§ 1. Like many mathematical articles, this one begins with a set, specifically S5 = { 0, 1, 2, 3, 4 }. Variables representing elements of this set are written a, b, c et cetera. Appearing also in the discussion are integers i, j, k, m and n, which are not necessarily elements of S5. Any arithmetic involving the members of S5 is done in modulo 5.
The principal undertaking is to introduce 240 ("dozen score" = 12 × 20) binary operations over this set, the operations satisfying two cancellation properties that can be explained after some notation is established. Each operation is denoted by a nonnegative integer followed by the nabla symbol: 0∇, 1∇, 2∇ … 239∇. Variables representing operations are written Z, Y, X et cetera. For an example of such an operation, here is the table that defines 58∇:
58∇ | second input | |||||
---|---|---|---|---|---|---|
0 | 1 | 2 | 3 | 4 | ||
first input | 0 | 1 | 0 | 4 | 3 | 2 |
1 | 3 | 2 | 1 | 0 | 4 | |
2 | 0 | 4 | 3 | 2 | 1 | |
3 | 2 | 1 | 0 | 4 | 3 | |
4 | 4 | 3 | 2 | 1 | 0 |
First-order cancellativity (FOC) is the well-known property where:
Second-order cancellativity (SOC) is achieved when:
There are 240 binary operations on S5 that satisfy both FOC and SOC, and they are the target of this report. They are all listed in a large table, and formatted as a C-language array.
This report generally uses prefix notation for applications of the operations, but infix notation is equally valid and sometimes clearer:
Z (a, b) ⇔ a Z b
Parentheses and the comma are retained in the prefix notation to reduce confusion, because the arguments are sometimes expressions involving more than one symbol.
§ 2. The nabla numbers of the dozen-scores are based on an obvious manner of lexicographic ordering. The table below is explicit about how the properties of operations m∇ and n∇ are used to establish a relation between integers m and n.
dozen-score lexicographic ordering | |
---|---|
first criterion | • If m∇(0, 0) < n∇(0, 0), then m < n.
• If m∇(0, 0) > n∇(0, 0), then m > n. • If m∇(0, 0) = n∇(0, 0), then use the second criterion. |
second criterion | • If m∇(0, 1) < n∇(0, 1), then m < n.
• If m∇(0, 1) > n∇(0, 1), then m > n. • If m∇(0, 1) = n∇(0, 1), then use the third criterion. |
third criterion | • If m∇(0, 2) < n∇(0, 2), then m < n.
• If m∇(0, 2) > n∇(0, 2), then m > n. • If m∇(0, 2) = n∇(0, 2), then use the fourth criterion. |
fourth criterion | • If m∇(0, 3) < n∇(0, 3), then m < n.
• If m∇(0, 3) > n∇(0, 3), then m > n. • If m∇(0, 3) = n∇(0, 3), then use the fifth criterion. |
fifth criterion | • If m∇(0, 4) < n∇(0, 4), then m < n.
• If m∇(0, 4) > n∇(0, 4), then m > n. • If m∇(0, 4) = n∇(0, 4), then use the sixth criterion. |
sixth criterion | • If m∇(0, 5) < n∇(0, 5), then m < n.
• If m∇(0, 5) > n∇(0, 5), then m > n. • If m∇(0, 5) = n∇(0, 5), then m = n. |
§ 3. The dozen-scores sometimes do, and sometimes do not, exhibit familiar algebraic properties.
Because of SOC, the operations are strictly non-commutative: if Z (a, b) equals Z (b, a), then a must equal b.
The standard associative property always fails, but some multiple-operation versions succeed. Also a distributive law works for some combinations of operations. These are discussed in section 6 below.
No operations have a two-sided identity element, although some have a left identity, which means that there exists an a such that Z (a, b) = b for all b. Some others have the corresponding right identity. FOC guarantees that when such a one-sided identity exists, it is unique. Examples:
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
left identity = 2 | right identity = 1 |
Only two operations are idempotent in that Z (a, a) = a:
|
|
§ 4. One operation can be systematically altered into other. If X (a, b) = W (a, b) for all a and b, the identity shorthand X ≡ W is often used.
§ 4a. Three self-inverse transformations:
|
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
100∇ | 110∇ ≡ 100∇T | 167∇ ≡ 100∇V | 213∇ ≡ 100∇H |
The elements unmoved in transposition lie on the principal diagonal.
Successive superscripts are evaluated in the natural fashion: ZHTV ≡ ((ZH)T)V.
Two notable relations are Z (b, a) = ZT (a, b) and ZTH ≡ ZVT.
The vertical and horizontal reversing operations suggest establishing a unary reversing operation, superscripted with I (for inverse) to avoid confusion with an operation introduced in the next section:
a | 0 | 1 | 2 | 3 | 4 |
aI = 4 − a | 4 | 3 | 2 | 1 | 0 |
Thus Z (a, b) = ZV (aI, b) = ZH (a, bI) = ZHV (aI, bI) = ZVH (aI, bI).
§ 4b. Two cyclical transformations:
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
62∇ | 107∇ ≡ 62∇R+3 | 161∇ ≡ 62∇C+4 |
These four statements are equivalent:
Row and column rotations are not independent:
There is not an transformation to in general exchange two rows, or two colums, as this could disturb SOC.
§ 4c. Elements of the table can be pairwise swapped, as denoted by a superscript S and two numbers from { 0, 1, 2, 3, 4 }. Only the output values, not the input values, are changed. The highlighted squares of this example show the effect of swapping:
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1∇ ≡ 43∇S1,4 | 43∇ ≡ 1∇S1,4 |
The swap being pairwise, ZSm,n ≡ ZSn,m. If m = n, nothing happens.
Canonization (or canonicalization) is the process of performing a sequence of pairwise swaps ultimately producing an operation where Z(0, a) = a; this condition is equivalent to saying that 0 is a left identity of Z. Such a canonical form helps to reveal the "shape" of the operation. Anywhere between zero and four swaps will be necessary. Example:
|
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
159∇ | 7∇ ≡ 159∇S0,3 | 3∇ ≡ 7∇S3,2 | 1∇ ≡ 3∇S4,3 |
For the dozen-scores:
If Z reduces to 0∇ and Y reduces to 1∇, then Z and Y taken together form a Graeco-Latin square. Each cell of the table contains a different ordered pair. Example:
60∇, 104∇ | second input | |||||
---|---|---|---|---|---|---|
0 | 1 | 2 | 3 | 4 | ||
first input | 0 | 1, 2 | 2, 0 | 0, 4 | 3, 1 | 4, 3 |
1 | 0, 1 | 3, 3 | 4, 2 | 1, 0 | 2, 4 | |
2 | 4, 0 | 1, 4 | 2, 1 | 0, 3 | 3, 2 | |
3 | 2, 3 | 0, 2 | 3, 0 | 4, 4 | 1, 1 | |
4 | 3, 4 | 4, 1 | 1, 3 | 2, 2 | 0, 0 |
§ 5. The dozen-scores can be partitioned into six subsets (quadragintas) of 40 operations each, as shown in the tables below.
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
The following formulas show more specifically how the dozen-scores within a column or row of a quadraginta table are connected, by giving as examples some relations within the first column and the first row of quadraginta B:
2∇ ≡ | 16∇T ≡ | 229∇V ≡ | 190∇H |
108∇ ≡ | 104∇T ≡ | 190∇V ≡ | 229∇H |
215∇ ≡ | 229∇T ≡ | 104∇V ≡ | 16∇H |
165∇ ≡ | 190∇T ≡ | 16∇V ≡ | 104∇H |
229∇ ≡ | 215∇T ≡ | 2∇V ≡ | 108∇H |
190∇ ≡ | 165∇T ≡ | 108∇V ≡ | 2∇H |
16∇ ≡ | 2∇T ≡ | 165∇V ≡ | 215∇H |
104∇ ≡ | 108∇T ≡ | 215∇V ≡ | 165∇H |
The operations in quadraginta A exhibit linearity in the following sense: for each operation there exist i, j and k such that Z(a, b) = i × a + j × b + k. With 128∇ for instance, i = 3, j = 1, and k = 2:
128∇ | second input | |||||
---|---|---|---|---|---|---|
0 | 1 | 2 | 3 | 4 | ||
first input | 0 | 2 | 3 | 4 | 0 | 1 |
1 | 0 | 1 | 2 | 3 | 4 | |
2 | 3 | 4 | 0 | 1 | 2 | |
3 | 1 | 2 | 3 | 4 | 0 | |
4 | 4 | 0 | 1 | 2 | 3 |
For a linear operation Z, the coëfficients are easily found:
With linear operations, four relations can never occur, because they would be connected with a violation of FOC or SOC:
In the next section are more examples of how quadraginta A is distinguished from the others.
§ 6. Combining operations.
§ 6a. There are 48,000 combinations (out of 2404 possibilities) of operations that satisfy this four-operation generalization of the associative law, here written in both prefix and infix notations:
Z (a, Y (b, c)) = X (W (a, b), c)
a Z (b Y c) = (a W b) X c
Necessary conditions are that operations Z and X must come from the same quadraginta, and operations Y and W must come from quadraginta A. Never does the particular case Z ≡ Y ≡ X ≡ W succeed.
§ 6b. There are 576,000 combinations (of 2404) of operations that satisfy this next version of associativity:
Z (Y (a, b), c) = X (W (a, b), c)
(a Y b) Z c = (a W b) X c
A necessary condition is that Z and X must come from the same quadraginta. Always successful is Z ≡ Y ≡ X ≡ W.
§ 6c. There are 160 combinations (of 2402) of operations, all from quadraginta A, that satisfy this distributive law:
Z (Y (a, b), c) = Y (Z (a, c), Z (b, c))
(a Y b) Z c = (a Z c) Y (b Z c)
Only 21∇ and 46∇, the idempotent operations, fulfill the constraint Y ≡ Z.
§ 6d. A general approach to composite operations is lengthy enough to need its own page.
§ 7. Thus far, great attention has been paid to a collection of 240 operations over the set S5 = { 0, 1, 2, 3, 4 }. Why not the more general Sn = { 0, 1, 2 … n − 1 }? There is an important result about binary operations over Sn that are first- and second-order cancellative:
For each of S7 and S11, there are two substantively different configurations. Examples, in canonical form, appear below preceded by 1∇ for comparison:
| |||||||||||||||||||||||||||||||||||||||||||
1∇(a, b) = 1∇(a + 1, b + 2) |
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
U1(a, b) = U1(a + 1, b + 2) | U2(a, b) = U2(a + 1, b + 3) |
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
U3(a, b) = U3(a + 1, b + 2) | U4(a, b) = U4(a + 1, b + 3) |
All the other FOC and SOC operations over S7 and S11 can be found by making the obvious adaptations to the following transformations that were defined for S5:
When n ≤ 11, each operation over Sn can, through canonization, be transformed into a linear function (section 5). However, this property no longer applies when n ≥ 13, the point at which the operations become far more complicated. The following is an operation over S13 that, although it is in canonical form, is conspicuously non-linear:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
U5(a, b) = U5(a + m, b + n) cannot be satisfied |
§ 8. The three-operand case can also be glimpsed. The following definitions are practically inevitable, and assume modulo-n arithmetic when the operation is over Sn:
FOC is satisfied when:
SOC is satisfied when:
Third-order cancellativity (TOC) is satisfied when:
Within the formulas there are more plus signs than minus signs, but this is not a problem because n itself can be positive or negative,.
SOC and TOC form two different kinds of pandiagonality. (Analogous are the face-centered and body-centered cubic lattices.)
Some existences: