arXiv:2108.10191v2 [math.CO] 24 Aug 2021 ubr,pa nagbaadanalysis. we and which pe algebra numbers and in crucial rational play the associated numbers, elucidate closely will which and papers numbers, three Catalan of first the is This Introduction 1 n hl hywr rtitoue yCtln[]i 84 hi true their 1874, in [4] Catalan by researchers. introduced eluded first were they while and colo ahmtc n Statistics and Mathematics of School Keywords: 00MteaisSbetClassification: Subject Mathematics 2020 nlssadacneto ocaatrsi eohroi a numbers. complem harmonic Catalan that zero super numbers circular (rational) characteristic additional to of variety connection wide a and of analysis phenomenon summatio a is polynomial which of connected, problem intricately the is geometry that show and geometries ntcrl vrafiiefil fodcharacteristic. odd algebra of an field of finite values a over of circle terms unit in interpretation algebraic an aldtecrua ue aaa ubr ( Ω numbers Catalan super circular the called hsivsiainoesu o nytepsiiiyt opr do to possibility the only not up opens investigation This ecnie he erclgoere,teEciengeome Euclidean the geometries, metrical three consider We egv h ue aaa numbers Catalan super the give We [email protected] ue aaa ubr,Crmgoer,and Chromogeometry, Numbers, Catalan Super ue aaa ubr,crmgoer,fiiefil,Fourie field, finite chromogeometry, numbers, Catalan super NWSydney UNSW ei Limanta Kevin o re umto vrFnt Fields Finite over Summation Fourier S ( ,n m, = ) m S (2 ! ( ue aaa numbers Catalan Super n ,n m, m ( ! !(2 )! S 10,1T5 30 Piay 51 (Secondary) 05E14 (Primary) 43-02 11T55, 11T06, m ( ) ,n m, + Abstract n ≡ )! n ,n m, )! m n hi soitdfml frtoa numbers, rational of family associated their and ) 1 (2 ! n ( Ω and n m endby defined ) ( ! colo ahmtc n Statistics and Mathematics of School !(2 )! m + [email protected] ,n m, n )! n )! n h ue aaa ubr and numbers Catalan super the ent omnJ Wildberger J. Norman citga fplnmasoe the over polynomials of integral ic = ) chromogeometry. ayi,btas h td fa of study the also but nalysis, r n w Einstein-Minkowski two and try S NWSydney UNSW vrteui icei each in circle unit the over n 4 ( r aua ubr fteform the of numbers natural are m hrceitcharmonic characteristic ime m ,n m, hp upiigrl htsuper that role surprising rhaps ilcl iclrsprCatalan super circular call will enn a ptl o largely now till up has meaning + summation. r n ) (1) These numbers generalize the well-known Catalan numbers
1 2n C(n) ≡ n +1 n since S (1, n)=2C(n). The fact that S(m, n) is always a whole number is a consequence of the recurrence relation 4S(m, n)= S(m +1, n)+ S(m, n + 1). (2) From the definition it is clear that S(n, m) = S(m, n). A further inspection of the formula shows that the zeroth row and zeroth column of the matrix of values S (m, n) agree with the diagonal, and these are the central binomial coefficients 2k S(k, 0) = S(0,k)= S(k,k)= . k The table below lists super Catalan numbers for small values of m and n from 0 up to 10, with S (m, n) in row m and column n, with indexing starting at 0.
m n 0 12345678910 \ 0 1 2 6 20 70 252 924 3432 12870 48620 184756 1 2 2 4 10 28 84 264 858 2860 9724 33592 2 6 4 6 12 28 72 198 572 1716 5304 16796 3 20 10 12 20 40 90 220 572 1560 4420 12920 4 70 28 28 40 70 140 308 728 1820 4760 12920 5 252 84 72 90 140 252 504 1092 2520 6120 15504 6 924 264 198 220 308 504 924 1848 3960 8976 21318 7 3432 858 572 572 728 1092 1848 3432 6864 14586 32604 8 12870 2860 1716 1560 1820 2520 3960 6864 12870 25740 54340 9 48620 9724 5304 4420 4760 6120 8976 14586 25740 48620 97240 10 184756 33592 16796 12920 12920 15504 21318 32604 54340 97240 184 756
Table 1: S(m, n) for 0 m, n 10. ≤ ≤ Somewhat surprisingly, there is no combinatorial understanding for S (m, n) to date, unlike the Catalan numbers which have hundreds of combinatorial interpretations, compiled for example in [15]; see also sequence A000108 in the OEIS [12]. There are, however, some combinatorial or weighted interpretations of S (m, n) for some values of m and n which we will describe in the next section. In this paper we introduce also a closely related family of rational numbers, the circular super Catalan numbers given by S(m, n) Ω(m, n) . (3) ≡ 4m+n
2 We also give a table of the circular super Catalan numbers for small values of m and n below.
m n 012345 6 7 8 \ 1 3 5 35 63 231 429 6435 0 1 2 8 16 128 256 1024 2048 32 768 1 1 1 5 7 21 33 429 715 1 2 8 16 128 256 1024 2048 32 768 65 536 3 1 3 3 7 9 99 143 429 2 8 16 128 256 1024 2048 32 768 65 536 262 144 5 5 3 5 5 45 55 143 195 3 16 128 256 1024 2048 32 768 65 536 262 144 524 288 35 7 7 5 35 35 77 91 455 4 128 256 1024 2048 32 768 65 536 262 144 524 288 4194 304 63 21 9 45 35 63 63 273 315 5 256 1024 2048 32 768 65 536 262 144 524 288 4194 304 8388 608 231 33 99 55 77 63 231 231 495 6 1024 2048 32 768 65 536 262 144 524 288 4194 304 8388 608 33 554 432 429 429 143 143 91 273 231 429 429 7 2048 32 768 65 536 262 144 524 288 4194 304 8388 608 33 554 432 67 108 864 6435 715 429 195 455 315 495 429 6435 8 32 768 65 536 262 144 524 288 4194 304 8388 608 33 554 432 67 108 864 2147 483 648
Table 2: Ω(m, n) for 0 m, n 8. ≤ ≤ This second table is still symmetric, namely Ω(n, m) =Ω(m, n). The relation between the zeroth row, or the zeroth column, and the diagonal is now 1 1 1 2k Ω(k,k)= Ω(k, 0) = Ω(0,k)= . 4k 4k 42k k A first indication of the possible interest of this second table is that the recurrence property of the super Catalan numbers becomes a Pascal-like property of the circular super Catalan numbers:
Ω(m, n) =Ω(m +1, n)+Ω(m, n + 1) so that each entry is the sum of the entry directly below it and to the right. It is not hard to see that these three conditions, together with the normalization condition that Ω(0, 0)= 1, determines the table.
Proposition 1 If a table of rational numbers c (m, n) for natural numbers m, n has the following properties:
1. c (0, 0)=1
2. c (m, n)= c (n, m) 1 1 3. c (m, m)= c (m, 0) = c (0, m) and 4m 4m 4. c (m, n)= c (m +1, n)+ c (m, n + 1)
then it must be that c (m, n)=Ω(m, n) for any natural numbers m and n.
3 We will see that the circular super Catalan numbers actually have a central algebraic and analytic interpretation in the context of integration or summation theory of polynomials over unit circles over finite fields of odd characteristic, in a manner that is to us quite remarkable, indeed almost unbelievable. Their role extends even to summations over suitably defined unit circles in relativistic geometries, and the curious relations between the Euclidean and relativistic formulas are an aspect of the three- fold symmetry of chromogeometry [16] which is an outgrowth of rational trigonometry [17]. This theory brings together Euclidean planar geometry with quadratic form x2 + y2 (blue) with two different forms of relativistic Einstein-Minkowski geometries with quadratic forms x2 y2 (red) and − xy (green). The relations and indeed symmetries between these three geometries play a crucial role in our derivations, as summations over unit circles in both Euclidean and relativistic geometries are closely related, and connections between them provide a powerful tool for explicit evaluations. In the rest of this Introduction we will explain the remarkable formulas in which the circular super Catalan numbers play the primary role and illustrate them with examples. We will also observe that the classical formulas for integration on the circle in the Euclidean plane can also be restated in terms of super Catalan numbers, so we obtain a bridge between characteristic zero and prime characteristic harmonic analysis. We will see that these formulas also suggest the study of a wide variety of additional (rational) numbers that play secondary roles.
1.1 Summing polynomials over unit circles over finite fields r We consider a finite field Fq where q = p is a power of an odd prime p, so that Fp Fq. The associated 2 ⊆ affine plane A consists of points [x, y] where x, y are in Fq, and we will in this paper identify this also with the associated vector space V2 of row vectors (x, y), so that linear transformations given by matrices act also on A2 on the right. Introduce three different metrical geometries: Euclidean (blue) and relativistic (red and green) with respective associated quadratic forms x2 + y2, x2 y2, and xy. The respective unit circles in − each geometry are then defined by
2 2 Sb (Fq)= Sb,q [x, y]: x, y Fq and x + y =1 ≡ ∈ 2 2 Sr (Fq)= Sr,q [x, y]: x, y Fq and x y =1 ≡ ∈ − Sg (Fq)= Sg,q [x, y]: x, y Fq and xy =1 ≡{ ∈ } where the subscripts b, r, and g naturally refer to blue, red, and green respectively; for brevity, we use the subscript c when we wish to refer to an undetermined one of these colors. We will also use Sc,q as a shorthand for Sc (Fq) when the underlying field is clear from context. The dichotomy of whether 1 is a square or not in Fq plays a role in the blue geometry and we − note that distinction through the Jacobi symbol, which is a generalization of the Legendre symbol. For any odd prime p, the Legendre symbol is defined as
1 1 if 1 is a square in Fp − − p ≡ 1 if 1 is not a square in Fp − − and for any prime power q = pr the Jacobi symbol is defined as 1 1 r − − . q ≡ p 4 This symbol comes into play in the following elementary count of the sizes of the unit circles.
Lemma 1 Let Sb,q, Sr,q, and Sg,q be the unit circles in the blue, red, and green geometries respectively over the field Fq. Then
1 Sg,q = q 1 Sr,q = q 1 and Sb,q = q − . | | − | | − | | − q Note that while we naturally consider the blue Euclidean geometry to be primary, we list the results in the order green, red, and blue which we will see generally aligns with the natural increasing complexity of formulas. Now consider the linear space Pol2 (Fq) of polynomials in two variables α and β with coefficients in Fq, which we will introduce in terms of variants called polynumbers which have been used by Wildberger [18] and allow us more clearly to distinguish between polynomials as algebraic objects and as functions. Polynumbers can be evaluated as functions, but their definition is solely in terms of an array of coefficients, and so in the context of polynumbers, the monomials αkβl as k and l range over all natural numbers (including 0) are independent. We will use the convention that α0β0 = 1 represents the constant polynumber whose value as a function is identically 1 = 1q in Fq. For c one of b, r, or g, define the Fq-valued normalized summation functional
ψc,q : Pol (Fq) Fq 2 → to be the linear functional whose values on a monomial αkβl for k and l natural numbers are given by
k l k l ψg,q α β x y ≡ − [x,y]∈S Xg,q k l k l ψr,q α β x y (4) ≡ − [x,y]∈S Xr,q k l 1 k l ψb,q α β − x y . ≡ − q [x,y]∈Sb,q X
We will show that ψc,q satisfies the following three conditions which we state loosely here to avoid technicalities, namely:
(Normalization) For the polynumber 1 in Pol2 (Fq) we have ψc,q (1)=1q.
(Locality) For any polynomial f that as a function restricts to zero on the circle Sc,q we have ψc,q (f) = 0.
(Invariance) ψc,q is invariant under any linear transformation that preserves the bilinear form of the associated geometry c.
These conditions will be presented in a more precise manner in Section 4 where we describe the linear groups of symmetries of each of the geometries. We call any linear functional satisfying the three conditions above a circular integral functional. The perhaps curious coefficients involved in the definitions are explained by the following main theorem.
5 Theorem 1 The linear functionals ψg,q, ψr,q, and ψb,q given in (4) are the unique circular integral functionals in the green, red, and blue geometries respectively.
We prove the above theorem in two steps. First, we establish that ψc,q is indeed a circular integral functional in each geometry c. Next, we prove that if such a circular integral functional in each geometry exists, then it must be unique. In fact, not only are we able to prove the uniqueness of ψc,q, we are also able to give an explicit formula for ψc,q in each geometry, and curiously the derivation in the red and blue geometries stems from the result in the green geometry. Moreover, we can describe ψr,q and ψb,q in terms of the elements
S (m, n) Ω(m, n) mod p = mod p 4m+n (2m)!(2n)! = mod p 4m+nm!n!(m + n)!
m+n of Fp Fq which are well-defined since S (m, n) is always an integer and the term 4 in the denominator⊆ is never a multiple of p. Here is a simplified and restricted version of the formulas for ψc,q in the red and blue geometries. This is the heart of our understanding of the true significance of the super Catalan numbers, and their rational cousins the circular super Catalan numbers.
r Theorem 2 Consider the finite field Fq where q = p with p> 2.
1. In the red geometry, for 0 k + l < q 1, ≤ − n k l ( 1) Ω(m, n) mod p if k =2m and l =2n, ψr,q α β = − 0 otherwise. 2. In the blue geometry, for 0 k + l < q 1, ≤ −
k l Ω(m, n) mod p if k =2m and l =2n, ψb,q α β = 0 otherwise. Here are some immediate consequences.
r Corollary 2.1 Consider the finite field Fq where q = p with p> 2 and any natural numbers k and l satisfying k + l < q 1. − k l k l 1. The normalized sums ψr,q α β and ψb,q α β lie in the prime field Fp and are independent of the power of p. 2. Suppose l =2n. We then have
k l n k l ψr,q α β =( 1) ψb,q α β . −