 ` Ý
Ñ
x 2
“ 1. (4.13)
Proposition 4.7. The polyadic matrices (4.11) form a subgroup SUrns
p2q of SUrns
p2q.
Proof. The restricted matrices M (4.11) correspond to the cyclic shift block matrices (3.3) having equal
entries. From the component multiplication rules (3.5)–(3.8), where only one equation remains now, it
follows that the set M
(
is a subgroup of SUrns
p2q.
In terms of the full Σ-matrices the n-ary multiplication in the subgroup SUrns
p2q is given by the single
relation (instead of (3.5)–(3.8))
n
hkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkikkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkj
´
x1
0Σ0 ` i
Ý
Ñ
x1 Ý
Ñ
Σ
¯ ´
x2
0Σ0 ` i
Ý
Ñ
x2Ý
Ñ
Σ
¯
. . .
´
x3
0 Σ0 ` i
Ý
Ñ
x3Ý
Ñ
Σ
¯
“ x0Σ0 ` iÝ
Ñ
x
Ý
Ñ
Σ. (4.14)
To obtain the multiplication in the subgroup, the Cayley table for the full Σ-matrices is needed. Some
obvious n-ary products (products of other lower arities are not closed) are the following
pΣ0qn
“ Σ0, pΣ0qn´1 Ý
Ñ
Σ “
Ý
Ñ
Σ, (4.15)
pΣjqn
“

Σ0, if n is even,
Σj, if n is odd.
, j “ 1, 2, 3. (4.16)
Therefore we have
Assertion 4.8. For odd arity all full Σ-matrices are n-ary idempotents, while for even arity Σj are n-ary
reflections, because Σ0 ” E.
Then to find the n-ary multiplication (4.14) in parameters xj, we need the expressions of σ-matrices up
to n, because of the exact form (4.10) of full Σ-matrices.
– 8 –

POLYADIC ANALOG OF SIGMA MATRICES
Example 4.9. For the case n “ 4 we have quaternary full Σ-matrices
Σ0 “
¨
˝
0 I2 0
0 0 I2
I2 0 0
˛
‚, Σj “
¨
˝
0 σj 0
0 0 σj
σj 0 0
˛
‚, j “ 1, 2, 3, (4.17)
such that only the product of 4 of them is closed (and of polyadic power ℓ, having 3ℓ ` 1 multipliers), and
Σj are 4-ary reflections, since pΣjq4
“ Er4s
“ E (see (3.19)).
We now try to construct the full polyadic group SUrns
p2q using the Hadamard product (d) (or element-
wise product, Schur product). Introduce the 2 pn ´ 1q ˆ 2 pn ´ 1q cyclic shift block matrix of parameters
of SUrns
p2q (see (3.1))
Xj “
¨
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˝
0 x
p1q
j I2 0 . . . 0 . . . 0
0 0 x
p2q
j I2 . . . 0 . . . 0
0 0 0
...
.
.
. . . .
.
.
.
0 0 0 . . . x
pkq
j I2 . . . 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.
.
0 0 0 . . . 0 . . . x
pn´2q
j I2
x
pn´1q
j I2 0 0 . . . 0 . . . 0
˛
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‚
, j “ 0, 1, 2, 3. (4.18)
´
x
pkq
0
¯2
`
´
Ý
Ñ
x pkq
¯2
“ 1, k “ 1, . . . n ´ 1. (4.19)
Then, instead of (4.4) and (4.11) we have
Theorem 4.10. The full polyadic group SUrns
p2q matrix (4.1) can be represented as an expansion in the
polyadic full Σ-matrices by
M “ X0 d Σ0 ` iX1 d Σ1 ` iX2 d Σ2 ` iX3 d Σ3, (4.20)
where pdq is the Hadamard product.
Proof. This follows from (4.1), (4.10) and (4.18), in both directions.
Note that (4.20) can be treated as a polyadic analog of the “multiplication by scalars” (some kind of
module) which is consistent with the cyclic shift block matrix structure.
5. TERNARY SUp2q AND Σ-MATRICES
Let us consider the smallest arity case n “ 3. In manifest form an element of the 6-parameter matrix
ternary group SUr3s
p2q is (cf. (2.1))
M “
¨
˚
˚
˚
˝
0 0 x
p1q
0 ` ix
p1q
3 ix
p1q
1 ` x
p1q
2
0 0 ix
p1q
1 ´ x
p1q
2 x
p1q
0 ´ ix
p1q
3
x
p2q
0 ` ix
p2q
3 ix
p2q
1 ` x
p2q
2 0 0
ix
p2q
1 ´ x
p2q
2 x
p2q
0 ´ ix
p2q
3 0 0
˛
‹
‹
‹
‚
, (5.1)
´
x
p1q
0
¯2
`
´
Ý
Ñ
x p1q
¯2
“
´
x
p2q
0
¯2
`
´
Ý
Ñ
x p2q
¯2
“ 1, x
p1q
j , x
p2q
j P R, j “ 0, 1, 2, 3. (5.2)
The direct multiplication table of SUr3s
p2q in terms of parameters (as (2.6)–2.7)) is too cumbersome to
display here.
– 9 –

TERNARY SUp2q AND Σ-MATRICES
There are eight 4 ˆ 4 ternary elementary Σ-matrices
Σ
p1q
j “
ˆ
0 σj
0 0
˙
, Σ
p2q
j “
ˆ
0 0
σj 0
˙
, j “ 0, 1, 2, 3 (5.3)
which are ternary nilpotent
´
Σ
p1q
j
¯3
“ 0,
´
Σ
p2q
j
¯3
“ 0. (5.4)
We recall the products of two and three σ-matrices
σkσl “ δklσ0 ` iǫklmσm, (5.5)
σkσlσm “ δklσm ´ δkmσl ` δlmσk ` iǫklmσ0. (5.6)
Proposition 5.1. The Cayley table of ternary elementary Σ-matrices (5.3) can be obtained from their
multiplication
µr3s
”
Σ
p1q
k , Σ
p2q
l , Σp1q
m
ı
“ Σ
p1q
k Σ
p2q
l Σp1q
m “ δklΣp1q
m ´ δkmΣ
p1q
l ` δlmΣ
p1q
k ` iǫklmΣ
p1q
0 , (5.7)
µr3s
”
Σ
p2q
k , Σ
p1q
l , Σp2q
m
ı
“ Σ
p2q
k Σ
p1q
l Σp2q
m “ δklΣp2q
m ´ δkmΣ
p2q
l ` δlmΣ
p2q
k ` iǫklmΣ
p2q
0 , (5.8)
while other products give zero 4 ˆ 4 matrix 0.
Proof. We have in components
Σ
p1q
k Σ
p2q
l Σp1q
m “
ˆ
0 σkσlσm
0 0
˙
, (5.9)
Σ
p2q
k Σ
p1q
l Σp2q
m “
ˆ
0 0
σkσlσm 0
˙
, (5.10)
and then use (5.6).
We introduce the ternary commutator and anti-commutator by analogy with (2.10)-(2.11)
ra, b, csr3s
“ abc ` bca ` cab ´ acb ´ bac ´ cba, (5.11)
ta, b, cur3s
“ abc ` bca ` cab ` acb ` bac ` cba, (5.12)
where a, b, c are Σ-matrices. Using (5.7)–(5.8), for elementary Σ-matrices we find that ternary commuta-
tors are proportional to the corresponding ternary units (cf. (2.10)–(2.11))
”
Σ
p1q
k Σ
p1q
l Σp2q
m
ır3s
“
”
Σ
p1q
k Σ
p2q
l Σp1q
m
ır3s
“
”
Σ
p2q
k Σ
p1q
l Σp1q
m
ır3s
“ 2iǫklmΣ
p1q
0 , (5.13)
”
Σ
p2q
k Σ
p2q
l Σp1q
m
ır3s
“
”
Σ
p2q
k Σ
p1q
l Σp2q
m
ır3s
“
”
Σ
p1q
k Σ
p2q
l Σp2q
m
ır3s
“ 2iǫklmΣ
p2q
0 , (5.14)
and the ternary anti-commutators are
!
Σ
p1q
k Σ
p1q
l Σp2q
m
)r3s
“ ´2δklΣp1q
m ` 2δkmΣ
p1q
l ` 2δlmΣ
p1q
k , (5.15)
!
Σ
p1q
k Σ
p2q
l Σp1q
m
)r3s
“ 2δklΣp1q
m ´ 2δkmΣ
p1q
l ` 2δlmΣ
p1q
k , (5.16)
!
Σ
p2q
k Σ
p1q
l Σp1q
m
)r3s
“ 2δklΣp1q
m ` 2δkmΣ
p1q
l ´ 2δlmΣ
p1q
k , (5.17)
– 10 –

TERNARY SUp2q AND Σ-MATRICES
!
Σ
p2q
k Σ
p2q
l Σp1q
m
)r3s
“ ´2δklΣp2q
m ` 2δkmΣ
p2q
l ` 2δlmΣ
p2q
k , (5.18)
!
Σ
p2q
k Σ
p1q
l Σp2q
m
)r3s
“ 2δklΣp2q
m ´ 2δkmΣ
p2q
l ` 2δlmΣ
p2q
k , (5.19)
!
Σ
p1q
k Σ
p2q
l Σp2q
m
)r3s
“ 2δklΣp2q
m ` 2δkmΣ
p2q
l ´ 2δlmΣ
p2q
k , (5.20)
while other (anti)commutators vanish.
The expansion of an SUr3s
p2q matrix (4.4) in terms of the elementary Σ-matrices becomes
M “ x
p1q
0 Σ
p1q
0 ` iÝ
Ñ
x p1qÝ
Ñ
Σ
p1q
` x
p2q
0 Σ
p2q
0 ` iÝ
Ñ
x p2qÝ
Ñ
Σ
p2q
, (5.21)
Ý
Ñ
Σ
p1,2q
“
´
Σ
p1,2q
1 , Σ
p1,2q
2 , Σ
p1,2q
3
¯
. (5.22)
The presentation of M PSUr3s
p2q in terms of the Hadamard product is given by (4.20), where now the
block-matrices of parameters (4.18) are
Xj “
˜
0 x
p1q
j I2
x
p2q
j I2 0
¸
, j “ 0, 1, 2, 3, (5.23)
and the full ternary Σ-matrices are
Σj “ Σ
p1q
j ` Σ
p2q
j “
ˆ
0 σj
σj 0
˙
, j “ 0, 1, 2, 3. (5.24)
Proposition 5.2. The Cayley table of the full ternary Σ-matrices is given by
ΣkΣlΣm “ δklΣm ´ δkmΣl ` δlmΣk ` iǫklmΣ0, (5.25)
ΣkΣlΣ0 “ ΣkΣ0Σl “ Σ0ΣkΣl “ δklΣ0 ` iǫklmΣm, (5.26)
ΣkΣ0Σ0 “ Σ0ΣkΣ0 “ Σ0Σ0Σk “ Σk, k, l, m “ 1, 2, 3. (5.27)
Proof. This follows from (5.7)-(5.8) and (5.24), taking into account the ternary nilpotency (5.4) of the
elementary Σ-matrices.
It follows from (4.15) and (5.25) that all full Σ-matrices are ternary idempotents (cf. (4.16) and (5.4))
pΣjq3
“ Σj, j “ 0, 1, 2, 3. (5.28)
In the same way, from (5.13)–(5.20) we get for full ternary Σ-matrices the ternary commutators (cf.
(2.10)–(2.11))
rΣkΣlΣmsr3s
“ 6iǫklmΣ0, k, l, m “ 1, 2, 3, (5.29)
and ternary anti-commutators
tΣkΣlΣmur3s
“ 2δklΣm ` 2δkmΣl ` 2δlmΣk. (5.30)
Using (4.20), (5.23) and (5.24), we obtain the SUr3s
p2q matrix (cf. (4.1))
M “
˜
0 x
p1q
0 σ0 ` iÝ
Ñ
x p1qÝ
Ñ
σ
x
p2q
0 σ0 ` iÝ
Ñ
x p2qÝ
Ñ
σ 0
¸
. (5.31)
The ternary multiplication in SUr3s
p2q can be derived from the general case (3.5)–(3.8), as follows
x
p1q
0 σ0 ` iÝ
Ñ
x p1qÝ
Ñ
σ “
´
x
p1q1
0 σ0 ` iÝ
Ñ
x p1q1Ý
Ñ
σ
¯ ´
x
p2q2
0 σ0 ` iÝ
Ñ
x p2q2Ý
Ñ
σ
¯ ´
x
p1q3
0 σ0 ` iÝ
Ñ
x p1q3Ý
Ñ
σ
¯
, (5.32)
x
p2q
0 σ0 ` iÝ
Ñ
x p2qÝ
Ñ
σ “
´
x
p2q1
0 σ0 ` iÝ
Ñ
x p2q1Ý
Ñ
σ
¯ ´
x
p1q2
0 σ0 ` iÝ
Ñ
x p1q2Ý
Ñ
σ
¯ ´
x
p2q3
0 σ0 ` iÝ
Ñ
x p2q3Ý
Ñ
σ
¯
. (5.33)
– 11 –

TERNARY SUp2q AND Σ-MATRICES The Pauli group
Using (5.5)–(5.6), we get the exact form of the ternary group SUr3s
p2q multiplication (5.32)–(5.33) in
terms of the group parameters as (cf. (2.6)–(2.7))
x
p1q
0 “ x
p1q1
0 x
p2q2
0 x
p1q3
0 ´ x
p1q1
0
´
Ý
Ñ
x p2q2Ý
Ñ
x p1q3
¯
´ x
p2q2
0
´
Ý
Ñ
x p1q1Ý
Ñ
x p1q3
¯
´ x
p1q3
0
´
Ý
Ñ
x p1q1Ý
Ñ
x p2q2
¯
` Ý
Ñ
x p1q1
´
Ý
Ñ
x p2q2
ˆ Ý
Ñ
x p1q3
¯
, (5.34)
Ý
Ñ
x p1q
“ Ý
Ñ
x p2q2
´
Ý
Ñ
x p1q1
ˆ Ý
Ñ
x p1q3
¯
´ Ý
Ñ
x p1q1
´
Ý
Ñ
x p2q2
ˆ Ý
Ñ
x p1q3
¯
´ Ý
Ñ
x p1q3
´
Ý
Ñ
x p2q2
ˆ Ý
Ñ
x p1q1
¯
´ x
p1q1
0
´
Ý
Ñ
x p2q2
ˆ Ý
Ñ
x p1q3
¯
´ x
p2q2
0
´
Ý
Ñ
x p1q1
ˆ Ý
Ñ
x p1q3
¯
´ x
p1q3
0
´
Ý
Ñ
x p1q1
ˆ Ý
Ñ
x p2q2
¯
, (5.35)
x
p2q
0 “ x
p2q1
0 x
p1q2
0 x
p2q3
0 ´ x
p2q1
0
´
Ý
Ñ
x p1q2Ý
Ñ
x p2q3
¯
´ x
p1q2
0
´
Ý
Ñ
x p2q1Ý
Ñ
x p2q3
¯
´ x
p2q3
0
´
Ý
Ñ
x p2q1Ý
Ñ
x p1q2
¯
` Ý
Ñ
x p2q1
´
Ý
Ñ
x p1q2
ˆ Ý
Ñ
x p2q3
¯
, (5.36)
Ý
Ñ
x p2q
“ Ý
Ñ
x p1q2
´
Ý
Ñ
x p2q1
ˆ Ý
Ñ
x p2q3
¯
´ Ý
Ñ
x p2q1
´
Ý
Ñ
x p1q2
ˆ Ý
Ñ
x p2q3
¯
´ Ý
Ñ
x p2q3
´
Ý
Ñ
x p1q2
ˆ Ý
Ñ
x p2q1
¯
´ x
p2q1
0
´
Ý
Ñ
x p1q2
ˆ Ý
Ñ
x p2q3
¯
´ x
p1q2
0
´
Ý
Ñ
x p2q1
ˆ Ý
Ñ
x p2q3
¯
´ x
p2q3
0
´
Ý
Ñ
x p2q1
ˆ Ý
Ñ
x p1q2
¯
, (5.37)
where Ý
Ñ
x ˆ Ý
Ñ
y is the standard 3D vector product.
The additional one parameter ternary idempotents in SUr3s
p2q are real 4 ˆ 4 matrices (cf. (3.15))
Eid paq “ Er3s
paq “
ˆ
0 aI2
1
a
I2 0
˙
, a P Rz t0u , (5.38)
which are also left and right identities, but not middle identities, according to (3.15)–(3.18).
The querelement of the ternary group SUr3s
p2q “
@
tMu | µr3s
, prq
D
coincides with the matrix inverse
Ă
M “ M´1
(in the ternary case only), as follows from the general formula (3.9).
6. n-ARY SEMIGROUPS AND GROUPS OF Σ-MATRICES
Let us briefly recall the concept of the (binary) group of σ-matrices (or the Pauli group), which is widely
used in the quantum error correction and stabilizer codes (see, e.g. NIELSEN AND CHUANG [2000], BALL
ET AL. [2023], DJORDJEVIC [2021]).
6.1. The Pauli group. The set of four σ-matrices do not form a group, because their multiplication is
not closed: the r.h.s. of (5.5) contains the multipliers, informally σkσl “ sσm, where tsu “ ˘1, ˘i, and
k, l, m “ 1, 2, 3. One approach to obtain a closed multiplication is by enlarging the set by new elements
which arise from consequent products. It is obvious that the enlarged set will be finite, if the multipliers
are cyclic. Indeed, in the case of σ-matrices the multipliers themselves form the cyclic group of order 4
generated by the imaginary unit
tsu “ xiy4 . (6.1)
In this way, the enlarged set contains 16 new elements σ̂j prjq which are constructed from σ-matrices
and powers of i as follows
σ̂j prjq “ irj
σj, j “ 0, 1, 2, 3, rj “ 0, 1, 2, 3, (6.2)
such that σ̂j p0q ” σj (2.8). The set of 2ˆ2 invertible matrices (6.2) (we call them extended σ-matrices or
sigma matrices with phase) is closed under multiplication being associative, as matrix products, moreover
it contains the identity I2 “ σ̂0 p0q, therefore
Ĝσ “
@
tσ̂j prjqu | p¨σq , p q´1
, I2
D
(6.3)
– 12 –

Groups of phase-shifted sigma matrices n-ARY SEMIGROUPS AND GROUPS OF Σ-MATRICES
is a finite binary group of order 16. We call the group Ĝσ a group of phase-shifted sigma matrices, which
is convenient for polyadic extentions. It is also called the Pauli group P1, diquaternion group DQ8, and
by structure it is isomorphic to the central product of a cyclic group C4 of order 4 and the dihedral group
D4 of order 8 as C4 ˝ D4 or to the semi-direct product of the quaternion group Q8 of order 8 and a cyclic
group C2 of order 2 as Q8 ¸2 C2, as well as D4 ¸2 C2, and pC4 ¸2 C2q ¸3 C2 (see, e.g. BAGARELLO
ET AL. [2021], BESCHE ET AL. [2002], CONWAY ET AL. [1986]).
The Cayley table of Ĝσ can be obtained from the manifest multiplication of the phase-shifted σ-matrices
σ̂k prkq ¨σ σ̂l prlq “ δklσ̂0 prk `4 rlq ` |ǫklm| σ̂m prk `4 rl `4 1 `4 p1 ´ ǫklmqq , k, l, m “ 1, 2, 3, (6.4)
where rk `4 rl ” prk ` rlq mod 4, which follows directly from (5.5) on multiplying both sides by irk
irl
.
In (6.4) we introduced the “phase presentation” of the Levi-Civita symbol
ǫklm “ |ǫklm| ei π
2
p1´ǫklmq
, k, l, m “ 1, 2, 3, (6.5)
which is convenient for calculation of the Cayley table for σ-matrices with phase (6.2).
6.2. Groups of phase-shifted sigma matrices. Let us generalize the multiplier construction of σ-
matrices (6.2) in the following way. Instead of the cyclic group xiy4 of order 4, we introduce higher
order cyclic groups which contain xiy4 as the subgroup. For that we rewrite the phase factor in (6.2) as the
4th root of unity
irj
“ ei 2π
q
rj
, q “ 4, rj “ 0, 1, 2, 3. (6.6)
To make the set of q’s finite, we propose that q should be among divisors of 360 which are multiples of
4 (e.g. to have an integer number of degrees, etc.), that is
q P Q12 “ t4, 8, 12, 20, 24, 36, 40, 60, 72, 120, 180, 360u. (6.7)
In this way we obtain 12 cyclic groups of order q in which the first one corresponds to the cyclic group
of order q “ 4 (the Pauli group, in bold) generated by the imaginary unit xiyq“4 (see (6.1) and (6.6)).
Then, now for the fixed q from (6.7), the enlarged set of q-generalized σ-matrices with multipliers from
the cyclic group Cq (with ei 2π
q as the primitive root, instead of the imaginary unit ei π
2 ) will contain q
different phase factors, and therefore we have
Definition 6.1. The enlarged set of the q-generalized phase-shifted σ-matrices (or multiplied by q-th roots
of unity) contains 4q elements
σ̂
pqq
j prjq “ ei 2π
q
rj
σj, j “ 0, 1, 2, 3, rj “ 0, 1, 2, . . . q ´ 1, q P Q12. (6.8)
The multiplication
`
¨σpqq
˘
of the phase-shifted σ-matrices (and the corresponding Cayley table) can be
obtained from (5.5) by multiplying both sides by the factors ei 2π
q
rk
ei 2π
q
rl
and using the “phase presentation”
of the Levi-Civita symbol (6.5), analogous to (6.4)
σ̂
pqq
k prkq ¨σpqq σ̂
pqq
l prlq “ δklσ̂
pqq
0 prk `q rlq ` |ǫklm| σ̂pqq
m
´
rk `q rl `q
q
4
`q
q
4
p1 ´ ǫklmq
¯
,
k, l, m “ 1, 2, 3, rk,l “ 0, 1, 2, . . .q ´ 1, q P Q12, (6.9)
where rk `q rl “ prk ` rlq mod q. The identity now is I2 “ σ̂
pqq
0 p0q. Instead of an involutory (2.9) we
have
σ̂
pqq
k prkq ¨σpqq σ̂
pqq
k prkq “ σ̂
pqq
0 pp2rkq mod qq . (6.10)
Proposition 6.2. The sets of the phase-shifted σ-matrices (6.8) form 12 non-isomorphic finite groups of
order 4q (6.7)
Ĝpqq
σ “
A!
σ̂
pqq
j prjq
)
|
`
¨σpqq
˘
, p q´1
, I2
E
, j “ 0, 1, 2, 3, rj “ 0, 1, 2, . . . q ´ 1, q P Q12. (6.11)
– 13 –

n-ARY SEMIGROUPS AND GROUPS OF Σ-MATRICES The n-ary group of full Σ-matrices
Proof. The set is
!
σ̂
pqq
j prjq
)
closed with respect to the product (6.9), because the factors in (6.8) form a
cyclic group, it contains the identity I2 “ σ̂
pqq
0 p0q, and every element has an inverse.
Note that now the finite binary groups up to the order 2000 are listed BESCHE ET AL. [2001].
6.3. The n-ary semigroup of elementary Σ-matrices. The set of 4 pn ´ 1q elementary Σ-matrices Σ
rks
j
(4.3) is not closed under n-ary multiplication µrns
(3.4) by the same reason as for the set of 4 binary
σ-matrices: the r.h.s. of their product (see (5.7)–(5.8) for ternary case) contains the same multipliers r,
where tru “ ˘1, ˘i, or tru “ xiy4, the cyclic group of order 4. So we will use now the same approach as
in the binary case: by enlarging the set of elementary Σ-matrices.
Definition 6.3. The enlarged set of the q-generalized phase-shifted elementary Σ-matrices contains
4q pn ´ 1q elements (see (6.2), (6.7) and (6.8))
Σ̂
rkspqq
j prjq “ ei 2π
q
rj
Σ
rks
j , j “ 0, 1, 2, 3, rj “ 0, 1, 2, . . . q ´ 1, q P Q12, k “ 1, . . . , n ´ 1. (6.12)
Assertion 6.4. The set
!
Σ̂
rkspqq
j prq
)
Y t0u, where 0 the 4 ˆ 4 matrix with zero entries, is closed with
respect to n-ary multiplication µrns
(3.4).
Proof. This follows from (4.3) that non-zero n-ary products of Σ-matrices are closed, as it is for the
products of n q-generalized extended σ-matrices (6.8). Other products are equal to 0.
Proposition 6.5. The set of the q-generalized phase-shifted n-ary elementary Σ-matrices forms 12 non-
isomorphic finite n-ary semigroups with zero, of order 4q pn ´ 1q ` 1, q P Q12.
Proof. It follows from Assertion 6.4, and total polyadic associativity follows from the fact that the mul-
tiplication of ordinary matrices is associative.
The formulas for n-ary products of q-generalized of Σ-matrices can be written in concrete cases.
Example 6.6 (Semigroup of ternary Σ-matrices). The multiplication of q-generalized phase-shifted ternary
elementary Σ-matrices can be obtained from (5.7)–(5.8) by multiplying both sides by three factors
ei 2π
q
rk
ei 2π
q
rl
ei 2π
q
rm
and use the “phase presentation” of the Levi-Civita symbol (6.5). In this way we get
the q-generalized ternary products
µr3s
q
”
Σ̂
r1spqq
k prkq , Σ̂
r2spqq
l prlq , Σ̂r1spqq
m prmq
ı
“ δklΣ̂r1spqq
m prk `q rl `q rmq ` δlmΣ̂
r1spqq
k prk `q rl `q rmq
` δkmΣ̂
r1spqq
l
´
rk `q rl `q rm `q
q
2
¯
` |ǫklm| Σ̂
r1spqq
0
´
rk `q rl `q rm `q
q
4
`q
q
4
p1 ´ ǫklmq
¯
, (6.13)
µr3s
q
”
Σ̂
r2spqq
k prkq , Σ̂
r1spqq
l prlq , Σ̂r2spqq
m prmq
ı
“ δklΣ̂r2spqq
r2spqq
k prk `q rl `q rmq
` δkmΣ̂
r2spqq
l
´
rk `q rl `q rm `q
q
2
¯
` |ǫklm| Σ̂
r2spqq
0
´
rk `q rl `q rm `q
q
4
`q
q
4
p1 ´ ǫklmq
¯
, (6.14)
while other ones are zero 0. The ternary total associativity follows from the ordinary associativity of
the matrix product (3.4). Therefore, the set of q-generalized phase-shifted ternary elementary Σ-matrices
form a finite ternary semigroup of order 8q ` 1, q P Q12 with the multiplication (6.13)–(6.14)
Ŝ
pqq
Σ “
A!
Σ̂
rkspqq
j prq
)
Y t0u | µr3s
q
E
. (6.15)
Even in the first case q “ 4 the semigroup of ternary elementary Σ-matrices is of 33th order. Nowadays
only binary semigroups up to 10th order are fully listed SLOANE AND PLOUFFE [1995], DISTLER ET AL.
[2012].
– 14 –

n-ARY SEMIGROUPS AND GROUPS OF Σ-MATRICES
6.4. The n-ary group of full Σ-matrices. Let us consider the set of full Σ-matrices (4.10), which is not
closed under n-ary multiplication µrns
(3.4): the r.h.s. of their product contains the same multipliers ˘1,
˘i (see (5.25)–(5.27) for ternary case).Therefore, we should enlarge this set, as for the σ-matrices (6.8).
Definition 6.7. The enlarged set of the q-generalized phase-shifted full Σ-matrices consists of 4q elements
(see (6.2), (6.7) and (6.8))
Σ̂
pqq
j prjq “ ei 2π
q
rj
Σj, j “ 0, 1, 2, 3, rj “ 0, 1, 2, . . . q ´ 1, q P Q12. (6.16)
!
Σ̂
pqq
j prq
)
is closed with respect to n-ary multiplication µrns
(3.4).
Proof. It follows from (4.10) that non-zero n-ary products of Σ-matrices are closed, as it is for the prod-
ucts of n q-generalized extended σ-matrices (6.8) which form a closed set, see (3.5)–(3.8).
To construct an n-ary group we have to find the querelement (3.9)–(3.10) for full Σ-matrices.
Proposition 6.9. The querelement for the q-generalized phase-shifted full Σ-matrix (6.16) is
r
Σ
pqq
j prjq “ ei 2π
q
prj `qp´nqrjq
Σj, (6.17)
where
r
Σj “

Σ0, if arity n is even,
Σj, if arity n is odd.
(6.18)
Proof. We use the equation (3.9) for phase factors and the involutory (2.9) to get (6.18).
Proposition 6.10. The set of the q-generalized phase-shifted n-ary full Σ-matrices forms 12 non-
isomorphic finite n-ary groups of order 4q, q P Q12.
Proof. It follows from Assertion 6.8, with the total polyadic associativity following from the fact that the
multiplication of ordinary matrices is associative, and the querelements are given in (6.17)–(6.18).
The formulas of n-ary multiplication for q-generalized full Σ-matrices are cumbersome, but for some
concrete cases can be written manifestly.
Example 6.11 (Ternary group of full Σ-matrices). We derive the multiplication of q-generalized
phase-shifted ternary full Σ-matrices from (5.25)–(5.27) by multiplying both sides by three factors
ei 2π
q
rk
ei 2π
q
rl
ei 2π
q
rm
and use (6.5). Thus, we get the q-generalized ternary products
µr3s
q
”
Σ̂
pqq
k prkq , Σ̂
pqq
l prlq , Σ̂
pqq
m prmq
ı
“ δklΣ̂
pqq
pqq
k prk `q rl `q rmq
` δkmΣ̂
pqq
l
´
rk `q rl `q rm `q
q
2
¯
` |ǫklm| Σ̂
pqq
0
´
rk `q rl `q rm `q
q
4
`q
q
4
p1 ´ ǫklmq
¯
. (6.19)
The querelement now satisfies ternary involutivity, which follows from the binary version (2.9) and (6.17)
r
Σ
pqq
j “ Σj. (6.20)
The ternary total associativity of the multiplication (6.19) follows from the ordinary associativity of the
matrix product (3.4). Therefore, the set of q-generalized phase-shifted ternary full Σ-matrices form a
finite ternary group of order 4q
Ĝ
r3spqq
Σ “
A!
Σ̂
pqq
j prq
)
| µr3s
q , Ă
p q
E
, q P Q12. (6.21)
In the minimal case q “ 4, the group of ternary full Σ-matrices Ĝ
pqq
Σ is of 16th order, while the Cayley
table contains 163
“ 4096 entries.
– 15 –

HETEROGENEOUS FULL Σ-MATRICES
7. HETEROGENEOUS FULL Σ-MATRICES
Let us consider a further extension of the full Σ-matrices: each entry in (4.10) contains a different
σ-matrix.
Definition 7.1. The heterogeneous n-ary full Σhet
-matrices are multi-index cyclic shift matrices of the
following form
Σhet
“ Σhet
j1j2...jn´1
“
¨
˚
˚
˚
˚
˚
˚
˚
˚
˚
˝
0 σj1 0 . . . 0 . . . 0
0 0 σj2 . . . 0 . . . 0
0 0 0
...
.
.
. . . .
.
.
.
0 0 0 . . . σjk
. . . 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.
.
0 0 0 . . . 0 . . . σjn´2
σjn´1 0 0 . . . 0 . . . 0
˛
‹
‹
‹
‹
‹
‹
‹
‹
‹
‚
, j1, j2, . . . , jn´1 “ 0, 1, 2, 3, (7.1)
and the 4 full Σ-matrices (4.10) can be called homogeneous.
The heterogeneous full Σhet
-matrices can be presented and the sum of the elementary Σ-matrices as
was done for the (homogeneous) full Σ-matrices (4.9)
Σ
p1q
j1
` Σ
p2q
j2
` . . . ` Σ
pn´1q
jn´1
“ Σhet
j1j2...jn´1
, j “ 0, 1, 2, 3. (7.2)
The number of heterogeneous full Σhet
-matrices is pn ´ 1q4
, which follows from the formula for ar-
rangements with repetitions.
The exact multiplication can be written in examples for concrete arities, mostly of the lowest ones.
Example 7.2 (Ternary heterogeneous full Σ-matrices). In the lowest ternary (n “ 3) case Σhet
(7.1) has
two indices
Σhet
j1j2
“
ˆ
0 σj1
σj2 0
˙
, j1, j2 “ 0, 1, 2, 3. (7.3)
There are 24
“ 16 ternary heterogeneous full Σhet
-matrices. The non-derived ternary product (3.4) of
Σhet
-matrices now has the form
Σhet
j1j2
“ µ
r3s
het
”
Σhet
j1
1j1
2
, Σhet
j2
1 j2
2
, Σhet
j3
1 j3
2
ı
“
ˆ
0 σj1
1
σj2
2
σj3
1
σj1
2
σj2
1
σj3
2
0
˙
, j1, j2, j1
1, j1
2, j2
1, j2
2 , j3
1 , j3
2 “ 0, 1, 2, 3,
(7.4)
and we can use the identities for σ-matrices (5.5)–(5.6). There are different cases, depending on the
presence of 0 among the indices j’s (or identities among σ’s):
1) There are no 0’s among all j’s, j1 “ k and j2 “ l, k, l “ 1, 2, 3. Then we use (5.6) to get
µ
r3s
het
“
Σhet
k1l1 , Σhet
k2l2 , Σhet
k3l3
‰
“
ˆ
0 δk1l2 σk3 ´ δk1k3 σl2 ` δl2k3 σk1 ` iǫk1l2k3 σ0
δl1k2 σl3 ´ δl1l3 σk2 ` δk2l3 σl1 ` iǫl1k2l3 σ0 0
˙
. (7.5)
2) There is one 0 among j1
1, j2
2, j3
1 , it is not important which one, and thus we get (e.g. for j3
1 “ 0)
µ
r3s
het
“
Σhet
k1l1 , Σhet
k2l2 , Σhet
0l3
‰
“
ˆ
0 δk1l2 σ0 ` iǫk1l2mσm
δl1k2 σl3 ´ δl1l3 σk2 ` δk2l3 σl1 ` iǫl1k2l3 σ0 0
˙
, m “ 1, 2, 3. (7.6)
– 16 –

In the same way, if one 0 is among j1
2, j2
1, j3
2 , then (e.g. for j3
2 “ 0)
µ
r3s
het
“
Σhet
k1l1 , Σhet
k2l2 , Σhet
k30
‰
“
ˆ
0 δk1l2 σk3 ´ δk1k3 σl2 ` δl2k3 σk1 ` iǫk1l2k3 σ0
δl1k2 σ0 ` iǫl1k2mσm 0
˙
, m “ 1, 2, 3. (7.7)
3) There are two 0’s in different entries of (7.4), e.g. j3
1 “ 0 and j3
2 “ 0, then
µ
r3s
het
“
Σhet
k1l1 , Σhet
k2l2 , Σhet
00
‰
“
ˆ
0 δk1l2 σ0 ` iǫk1l2mσm
δl1k2 σ0 ` iǫl1k2mσm 0
˙
, m “ 1, 2, 3. (7.8)
Note that each non-zero matrix entry in (7.5)–(7.8) is proportional to one σ-matrix multiplied by ˘1,
˘i, as in (5.5)–(5.6). However, these multipliers can be different for different entries, as opposed to the
case of homogeneous Σ-matrices, for which these multipliers coincide, and can be placed as common co-
efficients before matrices (see (5.25)–(5.27)). Therefore, the heterogeneous full Σhet
-matrices of the form
(7.1) are not closed with respect to the n-ary multiplication (3.4), as the four ordinary σ-matrices (5.5).
Nevertheless, the enlarged (in a special way) phase-shifted heterogeneous full Σhet
-matrices (analogous
to (6.8)) can be closed under n-ary multiplication.
Definition 7.3. The enlarged set of the q-generalized element-wise phase-shifted heterogeneous full Σhet
-
matrices have the form
Σhetpqq
“ Σ
hetpqq
j1j2...jn´1
`
rj1 , rj2 , . . . , rjn´1
˘
“
¨
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˚
˝
0 ei 2π
q
rj1 σj1 0 . . . 0 . . . 0
0 0 ei 2π
q
rj2 σj2 . . . 0 . . . 0
0 0 0
...
.
.
. . . .
.
.
.
0 0 0 . . . ei 2π
q
rjk σjk
. . . 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...
.
.
.
0 0 0 . . . 0 . . . ei 2π
q
rjn´2 σjn´2
ei 2π
q
rjn´1 σjn´1 0 0 . . . 0 . . . 0
˛
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‹
‚
, (7.9)
jk “ 0, 1, 2, 3, rjk
“ 0, 1, 2, . . . q ´ 1, q P Q12,
and this set consists of p4q pn ´ 1qq4
elements (see (6.2), (6.7) and (6.8)).
!
Σ
hetpqq
j1j2...jn´1
`
rj1 , rj2 , . . . , rjn´1
˘)
is closed with respect to the non-derived n-ary
multiplication µrns
(3.4).
Proof. This follows from the fact that the products of q-generalized extended σ-matrices (6.8) form a
closed set, see (3.5)–(3.8) for general cyclic shift matrices.
Proposition 7.5. The set of the q-generalized element-wise phase-shifted heterogeneous full Σhet
-
matrices (7.9) is a finite n-ary group of order p4q pn ´ 1qq4
Ĝ
hetrnspqq
Σ “
A!
Σ
hetpqq
j1j2...jn´1
`
rj1 , rj2 , . . . , rjn´1
˘)
| µ
rns
hetpqq, Ă
p q
E
, q P Q12. (7.10)
Proof. The set
!
Σ
hetpqq
j1j2...jn´1
`
rj1 , rj2 , . . . , rjn´1
˘)
is closed by Assertion 7.4. The total polyadic associa-
tivity is governed by the associativity of ordinary matrix multiplication. Each element has its querelement
– 17 –

Č
Σhetpqq
defined by the general formula (3.9), where the entries are of the form (7.9), because of (3.10) and
the fact that the product of n ordinary σ-matrices is proportional to one σ-matrix multiplied by the cyclic
group factor.
Example 7.6 (Ternary group of element-wise phase-shifted heterogeneous full Σhet
-matrices). In the
ternary n “ 3 case, for Σhetpqq
(7.9) we have
Σ
hetpqq
j1j2
prj1 , rj2 q “
˜
0 ei 2π
q
rj1 σj1
ei 2π
q
rj2 σj2 0
¸
j1,2 “ 0, 1, 2, 3, rjk
“ 0, 1, 2, . . .q ´ 1, q P Q12. (7.11)
The order of the ternary group Ĝ
hetr3spqq
Σ is 4096q4
, which for minimal case q “ 4 becomes 1048 576.
The ternary multiplication can be obtained from (7.4)–(7.8) by adding the phase factors
Σ
hetpqq
j1j2
prj1 , rj2 q “ µ
r3s
hetpqq
”
Σhet
j1
1j1
2
`
rj1
1
, rj1
2
˘
, Σhet
j2
1 j2
2
`
rj2
1
, rj2
2
˘
, Σhet
j3
1 j3
2
`
rj3
1
, rj3
2
˘ı
“
¨
˝ 0 e
i 2π
q
´
rj1
1
`rj2
2
`rj3
1
¯
σj1
1
σj2
2
σj3
1
e
i 2π
q
´
rj1
2
`rj2
1
`rj3
2
¯
σj1
2
σj2
1
σj3
2
0
˛
‚, (7.12)
j1, j2, j1
1, j1
2, j2
1, j2
2, j3
1 , j3
2 “ 0, 1, 2, 3, rj1
1,2
, rj2
1,2
, rj3
1,2
“ 0, 1, 2, . . . q ´ 1, q P Q12.
We consider 3 cases as well and use (5.5)–(5.6) with (6.5) to get
1) There are no 0’s among all j’s, j1 “ k and j2 “ l, k, l “ 1, 2, 3. Then we use (5.6) to get
µ
r3s
hetpqq
”
Σ
hetpqq
k1l1 prk1 , rl1 q , Σ
hetpqq
hetpqq
k3l3 prk3 , rl3 q
ı
“
¨
˚
˚
˚
˚
˚
˝
0
e
i 2π
q prk1 `rl2 `rk3 qpδk1l2 σk3 `δk1k3 eiπ
σl2
`δl2k3 σk1 ` |ǫk1l2k3 | ei π
2
p2´ǫk1l2k3 q
σ0
¯
e
i 2π
q prl1 `rk2 `rl3 q `
δl1k2 σl3 ` δl1l3 eiπσk2
`δk2l3 σl1 ` |ǫl1k2l3 | ei π
2
p2´ǫl1k2l3 q
σ0
¯ 0
˛
‹
‹
‹
‹
‹
‚
. (7.13)
2) There is one 0 among j1
1, j2
2, j3
1 , it is not important which one, and thus we get (e.g. for j3
1 “ 0)
µ
r3s
hetpqq
”
Σ
hetpqq
hetpqq
hetpqq
0l3 pr0, rl3 q
ı
“
¨
˚
˚
˚
˝
0 e
i 2π
q prk1 `rl2 `r0q
´
δk1l2 σ0 ` |ǫk1l2m| ei π
2
p2´ǫk1l2mq
σm
¯
e
i 2π
q prl1 `rk2 `rl3 qpδl1k2 σl3 `δl1l3 eiπ
σk2
`δk2l3 σl1 ` |ǫl1k2l3 | ei π
2
p2´ǫl1k2l3 q
σ0
¯ 0
˛
‹
‹
‹
‚
. (7.14)
In the same way, if there is one 0 among j1
2, j2
1, j3
2 , then (e.g. for j3
2 “ 0)
µ
r3s
hetpqq
”
Σ
hetpqq
hetpqq
hetpqq
k30
prk3 , s0q
ı
“
¨
˚
˚
˝
0
e
i 2π
q prk1 `rl2 `rk3 q `
δk1l2 σk3 ` δk1k3 eiπσl2
`δl2k3 σk1 ` |ǫk1l2k3 | ei π
2
p1´ǫk1l2k3 q
σ0
¯
e
i 2π
q prl1 `rk2 `s0q
´
δl1k2 σ0 ` |ǫl1k2m| ei π
2
p1´ǫl1k2mq
σm
¯
0
˛
‹
‹
‚. (7.15)
3) There are two 0’s in different entries of (7.4), e.g. j3
1 “ 0 and j3
2 “ 0, then
µ
r3s
hetpqq
”
Σ
hetpqq
hetpqq
hetpqq
00 pr0, s0q
ı
“
¨
˝
0 e
i 2π
q prk1 `rl2 `r0q
´
δk1l2 σ0 ` |ǫk1l2m| ei π
2
p2´ǫk1l2mq
σm
¯
e
i 2π
q prl1 `rk2 `s0q
´
δl1k2 σ0 ` |ǫl1k2m| ei π
2
p2´ǫl1k2mq
σm
¯
0
˛
‚, (7.16)
r0, s0 “ 0, 1, 2, . . . q ´ 1, m “ 1, 2, 3.
– 18 –

Each non-zero matrix entry in (7.13)–(7.16) is equal to one σ-matrix multiplied by a phase factor from
Cq, that is, it has the form (7.11), and therefore the ternary multiplication µ
r3s
hetpqq is closed and non-derived.
Since in the ternary case (only) the querelement (3.9) coincides with the inverse matrix (see the note after
(5.38)), and σ-matrices are reflections (2.9), we have
r
Σ
hetpqq
j1j2
prj1 , rj2 q “
˜
0 e´i 2π
q
rj2 σj2
e´i 2π
q
rj1 σj1 0
¸
j1,2 “ 0, 1, 2, 3, rjk
“ 0, 1, 2, . . . q ´ 1, q P Q12.
(7.17)
The ternary associativity follows from the associativity of ordinary matrix multiplication. Therefore,
Ĝ
hetr3spqq
Σ “
A!
Σ
hetpqq
j1j2
prj1 , rj2 q
)
| µ
r3s
hetpqq, Ă
p q
E
, (7.18)
is a finite ternary group of element-wise phase-shifted heterogeneous full Σhet
-matrices of order 4096q4
.
Acknowledgements. The author is deeply grateful to Mike Hewitt, Thomas Nordahl, Vladimir Tkach
and Raimund Vogl for useful discussions and valuable help.
REFERENCES
BAGARELLO, F., Y. BAVUMA, AND F. G. RUSSO (2021). Topological decompositions of the Pauli group and their influence
on dynamical systems. Math. Phys. Anal. Geom. 24 (2), 1–20. [Cited on page 13]
BALL, S., A. CENTELLES, AND F. HUBER (2023). Quantum error-correcting codes and their geometries. Ann. Inst. Henri
Poincare D, Comb. Phys. Interact. 10 (2), 337–405. [Cited on page 12]
BESCHE, H. U., B. EICK, AND E. O’BRIEN (2002). A millennium project: constructing small groups. Int. J. Algebra
Comput. 12 (05), 623–644. [Cited on page 13]
BESCHE, H. U., B. EICK, AND E. A. O’BRIEN (2001). The groups of order at most 2000. Electron. Res. Announc. Amer.
Math. Soc. 7 (1), 1–4. [Cited on page 14]
CONWAY, J. H., R. T. CURTIS, S. P. NORTON, R. A. PARKER, AND R. A. WILSON (1986). Atlas of Finite Groups.
Oxford: Oxford Univ. Press. [Cited on page 13]
DISTLER, A., C. JEFFERSON, T. KELSEY, AND L. KOTTHOFF (2012). The semigroups of order 10. In: M. MILANO (Ed.),
Principles and Practice of Constraint Programming, Vol. 7514 of Lecture Notes in Computer Science, Berlin-Heidelberg:
Springer, pp. 883–899. [Cited on page 14]
DJORDJEVIC, I. (2021). Quantum Information Processing, Quantum Computing, and Quantum Error Correction. An Engi-
neering Approach (2nd ed.). Amsterdam: Elsevier/Academic Press. [Cited on page 12]
DÖRNTE, W. (1929). Unterschungen über einen verallgemeinerten Gruppenbegriff. Math. Z. 29, 1–19. [Cited on page 4]
DUPLIJ, S. (2022). Polyadic Algebraic Structures. Bristol: IOP Publishing. [Cited on pages 2 and 3]
FEYNMAN, R. P., R. B. LEIGHTON, AND M. SANDS (1965). The Feynman Lectures on Physics, Vol. 3: Quantum Mechan-
ics. Reading, MA: Addison-Wesley. [Cited on page 2]
KIBLER, M. R. (2009). An angular momentum approach to quadratic Fourier transform, Hadamard matrices, Gauss sums,
mutually unbiased bases, the unitary group and the Pauli group. J. Phys. A 42 (35), 353001. [Cited on page 2]
LIBOFF, R. L. (2002). Introductory Quantum Mechanics. Addison-Wesley. [Cited on page 2]
NIELSEN, M. A. AND I. L. CHUANG (2000). Quantum Computation and Quantum Information. Cambridge: Cambridge
University Press. [Cited on pages 2 and 12]
SCHIFF, L. I. (1968). Quantum Mechanics. New York: McGraw-Hill. [Cited on page 2]
SLOANE, N. J. A. AND S. PLOUFFE (1995). The Encyclopedia of Integer Sequences. Academic Press. [Cited on page 14]
– 19 –

"Polyadic sigma matrices" by S. Duplij, arxiv: 2403.19361

More Related Content

Similar to "Polyadic sigma matrices" by S. Duplij, arxiv: 2403.19361 (20)

More from Steven Duplij (Stepan Douplii) (20)

Recently uploaded (20)

"Polyadic sigma matrices" by S. Duplij, arxiv: 2403.19361