Abstract
This paper is devoted to the newly defined families of associated sequences of real polynomials and numbers that arose on a base of quaternions. We present not only the explicit and recurrent formulae for these sequences, but also summation and reduction ones. Some (but not all) of these sequences can be found in OEIS. Moreover, the obtained results also speak a lot about the quaternions structure itself. Also, while examining matrices related to these sequences we discovered some general formulae for powers of so-called arrowhead matrices.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Wituła and Słota in their papers [21,22,23] introduced original number systems called quasi-Fibonacci numbers of n-th order. These systems arose naturally on a base of two well-known equivalent relations for Fibonacci numbers, that is
and
where \(F_n\) is the n-th Fibonacci number and \(n\in {\mathbb {N}}_0,\ k\in {\mathbb {N}}{\setminus } 5{\mathbb {N}},\ \eta \in {\mathbb {C}}{\setminus }\{1\},\) \(\eta ^5=1\), i.e. k is not divisible by 5 and \(\eta \) is a primitive 5-th root of unity (see also [14, 19]).
The simplest system of this type consists of \(\delta \)-Fibonacci numbers discussed in many papers (e.g. [10, 11, 23]). Recall that \(\delta \)-Fibonacci numbers are members of sequences \((a_n(\delta )),\ (b_n(\delta ))\) defined in the following way
where \(n\in {\mathbb {N}}_0,\ k\in {\mathbb {N}}{\setminus } 5{\mathbb {N}},\ \eta \in {\mathbb {C}}{\setminus }\{1\},\ \eta ^5=1\).
They can be also defined recurrently, namely \(a_0(\delta )=1,\ b_0(\delta )=0\) and
One of the most important properties of \(\delta \)-Fibonacci numbers is the fact that they represent the binomials transformation of the scaled Fibonacci numbers, that is the following identities hold
for \( n\in {\mathbb {N}}\). Also, interesting sequences were obtained for some values of \(\delta \):
Moreover, for odd natural n we have
and for even natural n we have
where \(L_n\) denotes the n-th Lucas numberFootnote 1 whereas \(T_n\) and \(U_n\) denote the Chebyshev polynomials of the first and second kind, respectivelyFootnote 2 (see [8]).
It turned out that it is even more natural and proper to consider generalizations of Fibonacci numbers depending on more parameters [17]. These are quasi-Fibonacci numbers of order n (which is assumed to be odd) and depend on \(\frac{1}{2}\varphi (n)-1\) parameters, where \(\varphi =\varphi (n)\) is the Euler’s totient function. In [17] the first nontrivial cases were discussed in details, namely quasi-Fibonacci numbers of order 7 and 9 (these are the only two-parametric numbers). For instance, two-parametric quasi-Fibonacci numbers of 9-th order are members of sequences \((A_{n,9}(\delta ,\gamma )),\ (B_{n,9}(\delta ,\gamma ))\) and \((C_{n,9}(\delta ,\gamma ))\) defined by the following relations
where \( n \in {\mathbb {N}}_0,\ k \in {\mathbb {N}}{\setminus } 9{\mathbb {N}}, \ \delta , \gamma \in {\mathbb {C}},\ \zeta \in {\mathbb {C}}\) is a primitive 9-th root of unity.
They can be also defined recurrently, namely we have \(A_{0,9}(\delta ,\gamma )=1,\) \(B_{0,9}(\delta ,\gamma )=C_{0,9}(\delta ,\gamma )=0\) and for all \(n\in {\mathbb {N}}_0\)
Note that for \(\gamma =0\) we get the (one-parameter) quasi-Fibonacci numbers (this type of numbers was discussed e.g. in [21, 22]).
From one side n-th roots of unity form a cyclic group under multiplication, and from the other some sums of these roots are linearly independent (like \(1,\ \zeta ^k +\zeta ^{8k}\) and \(\zeta ^{2k} +\zeta ^{7k}\) used in the above definition). That made us extend the idea of quasi-Fibonacci numbers to real quaternions. During our investigations we obtained a lot of interesting results, also on quaternions structure itself, and they will be presented in this paper.
We first remind the definition and basic properties of real quaternions. They arose as a natural extension of complex numbers [15].
Definition 1
Let \({\mathbb {H}}\) be a 4-dimensional vector space over \({\mathbb {R}}\) with the basis \(\{1,i,j,k\}\). Then a (real) quaternion q is an element of \({\mathbb {H}}\) written with respect to this basis, that is
where \(a_0\;,a_1,\;a_2,\;a_3 \in {\mathbb {R}}\) with i, j, k called imaginary units.
Now, let the multiplication be defined on imaginary units by the following rules, called Hamilton formulae,
and extended to \({\mathbb {H}}\) by requirements that it is associative, distributive over addition and commutative with respect to scalar multiplication. The resulting division ring is also denoted by \({\mathbb {H}}\).
Moreover, a unary operation \({\overline{\omega }}:{\mathbb {H}}\rightarrow {\mathbb {H}}\) called conjugation is defined in the following way
and the norm of a quaternion q is defined as
From now on i, j, k will be used only as imaginary units.
Note that quaternions can be also represented by matrices in such a manner that multiplication of quaternions is compatible with that of matrices.Footnote 3 Namely, let I be the identity matrix of order 4 and let H, J, K be matrices of order 4 representing i, j, k, respectively, i.e. satisfying the relations like in (2). Then \(q\in {\mathbb {H}}\) can be written in a matrix form Q as
It was proven in [4] that there are 48 distinct representations of quaternions by skew-symmetric signed permutation matrices. More precisely, H, J, K can be chosen from two sets \(\{\pm X_s,\pm Y_s,\pm Z_s\},\ s=1,2\) under assumption that we take either a matrix or its negation and formulae (2) are satisfied where
Now, let \(S^3\) denote the set of unit quaternions, that is
Note that \(S^3\) is a subgroup of the multiplicative group of \({\mathbb {H}}\). Moreover, in \({\mathbb {H}}\simeq {\mathbb {R}}^4\) it is a surface of a unit sphere with the centre at the origin (see [13]).
Next, each quaternion \(q\in S^3{\setminus }\{\pm 1\}\) can be written in a polar form
where \(\cos \theta = a_0\) and \(\omega =\frac{a_1 i+a_2j +a_3 k}{\sqrt{a_1^2+a_2^2+a_3^2}}=\frac{a_1 i+a_2j +a_3 k}{\sqrt{1-a_0^2}}\), from which Cho in [2] obtained an analogue of de Moivre’s formula, i.e.
Actually, each nonreal quaternion \(q\in {\mathbb {H}}\), i.e. such that \(a_1^2+a_2^2+a_3^2\ne 0\), can be written in a polar form
where
such that the generalization of (4), still called de Moirve’s formula, holds
2 Quaternaccis
Definition 2
Quaternion equivalents for quasi-Fibonacci numbers (shortly Quaternaccis) are members of sequences \((A_{n}(b,c,d)), (B_{n}(b,c,d)),(C_{n}(b,c,d)),\) \((D_{n}(b,c,d))\) defined by the following relation
where \(n \in {\mathbb {N}}_0\), \(b,c,d\in {\mathbb {R}}\).
Example
Values of Quaternaccis for \(n\leqslant 6\) are given at the end of this paper.
Notation From now on, we will use (b, c, d) only to denote arguments of Quaternaccis. Moreover, while describing properties valid for any argument (b, c, d) we shall omit them in formulation of results, in particular we will use the following abbreviations:
Theorem 1
For all \(n\in {\mathbb {N}}_0\) the following recurrence relation holds
Proof
It is obviously true for \(n=0\). For \(N\in {\mathbb {N}}\) from Definition 2 we get
whereas from the properties of exponentiation of quaternions it follows that
Since \(\{1,i,j,k\}\) is the basis of \({\mathbb {H}}\), we obtain the statement by comparison of coefficients of 1, i, j, k in (8) and (9).\(\square \)
Proposition 1
For all \(m,n\in {\mathbb {N}}_0\) the following relations hold
and
Proof
From associativity of powers for \(N\in {\mathbb {N}}\) we also have
Using the definition of Quaternaccis for the later factor and comparing with (9) we get (10). The remaining three formulae are derived in an analogous way from \(q^nq^m=q^{m+n}=q^mq^n\) for \(q=1+bi+cj+dk\).\(\square \)
Formulae (7) and (10) immediately imply the following relations:
Corollary 1
For all \(n\in {\mathbb {N}}_0\) the following recurrence relations are satisfied
and
Remark 1
Note that formulae (10) actually allow to write a form of the recurrence relations for Quaternaccis which generalize (7), (12) and (13), by introducing parameters \(a_1,a_2,a_3\in {\mathbb {R}}\) as follows
Remark 2
The matrix given in Theorem 1 is one of the basic matrix representations of quaternion \(1+bi+cj+dk\), obtained from (3) for \(H=-X_2,\) \(J=-Y_2,K=-Z_2\), similarly as the matrix in (12), which is obtained for \(H=-Y_1,J=-X_1,K=-Z_1\).
Our next results provide more properties of transition matrices of recurrence relations (7), (12) and (13). We first focus on determining the n-th power of these matrices. For instance, after n successive applications of formula (7) we obtain
Note that since \(A_{0}=1,\ B_{0}=C_{0}=D_{0}=0\), then the first column of the above matrix is \([A_{n},B_{n},C_{n},D_{n}]^T\). Analogous properties are true for (12), (13) and (14). But it turns out that the situation is even more interesting.
Theorem 2
For all \(n\in {\mathbb {N}}_0\) the following identities hold
and
where \(E_{n}:={\left\{ \begin{array}{ll}0&{} b=c=d=0\\ \frac{A_{n}-1}{b^2+c^2+d^2} &{} \text {otherwise}\end{array}\right. }\).
Proof
We shall prove (15) by induction on n. By definition of Quaternaccis the statement follows for \(n=0\) and moreover, for \(n=1\) we have
Next, assume that (15) is true for some \(n\in {\mathbb {N}}\). Then
Therefore by the induction principle formula (15) is valid for all \(n\in {\mathbb {N}}_0\).
Formula (16) is proven analogously. To prove (17) we need to justify auxiliary identities, namely that \(E_0=E_1=0\) and for \(n\geqslant 2,\ bcd\ne 0\)
whereas for \(b=0\) or \(c=0\) or \(d=0\) the corresponding sum is 0.
These identities can be proven by induction, however a more elegant and interesting proof will be given later (see Corollary 2). \(\square \)
Note that if in the proof we use the inductive assumption as follows
then we obtain a matrix that is equal to the desired one by means of (10).
Remark 3
In Appendix 1 we give two general methods for exponentiation of so-called arrowhead matrices which could be also used for proving (17).
Remark 4
Note that matrices in (15) and (16) are actually of the very special form (mind the colors)—we point it out below for (15).
More precisely, for every \(n\in {\mathbb {N}}\) the matrix on the right results form the one on the left by substituting 1, b, c, d by \(A_{n}, B_{n},C_{n}\) and \(D_{n}\), respectively. Or, on the other hand, the matrix on the right is obtained from the middle one by replacing the index by the exponent.
An analogous property is true for two-parameter quasi-Fibonacci numbers of order 7 (see [17]) and 9 (the formula presented below is new)
Notation From now on for \(b,c,d \in {\mathbb {R}}\) we set \(\lambda :=\sqrt{b^2+c^2+d^2}\).
Straightforward calculations lead to the following
Proposition 2
Transition matrices in (7), (12) and (13) are diagonalizable and their Jordan decompositions are as follows:
a)
where
b)
where
c)
where
As a consequence of Theorem 2 we get the following
Theorem 3
Quaternacci sequences \((A_{n}), (B_{n}),(C_{n})\) and \((D_{n})\) satisfy the following recurrence relation
with initial conditions
Proof
Let \({\mathcal {W}}^n=[w_{ij}(n)]_{4\times 4}\) for all \(n \in {\mathbb {N}}\). If the matrix \({\mathcal {W}}\) is diagonalizable, then its minimal polynomial \(\psi \) is a product of linear factors for distinct eigenvalues. Since \(\psi ({\mathcal {W}})=0\) and \(\psi ({\mathcal {W}})\) is a linear combination of some powers of \({\mathcal {W}}\), for each pair of indices i, j we obtain a linear equation involving \(w_{ij}(n)\) for different n. Hence each sequence \((w_{ij}(n)), \; n\in {\mathbb {N}}\) can be defined by a linear recurrence relation with characteristic polynomial equal to the minimal polynomial for \({\mathcal {W}}\).
So if we take a transition matrix in formula (7) as \({\mathcal {W}}\), then from (15) in particular we have \(w_{11}(n)=A_{n}, w_{21}(n)=B_{n}, w_{31}(n)=C_{n}, w_{41}(n)=D_{n}\). Therefore (23) follows from b) in Proposition 2. \(\square \)
Remark 5
There is a natural connection between Quaternaccis and Lucas numbers of the first and second kind, i.e. \(U_n(P,Q)\) and \(V_n(P,Q)\), respectively, which are given by a recurrence relation
with \(P,Q\in {\mathbb {Z}},\ U_0(P,Q)=0,U_1(P,Q)=1\) and \(V_0(P,Q)=2, V_1(P,Q)=P\).
For instance, for all \(n\in {\mathbb {N}}_0\) we have \(B_{n}(1,c,d)=U_n(2,2+c^2+d^2)\) and \(2A_{n}(b,c,d)=V_n(2,1+b^2+c^2+d^2)\).
Corollary 2
Formulae (18) are valid, that is for \(n\geqslant 2,\ bcd\ne 0\) we have
and for \(b=0\) or \(c=0\) or \(d=0\) the corresponding sum equals 0.
Proof
Let \((x,X)\in \{(b,B),(c,C),(d,D)\}\). If we fix \(n\in {\mathbb {N}}\), then from (23) we get
Adding the above (a telescoping summing) and substituting \(X_0=0,\) \(X_1=x\) we obtain
Now, if we take \(X_{n}=x{\tilde{X}}_{n}\), where
and \(Y_n={\tilde{X}}_{n}-{\tilde{X}}_{n-1}\) for \(n\in {\mathbb {N}}\), then from above we get
Hence by Theorem 3, \(Y_n=A_{n-1}\) and we have
which implies that the sum is 0 if \(x=0\) and for \(x\ne 0\) finally gives
\(\square \)
A consequence of Theorem 3 are Binet’s formulae for Quaternaccis
Corollary 3
If \(n\in {\mathbb {N}}_0,\ bcd\ne 0\) then Binet’s formulae for Quaternaccis are the following
which is equivalent to
If \(x=0\), where \(x=b,c,d\), then \(X_{n}\equiv 0\) for \(X=B,C,D\), respectively.
Note that the proof of Corollary 2 can be also derived from Binet’s formulae. In particular, from (25) and (26) we can obtain
where \(\varphi =\frac{1+\sqrt{5}}{2}\) and \({\overline{\varphi }}=\frac{1-\sqrt{5}}{2}\).
Another immediate consequence of Binet’s formulae is the following
Corollary 4
For each \(n\in {\mathbb {N}}_0\) the following cyclic relations for b, c, d holds
and \((A_{n}(b,c,d))\) is invariant with respect to permutations of b, c, d.
Assume now that \(\alpha ,\beta ,\gamma ,\delta \in {\mathbb {R}},\ \alpha \ne 0\) are such that \(q=\alpha +\beta i+\gamma j+\delta k \in S^3\). Then de Moivre’s fomula (4) and the following application of Definition 2
yields
Corollary 5
Let \(\alpha =\cos \theta \) and \(\beta ,\gamma ,\delta \in {\mathbb {R}}\) be such that \(\alpha \beta \gamma \delta \ne 0\) and \(\alpha ^2+\beta ^2+\gamma ^2+\delta ^2=1\). Then for every \(n\in {\mathbb {N}}\), Quaternaccis and Chebyshev polynomials are related by the following identities
where \({\tilde{U}}_n(\alpha )\) is a scaled Chebyshev polynomial of the second kind, that is
Finally, Horadam in [12] introduced the number systems called Fibonacci quaternions \(Q_n\) and Lucas quaternions \(K_n\), defined recurrently by
where \(n\in {\mathbb {N}}_0\) and \(F_n,L_n\) denote the n-th Fibonacci and Lucas number, respectively (see also [5, 6, 18]).
He also proved equivalents of identities for Fibonacci numbers, like e.g.
Note that Quaternaccis and Fibonacci quaternions are related by the following identity, valid for all \(m,n \in {\mathbb {N}}_0\):
Moreover, from (40) and (41) we obtain
Corollary 6
For all \(m,n\in {\mathbb {N}}_0\) the following identities holds
3 Reduction Formulae
By the properties of exponentiation and from Proposition 1 we obtain the basic reduction formulae for Quaternaccis, that is
Theorem 4
Let \(X\in \{B,C,D\},\ m,n\in {\mathbb {N}}_0\). Then
And the special cases are the following
Corollary 7
Let \(X\in \{B,C,D\},\ n\in {\mathbb {N}}_0\). Then
a)
where \(\Delta =1-\lambda ^2\),
b)
where
Quaternaccis can be used for obtaining decompositions of polynomials.
Theorem 5
a) For all \(n\in {\mathbb {N}}_0\) the following identity holds
where \(X_b=X_n(b,0,0)\) for \(X\in \{A,B,C,D\}\).
b) Let \(\alpha _n=(1+bi+cj)^n,\ \beta _n=(1+bj+ck)^n,\ \gamma _n=(1+bi+ck)^n,\ n\in {\mathbb {N}}_0 \). Then Vieta’s formulae for \(({\mathbb {X}}-\alpha _n)({\mathbb {X}}-\beta _n)({\mathbb {X}}-\gamma _n)\) are the following
andFootnote 4
where \(X_{bc}=X_n(b,c,0)\) for \(X\in \{A,B,C,D\}\).
Proof
We will prove the case a), b) is analogous. After multiplying and rearranging the left side we have
Then the statement follows from Definition 2 and Corollary 4. \(\square \)
From Binet’s formulae we obtain further reduction formulae.
Theorem 6
For \(X \in \{B,C,D\}\) and \(n\in {\mathbb {N}}_0\) the following identities hold
a) for any real a:
b) for \(bcd \ne 0\) such that \(b\ne c\):
c) for \(bcd\ne 0\) such that \(b^2+c^2=d^2\):
Some index-parameter reduction formulae were also obtained.
Theorem 7
For \(X\in \{A,B,C,D\}\) and \(m,n \in {\mathbb {N}}\) we have
Proof
Definition of Quaternaccis and properties of exponentiation yield
Comparison of coefficients of 1, i, j, k finishes the proof. \(\square \)
Example
From (25), that is Binet’s formula for \(A_{n}\), we get for instance
and (26) gives \(B_{4}(1,2,0)=-16, C_{4}(1,2,0)=-32, D_{4}(1,2,0)=0.\) Thus from Theorem 7 we obtain
Similarly we have
An interesting trigonometric case follows from Theorem 7 and Corollary 5:
Corollary 8
Let \((x,X)\in \{(\beta , B),(\gamma , C),(\delta ,D)\}\). If \(\alpha ^2+\beta ^2+\gamma ^2+\delta ^2=1\) and \( \beta \gamma \delta \ne 0\) then for every \(m,n\in {\mathbb {N}}\) we have
Remark 6
Formulae in Theorem 7 significantly simplify the evaluation of Quaternaccis for high indices, e.g. for \(X\in \{A,B,C,D\},\ n\in {\mathbb {N}}\) we have
That means that if we want to evaluate Quaternaccis at some point (b, c, d) for index 1000, say, it is enough to determine their values for index 10, although the calculations has to be done at some other points (specified above).
4 Applications
1. Professor Webb in 2008, during the 13th International Conference on Fibonacci Numbers and Their Applications in Patras, posed a question of finding compact formulae for sums of the form
where \(r\in {\mathbb {N}}\). Thanks to properties of \(\delta \)-Fibonacci numbers (recalled in Sect. 1) this problem was reduced to finding sums of powers of Fibonacci numbers with indices not greater than N. More precisely, by reduction formulae for \(\delta \)-Fibonacci numbers (similar to those in Theorem 7) we can get (see [23])
whence for instance
Similarly, reduction formulae given in Theorem 7 can be used for solving analogous problems in which Quaternaccis can be involved. For instance, high powers of Fibonacci quaternions \(Q_n\) can be replaced by high powers of Quaternaccis (due to formula (41)) which in turn can be lowered significantly by Theorem 7 and Corollary 6, which can be lowered even further by (43) and (44).
For instance, for \(s,r,n,N,M\in {\mathbb {N}},\ M\leqslant N\) we have
where \(X_{n,F}=X_{n}\left( \frac{F_{{n^s}+1}}{F_{n^s}},\frac{F_{{n^s}+2}}{F_{n^s}},\frac{F_{{n^s}+3}}{F_{n^s}}\right) \) for \(X\in \{A,B,C,D\}\) and indices of Fibonacci numbers \(F_{n^s}\) can be reduced in a way mentioned above.
Remark 7
Since every nonconstant quaternion can be written in a polar form (see (5)), then sums of the form \({\sum _{k=M}^NQ_t^{2m}}\) can be evaluated by determining some trigonometric sums of the form \({\sum _{l=1}^t \alpha ^l \cos l\phi }\) and \({\sum _{l=1}^t \alpha ^l \sin l\phi }\), where \(\alpha ,\phi \in {\mathbb {R}}\). Formulae for these sums are known (see e.g. [20, p.13]). Note however, that our approach, like in (45), allows to reduce arguments of cosine, sine and powers of \(\alpha \), unlike the other known formulae.
2. Quaternaccis can be used also to determine some trigonometric sums. What is interesting in this approach is that we do not use known trigonometric identities but we use only those we obtained from properties of exponentiation of quaternions, i.e. (33)–(36).
Theorem 8
Set \(\alpha =\cos \theta \), where \(\theta \in {\mathbb {R}}{\setminus } \{\frac{k\pi }{2},\ k\in {\mathbb {Z}}\}\). Then for \(N\in {\mathbb {N}}\)
Proof
We will prove case a), the remaining cases are proven analogously.
Since by assumption \(\alpha ^2< 1\) and \(\alpha \ne 0\), there exist \(\beta ,\gamma ,\delta \in {\mathbb {R}},\ \beta \ne 0\) such that \(\alpha ^2+\beta ^2+\gamma ^2+\delta ^2=1\) and \(q=\left( 1+\frac{\beta }{\alpha }i+\frac{\gamma }{\alpha }j+\frac{\delta }{\alpha }k\right) \in {\mathbb {H}}\). Let \({{\,\mathrm{Im}\,}}_i(q)\) denote the imaginary part of q corresponding to the imaginary unit i and—in this proof only—we denote
Then
Now, from (33) and (34) we finally get (46). \(\square \)
Remark 8
In view of (5), \(\alpha , \beta ,\gamma ,\delta \) can be any real numbers such that \(\alpha \beta \gamma \delta \ne 0\) and the above statements remain true if \(\displaystyle {\cos \theta =\frac{\alpha }{|q|}},\) \(\displaystyle {\sin \theta =\frac{\sqrt{\beta ^2+\gamma ^2+\delta ^2}}{|q|}}\).
5 OEIS
Note that some Quaternacci sequences can be found in OEIS [25], e.g.:
-
\(A_{n}(1,0,0)=A146559(n)\) is a difference counterpart of cosine (see [16]);
-
For \(\varphi =\arccos \left( \frac{1}{3}\right) \) we have \(\cos (n \varphi )=\frac{A_{n}(2,2,0)}{3^n}=\frac{A025172(n)}{3^n}\). The sequence appears in the negative answer to the third Hilbert Problem (see [1]);
-
\(B_{n}(1,0,1)=D_{n} (1,0,1)=A088137(n),\ B_{n}(1,1,1)=C_{n} (1,1,1)=D_{n}(1,1,1) = A_{n-1}(1,-1,1)+B_{n-1}(1,-1,1)+C_{n-1}(1,-1,1)+D_{n-1}(1,-1,1) =A088138(n)\) are Generalized Gaussian Fibonacci integers;
-
\(A_{n}(1,-1,-1)+B_{n}(1,-1, -1)+C_{n}(1,-1,-1) +D_{n}(1,-1,-1)=A_{n}(1,1,1)- C_{n}(1,1,1)=A120580(n)\) is the Hankel transform of the sequence of sums \({\sum _{k=0}^n {2k \atopwithdelims ()k}}\);
-
\(|A_{n}(0,0,-1)|=|A009116(n)|\) and members of A009116(n) are coefficients in the expansion of \(\frac{\cos x}{e^x}\);
-
We have \(2^nA_{n}\left( \frac{1}{2},\frac{1}{2},\frac{1}{2}\right) =A213421(n)=\text {Re}\bigl ((2+i+j+k)^n\bigr )\) and also \(2^n\left( A_{n}\left( 0,\frac{1}{2},\frac{1}{2}\right) +C_{n}\left( 0,\frac{1}{2},\frac{1}{2}\right) +D_{n}\left( 0,\frac{1}{2},\frac{1}{2}\right) \right) =A190965(n+1) ={{\,\mathrm{Im}\,}}_j\bigl ((2+j+k)^{n +1}\bigr )\), where \({{\,\mathrm{Im}\,}}_j\) denotes the imaginary part of a quaternion corresponding to the imaginary unit j. Moreover, \(2^nA_{n}\left( 0,\frac{1}{2},\frac{1}{2}\right) = A266046(n)=\text {Re}\bigl ((2+j+k)^n\bigr )\). Also, in the description of the later sequence we have found a remark by Stanislav Sykora in which he suggests to consider the coefficients of 1, i, j, k in powers of a quaternion as recurrent sequences of coefficients of 1, i, j, k in that quaternion;
-
\(\displaystyle {\frac{B_{n}(1,0,1)}{A_{n}(1,0, 1)}=\frac{D_{n}(1,0,1)}{A_{n}(1,0, 1)}=\frac{A088137(n)}{A087455(n)}=\underbrace{1 \circ 1\circ \cdots \circ 1}_{n \text { times}}},\) where \(a\circ b=\displaystyle {\frac{a+b}{1-2ab}}\).
In OEIS there is a number of Quaternacci sequences connected with Lucas sequences of the first and second kind, e.g.:
-
\(A_{n}(1,0,1)=A_{n}(0,1, -1)=A_{n}(1,0,-1)=A087455(n) =\frac{1}{2}(-1)^n V_n(-2,3) \);
-
\(A_{n}(2,0,0)=A006495(n)=\frac{(-1)^n}{2} V_n(-2,5)\);
-
\(B_{n}(1,1,1)=C_{n} (1,1,1)=D_{n} (1,1,1)=A088138(n)=(-1)^{n+1}U_{n}(-2,4) =U_{n}(2,4)\);
-
\(B_{n}(1,0,1)=D_{n}(1,0,1)=A088137(n)=U_n(2,3)\);
-
\(2^n A_{n}\left( \frac{1}{2},\frac{1}{2},\frac{1}{2}\right) =A213421(n)=V_n(4,7)\);
-
\(2^nA_{n}\left( \frac{1}{2},0,0\right) =A139011(n)=V_n(4,5)\);
-
\(2^nB_{n}\left( \frac{1}{2},0,0\right) =A099456(n)=U_n(4,5)\);
-
From (23) we get
$$\begin{aligned} 2^nB_{n}\left( \frac{1}{2},\frac{1}{2},\frac{1}{2}\right) =2^nC_{n}\left( \frac{1}{2},\frac{1}{2},\frac{1}{2}\right) =2^nD_{n}\left( \frac{1}{2},\frac{1}{2},\frac{1}{2}\right) =U_n(4,7). \end{aligned}$$Hence from Theorem 7 we obtain the following reduction formulae
$$\begin{aligned} V_{n^2}(4,7)=V_n^n(4,7)A_{n}\left( \frac{U_n(4,7)}{V_n(4,7)},\frac{U_n(4,7)}{V_n(4,7)},\frac{U_n(4,7)}{V_n(4,7)}\right) . \end{aligned}$$A similar argument can be used for \(A_{n}(1,1,1)=A138230(n)={\mathcal {U}}_n\), where \({\mathcal {U}}_n=2{\mathcal {U}}_{n-1}-4{\mathcal {U}}_{n-2}\), \({\mathcal {U}}_0={\mathcal {U}}_1=1\), which gives
$$\begin{aligned} {\mathcal {U}}_{n^2}={\mathcal {U}}_{n}^nA_{n}\left( \frac{(-1)^{n+1}U_n(2,4)}{{\mathcal {U}}_n},\frac{(-1)^{n+1}U_n(2,4)}{{\mathcal {U}}_n},\frac{(-1)^{n+1}U_n(2,4)}{{\mathcal {U}}_n}\right) . \end{aligned}$$
We have also found some other Quaternaccis (we omit their descriptions):
-
\(A_{n}(1,2,0)=A_{n}(2,1,0)=A_{n}(2,0,1)=A138229(n);\)
-
\(A_{n}(1,1,1)=A138230(n);\)
-
\(A_{n}(0,0,-1)=A146559(n);\)
-
\(B_{n}(1,2,0)=A088139(n);\)
-
\(A_{n}(1,1,1)+B_{n}(1,1,1)+C_{n}(1,1,1)+D_{n}(1,1,1)=A137717(n);\)
-
\(3^n A_{n}\left( \frac{2}{3},0,0\right) =A121622(n);\)
-
\(3^nB_{n}\left( \frac{1}{3},0,0\right) =A106392(n+1);\)
-
\(2^nA_{n}\left( \frac{1}{2},1,1\right) =2^nA_{n}\left( \frac{3}{2},0,0\right) =A121621(n);\)
-
\(2^nB_{n}\left( \frac{1}{2},1,\frac{1}{2}\right) =A190968(n);\)
-
\(2^nB_{n}\left( \frac{1}{2},1,0 \right) =A190967(n);\)
-
\(2^n\left( A_{n}\left( \frac{1}{2},1,0\right) +B_{n}\left( \frac{1}{2},1,0\right) -C_{n}\left( \frac{1}{2},1,0\right) \right) =A176333(n).\)
Note that many Quaternacci sequences for small integer values of b, c, d does not appear in OEIS in any form, e.g. \(B_{n}(2,1,0),C_{n}(2,1,0),B_{n}(2,2,0),\) \(C_{n}(2,2,0),\ A_{n}(1,-1,1),\ B_{n}(1,-1,1),\ C_{n}(1,-1,1),\ D_{n}(1,-1,1),A_{n}(1,0,3),\) \(B_{n}(1,0,3),\ D_{n}(1,0,3),\ A_{n}(-1,3,1),\ B_{n}(-1,3,1),\ C_{n}(-1,3,1),\ A_{n}(3,1,0),\) \(B_{n}(3,1,0),C_{n}(3,1,0), A_{n}(3,0,0)\).
Notes
Defined recurrently as follows: \(L_0=2,\ L_1=1,\ L_n=L_{n-1}+L_{n-2}\) for \(n\geqslant 2\).
\(T_n(\cos \theta )=\cos n \theta , \quad {U}_n(\cos \theta )=\frac{\sin ((n+1)\theta )}{\sin \theta }\) for \(\theta \in {\mathbb {R}},\ n\in {\mathbb {N}}_0\).
That is, a product of matrices representing p, q equals to the matrix representing pq.
The last formula is not given explicitly because of its size
References
Boltjansky, V.G.: Hilbert’s Third Problem (Scripta Series in Mathematics). Wiley, Oxford (1978)
Cho, E.: De Moivre’s formula for quaternions. Appl. Math. Lett. 11(6), 33–35 (1998)
Çimen, C.B., İpek, A.: On Pell quaternions and Pell–Lucas quaternions. Adv. Appl. Clifford Algebras 26, 39–51 (2016)
Farebrother, R., Groß, J., Troschke, S.O.: Matrix representation of quaternions. Linear Algorithm Appl. 362, 251–255 (2003)
Flaut, C., Shpakivskyi, V.: On generalized Fibonacci quaternions and Fibonacci–Narayana quaternions. Adv. Appl. Clifford Algebras 23, 673–688 (2013)
Halici, S.: On Fibonacci quaternions. Adv. Appl. Clifford Algebras 22, 321–327 (2012)
Halici, S., Karatas, A.: On a generalizaton for Fibonacci quaternions. Chaos Solitons Fractals 98, 178–182 (2017)
Handscomb, D.C., Mason, J.C.: Chebyshev Polynomials. CRC, Boca Raton (2003)
Hetmaniok, E., Różański, M., Słota, D., Szweda, M., Trawiński, T., Wituła, R.: Determinants of the block arrowhead matrices. In: Wituła, R., Bajorska-Harapińska, B., Hetmaniok, E., Słota, D., Trawiński, T. (eds.) Selected Problems on Experimental Mathematics, pp. 73–88. Gliwice, Wydawnictwo Politechniki Śląskiej (2017)
Hetmaniok, E., Piątek, B., Wituła, R.: Binomials transformation formulae for scaled Fibonacci numbers. Open Math. 15, 477–485 (2017)
Hetmaniok, E., Pleszczyński, M., Słota, D., Wituła, R.: \(\delta \)-Fibonacci and \(\delta \)-Lucas numbers, \(\delta \)-Fibonacci and \(\delta \)-Lucas polynomials. Math. Slovaca 67, 51–70 (2017)
Horadam, A.F.: Complex Fibonacci numbers and Fibonacci quaternions. Am. Math. Mon. 70, 289–291 (1963)
Koecher, M., Remmert, R.: Hamilton’s Quaternions. In: Ebbinghaus, H.-D., Hermes, H., Hirzebruch, F., Koecher, M., Mainzer, K., Neukirch, J., Prestel, A., Remmert, R. (eds.) Numbers. Springer, Berlin (1995)
Koshy, T.: Fibonacci and Lucas Numbers with Applications. Wiley, Oxford (2011)
Rodman, L.: Topics in Quaternion Linear Algebra. Princeton University Press, Princeton (2014)
Shevelev, V.: Combinatorial Identities Generated by Difference Analogs of Hyperbolic and Trigonometric Functions of Order \(n\). arXiv:1706.01454v4
Smoleń, B., Wituła, R.: Two-parametric quasi-Fibonacci numbers. Silesian J. Pure Appl. Math. 7, 1, 99–121 (2017). mat.polsl.pl/sjpam/zeszyty/z7/Silesian\_J\_Pure\_Appl\_Math\_v7\_i1\_str\_099-121.pdf. Accessed 18 May 2019
Tan, E., Yilmaz, S., Sahin, M.: On a new generalization of Fibonacci quaternions. Chaos Solitons Fractals 82, 1–4 (2016)
Vajda, S.: Fibonacci and Lucas Numbers, and the Golden Section. Theory and Applications. Dover Publ. Inc., Mineola (2008)
Wituła, R.: On Some Applications of Formulae for Sums of Unimodular Complex Numbers. Wydawnictwo Pracowni Komputerowej Jacka Skalmierskiego, Gliwice (2011). (in Polish)
Wituła, R., Słota, D., Warzyński, A.: Quasi-Fibonacci numbers of the seventh order. J. Int. Seq. 9 (2006) *(Article 06.4.3)
Wituła, R., Słota, D.: Quasi-Fibonacci numbers of order 11. J. Int. Seq. 10 (2007) (Article 07.5.6)
Wituła, R., Słota, D.: \(\delta \)-Fibonacci numbers. Appl. Anal. Discret. Math. 3, 310–329 (2009)
Wyss, W.: Two Non-commutative Binomial Theorems. arXiv:1707.03861v2
The On-line Encyclopedia of Integer Sequences, founded by J.N.A. Sloane. https://oeis.org/. Accessed 18 May 2019
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Vladislav Kravchenko.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1
We present here two methods for exponentiation of some arrowhead matrices. The first generalizes formula (17) to matrices of order \(n\times n\), \(n \in {\mathbb {N}}\) which are antisymmetric transition matrices for some recurrence relations. The second method is more general as it can be used for finding powers of any arrowhead matrix, not only antisymmetric or even not only the one defining some recurrence relation.
Method 1 Let \(a_1,a_2,\ldots , a_n \in {\mathbb {R}}\), \(a_1^2+a_2^2+\cdots +a_n^2\ne 0\) and denote
If the sequences \(A_1,\ldots A_{n+1}\) are such that \(A_{1,0}=1, A_{2,0}=\cdots =A_{n+1,0}=0\) and
then it can be proven by induction that
where
Note that (17) is a special case of the above formula, for \(A_1=A, A_2=B,\) \(A_3=C,A_4=D\) and \(a_1=b, a_2=c, a_3=d\).
Method 2 Now consider an arrowhead matrix S over \({\mathbb {C}}\) written in a block form
where \(k,l\in {\mathbb {N}}, \omega \in {\mathbb {C}}, X\) is a complex matrix of order \({k\times l}\) and \(I_{k\times k}\) is the identity matrix of order k. As it was shown in [9], the determinant of such matrix is
Now we define recurrently for any matrix Y over \({\mathbb {C}}\) of order \({r\times s}\)
Note that \(Y^{(n)}\) is of order \(r \times r\) for even n whereas of order \(r\times s\) for odd. Moreover,
Then it follows by induction for \(n\in {\mathbb {N}}\) that
Note that Method 1 is a special case of this one for \(\omega =-1, X=[a_1,\ldots , a_n]\).
Appendix 2: Number Systems Related to Quaternaccis
Halici and Karatas in 2017 in paper [7] considered generalized Fibonacci quaternions, based on so-called general Fibonacci sequence \((W_n)_{n=0}^\infty \) defined as follows
Then the generalized Fibonacci quaternions (gFQ) are the ones of the form
Note that Fibonacci and Lucas quaternions described before are the special cases of gFQ, namely \(Q_n=QW_n(0,1,1,1), K_n=QW_n(2,1,1,1)\).
Also, two sequences considered by Çimen and İpek in 2016 [3] are gFQ, namely Pell quaternions \(QP_n=QW_n(0,1,2,1)\) and Pell–Lucas quaternions \(QPL_n=QW_n(2,1,2,1)\), built up from the classic Pell and Pell–Lucas sequences, respectively.
The Binet’s formula for gFQ is
where
These quaternions satisfy the following recurrence relation
where
Note that, similarly as Fibonacci quaternions, qeneralized Fibonacci quaternions are related to Quaternaccis by the following identity valid for all \(m,n \in {\mathbb {N}}_0\):
To give the final formulae we need the equivalent for the binomial theorem in noncommutative algebras (see [24]), from which for all \(q_1,q_2\in {\mathbb {H}}\) we get
where \(L_n=L_n(q_2,q_1)=d_{q_2}q_1^{n-1}+(q_1+d_{q_2})L_{n-1}(q_2,q_1)\), \(L_0(q_2,q_1)=0\) and \(d_{q_2}=d_{q_2}(X)=q_2X-Xq_2\) is a linear operator on \({\mathbb {H}}\).
Theorem 9
Let \(q_1=R\alpha ^n(1+\alpha i+\alpha ^2j+\alpha ^3k),\ q_2=-S\beta ^n(1+\beta i+\beta ^2 j+\beta ^3 k)\), where \(\alpha ,\beta , R,S\) are as in (51). Then for all \(m,n\in {\mathbb {N}}\)
where for \(X\in \{A,B,C,D\}\) we denote
Similar identities hold for \(C_{m}\left( \frac{W_{n +1}}{W_n},\frac{W_{n+ 2}}{W_n},\frac{W_{n +3}}{W_n}\right) \) and \(D_{m}\left( \frac{W_{n+1}}{W_n}, \frac{W_{n+2}}{W_n},\frac{W_{n+3}}{W_n} \right) \).
Proof
From (51) it follows that
whence from (52) we get
Transforming the right side yields
Thus the statement follows from comparison of coefficients to 1, i, j, k in (55) and (56). \(\square \)
A Table of Values for Quaternaccis for \(n\leqslant 6\).
n | \(A_{n}(b,c,d)\) | \(B_{n}(b,c,d)\) |
---|---|---|
0 | 1 | 0 |
1 | 1 | b |
2 | \(1 - b^2 - c^2 - d^2\) | 2b |
3 | \(1 - 3 b^2 - 3 c^2 - 3 d^2\) | \(3 b - b^3 - b c^2 - b d^2\) |
4 | \(\begin{array}{l}1 - 6 b^2 + b^4 - 6 c^2 + 2 b^2 c^2\\ + c^4 - 6 d^2 + 2 b^2 d^2 + 2 c^2 d^2 + d^4\end{array}\) | \(4 b - 4 b^3 - 4 b c^2 - 4 b d^2\) |
5 | \(\begin{array}{l}1 - 10 b^2 + 5 b^4 - 10 c^2 + 10 b^2 c^2 \\ + 5 c^4 - 10 d^2 + 10 b^2 d^2 \\ + 10 c^2 d^2 + 5 d^4\end{array}\) | \(\begin{array}{c}5 b - 10 b^3 + b^5 - 10 b c^2 \\ + 2 b^3 c^2 + b c^4 - 10 b d^2 + 2 b^3 d^2 \\ + 2 b c^2 d^2 + b d^4 \end{array}\) |
6 | \(\begin{array}{l} 1 - 15 b^2 + 15 b^4 - b^6 - 15 c^2\\ + 30 b^2 c^2 - 3 b^4 c^2 + 15 c^4 - 3 b^2 c^4\\ - c^6 - 15 d^2 + 30 b^2 d^2 - 3 b^4 d^2 \\ + 30 c^2 d^2 - 6 b^2 c^2 d^2 - 3 c^4 d^2 + 15 d^4 \ 3 b^2 d^4 - 3 c^2 d^4 - d^6 \end{array}\) | \(\begin{array}{l} 6 b - 20 b^3 + 6 b^5 - 20 b c^2 \\ + 12 b^3 c^2 + 6 b c^4 - 20 b d^2 + 12 b^3 d^2 \\ + 12 b c^2 d^2 + 6 b d^4 \end{array}\) |
n | \(C_{n}(b,c,d)\) | \(D_{n}(b,c,d)\) |
---|---|---|
0 | 0 | 0 |
1 | c | d |
2 | 2c | 2d |
3 | \(3 c - b^2 c - c^3 - c d^2\) | \(3 d - b^2 d - c^2 d - d^3\) |
4 | \(4 c - 4 b^2 c - 4 c^3 - 4 c d^2\) | \(4 d - 4 b^2 d - 4 c^2 d - 4 d^3\) |
5 | \(\;\begin{array}{l}5 c - 10 b^2 c + b^4 c - 10 c^3 \\ +2 b^2 c^3 + c^5 - 10 c d^2 + 2 b^2 c d^2\\ + 2 c^3 d^2 + c d^4\end{array}\;\) | \(\;\begin{array}{c}5 d - 10 b^2 d + b^4 d - 10 c^2 d \\ + 2 b^2 c^2 d + c^4 d - 10 d^3 + 2 b^2 d^3\\ + 2 c^2 d^3 + d^5\end{array}\;\) |
6 | \(\begin{array}{l} 6 c - 20 b^2 c + 6 b^4 c - 20 c^3 \\ + 12 b^2 c^3 + 6 c^5 - 20 c d^2 + 12 b^2 c d^2 \\ + 12 c^3 d^2 + 6 c d^4 \end{array}\) | \(\begin{array}{l} 6 d - 20 b^2 d + 6 b^4 d - 20 c^2 d \\ + 12 b^2 c^2 d + 6 c^4 d - 20 d^3 + 12 b^2 d^3 \\ + 12 c^2 d^3 + 6 d^5 \end{array}\) |
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Bajorska-Harapińska, B., Smoleń, B. & Wituła, R. On Quaternion Equivalents for Quasi-Fibonacci Numbers, Shortly Quaternaccis. Adv. Appl. Clifford Algebras 29, 54 (2019). https://doi.org/10.1007/s00006-019-0969-9
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00006-019-0969-9