Abstract
In this paper, we present a construction to extend number system by using the vector generators of the matrix algebra \(L_{0,n}(\mathbb {R})\) that is isomorphic with the Clifford algebra \(C\ell _{0,n}\). Moreover, we investigate the existence of solutions for some linear equations in \(C\ell _{0, 3}\) by means of real matrix representation.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
In 1878, Clifford algebras having n number of generators were discovered. Various useful and important results related with Clifford algebras have been studied up to recently.
Let \(\mathbb {R}^{0,n}\) be the standard n-dimensional pseudo-Euclidean space endowed with the quadratic form of signature (0, n). Furthermore, let \(C\ell _{0,n}\) be the corresponding real Clifford algebra of \(\mathbb {R}^{0,n}\).
Some different matrix representations of \(C\ell _{0,n}\) have been established and \(C\ell _{0,n}\) is isomorphic to a matrix algebra [1, 2, 4, 8]. The \(2^{n}\)-dimensional matrix algebra \(L_{2^{n}}(\mathbb {R})\) is proved to be isomorphic to \(C\ell _{0,n}\) [2]. Throughout this paper, we will denote \(L_{2^{n}}(\mathbb {R})\) by \(L_{0, n}(\mathbb {R})\).
The vector generators of \(L_{p,q}(\mathbb {R})\) that is isomorphic to \(C\ell _{p,q}\) are obtained using the set of vector generators \(\left\{ g_{2}, g_{3}, g_{7}, \ldots , g_{2^{n}-1}\right\} \) of \(L_{0, n}(\mathbb {R})\) by multiplying special diagonal matrices to the vector generators \(g_{2}, g_{3}, g_{7}, \ldots , g_{2^{n}-1}\) [5]. Thus, algebraic properties of \(C\ell _{p,q}\) can be derived from the properties of \(C\ell _{0,n}\).
The Cayley–Dickson construction is a well-known process to extend number system. Using the Cayley–Dickson construction, one can obtain complex numbers, quaternions, octonions and sedenions from real numbers in the sequel. A lot of works associated with Cayley–Dickson process have been done [1, 3, 7].
In Sect. 2, We establish a method to construct \(L_{0, n}(\mathbb {R})\) using \(L_{0, n-1}(\mathbb {R})\) and so one can obtain a sequence of algebras basically different from the algebras induced by the Cayley–Dickson construction. Since \(L_{0, 0}(\mathbb {R})\) is isomorphic to \(\mathbb {R}\) and \(L_{0, 1}(\mathbb {R})\) is isomorphic to \(\mathbb {C}\), likewise, we can extend the number system by using \(L_{0, n}(\mathbb {R})\).
In [6], Tian present some properties of quaternion matrices and investigate the solution of some linear equations over real quaternion algebra \(\mathbb {H}\).
In Sect. 3, we investigate the existence of solutions for some linear equations such as \(ax=0, \,\,xa=0,\,\, ax=xa,\,\,ax-xb=c,\,\,axa=a\) over the Clifford algebra \(C\ell _{0,3}\) by means of matrix representation.
2 Extension of Number System by Means of \(L_{0,n}(\mathbb {R})\)
Recall that, the algebra \(L_{0, n}(\mathbb {R})\) is constructed through the following process [2]. Let
be the Pauli matrices. Then, we first define \(A^{(1)}=A_1^{1}\sigma _{0}+A_{2}^{1}\sigma _{2}(-i),\,\,\, A_1^{1}, A_{2}^{1}\in {\mathbb {R}}\) and define \(A^{(n)}\) by using \(A^{(n-1)}\in L_{0,n-1}(\mathbb {R})\) as follows. Let \(A^{(n-1)} \in L_{0,n-1}(\mathbb {R})\) and let the first column of \(A^{(n-1)}\) be \((A_1^{n-1} , A_2^{n-1} , \cdots , A_{2^{n-1} }^{n-1})^{\mathrm{T}}\). If i is odd, replace \(A_i^{n-1}\) by \( \left( \begin{array}{cc} A_{2i-1}^{n} &{} -A_{2i}^{n} \\ A_{2i}^{n} &{} A_{2i-1}^{n} \end{array} \right) \). If i is even, replace \(A_i^{n-1}\) by \(\left( \begin{array}{cc} A_{2i-1}^{n} &{} A_{2i}^{n} \\ A_{2i}^{n} &{} -A_{2i-1}^{n} \end{array} \right) \). Through this process, the \(2^{n} \times 2^{n}\) matrix \(A^{(n)}\) and \(L_{0, n}(\mathbb {R}) = \{ A^{(n)} \in M_{2^{n}}(\mathbb {R}) \,| \, A_{i}^{n} \in \mathbb {R} \,\, , \,\, i= 1 ,2, \ldots , 2^{n}-1 \}\) are obtained. Moreover, \(g_i \) is a \(2^n \times 2^n\) matrix in \(L_{0,n} (\mathbb {R})\) such that (i, 1)-th entry is 1, and (j, 1)-th entry is 0 for \(j \not = i\). Thus, \(g_{1}=I_{2^{n}}\) and \(L_{0,0}(\mathbb {R})=\left\{ a_{1}g_{1} | a_{1}\in \mathbb {R}\right\} \), \(L_{0,1}(\mathbb {R})=\left\{ a_{1}g_{1}+a_{2}g_{2} | a_{1}, a_{2}\in \mathbb {R}\right\} \). For \(n\ge 2\), \(g_{2}, g_{3}, g_{7}, \ldots , g_{2^{n}-1}\) are vector generators of \(L_{0,n}(\mathbb {R})\) and
is a vector space basis of \(L_{0,n}(\mathbb {R})\). In what follows, A in \(L_{0,n-1}(\mathbb {R})\) will be identified with the block matrix \( \left( \begin{array}{c@{\quad }c} A &{} O \\ O &{} A \end{array}\right) \) in \(L_{0,n}(\mathbb {R}).\)
Obviously, \(A \in L_{0,1}(\mathbb {R})\) can be expressed uniquely by \(A = a_{1}g_{1}+(b_{1}g_{1})g_{2}\) for some \(a_{1},b_{1} \in \mathbb {R}\). Furthermore, we can generalize the expression of elements in \(L_{0,n}(\mathbb {R})\) as following theorem.
Theorem 2.1
Let \(A \in L_{0,n}(\mathbb {R})\) for \(n\ge 2\). Then, A can be expressed uniquely as \(A = v_{1}+v_{2}g_{2^{n}-1}\) for some \(v_{i} \in L_{0,n-1}(\mathbb {R})\), \(i=1,2\).
Proof
Since \(\mathcal {B}\) is a vector space basis of \(L_{0,n}(\mathbb {R})\), \(A\in span(\mathcal {B})\), the spanned subspace of \(\mathcal {B}\). Thus, the terms in A can be expressed by the sum of terms without the factor of \(g_{2^{n}-1}\) and terms with the factor of \(g_{2^{n}-1}\). Let \(v_{1}\) be the sum of all the terms without the factor of \(g_{2^{n}-1}\) and let w be the sum of all the terms with the factor of \(g_{2^{n}-1}\). Then, \(w=v_{2}g_{2^{n}-1}\) for some \(v_{2}\), where \(v_{2}\) has no terms with factor of \(g_{2^{n}-1}\). Thus, \(v_{i} \in L_{0,n-1}(\mathbb {R})\), for all \(i=1,2\) and A can be expressed as \(A = v_{1}+v_{2}g_{2^{n}-1}\).
For the uniqueness, suppose that \(A = v_{1}+v_{2}g_{2^{n}-1}= w_{1}+w_{2}g_{2^{n}-1}\) for some \(v_{i}, w_{i} \in L_{0,n-1}(\mathbb {R})\), \(i=1,2\). Then, \(v_{1}=w_{1}\) and \(v_{2}g_{2^{n}-1}=w_{2}g_{2^{n}-1}\). Since \(g_{2^{n}-1}\) is invertible, \(v_{2}=w_{2}\) and so the uniqueness is proved. \(\square \)
Example 2.2
Let \(A \in L_{0,3}(\mathbb {R})\). Then, \(A=a_{1}g_{1}+a_{2}g_{2}+a_{3}g_{3}+a_{4}g_{2}g_{3} +b_{1}g_{7}+b_{2}g_{2}g_{7}+b_{3}g_{3}g_{7} +b_{4}g_{2}g_{3}g_{7}\) for some \(a_{i}, b_{i}\in \mathbb {R}\), \(i=1,2,3,4\) and so we can write \(A = v_{1}+v_{2}g_{7}\), where \(v_{1}=a_{1}g_{1}+a_{2}g_{2}+a_{3}g_{3}+a_{4}g_{2}g_{3}\) and \(v_{2}=b_{1}g_{1}+b_{2}g_{2}+b_{3}g_{3}+b_{4}g_{2}g_{3} \in L_{0,2}(\mathbb {R})\). \(\Box \)
To simplify the expression of the multiplication of elements in \(L_{0, n}(\mathbb {R})\), we define \(\overline{A}\), the conjugate of A, as follows.
Definition 2.3
Let \(A \in L_{0, n}(\mathbb {R})\). Then, we define the conjugate \(\overline{A}\) of A according to n as follows.
- (1)
For \(n=0\), \(\overline{A} = A\).
- (2)
For \(n=1\), \(\overline{A} = \overline{v_{1}}-\overline{v_{2}}g_{2}\), where \(A = v_{1}+ v_{2}g_{2}\), \(v_{i}\in L_{0, 0}(\mathbb {R})\), \(i=1,2\).
- (3)
For \(n\ge 2\), \(\overline{A} = \overline{v_{1}}-\overline{v_{2}}g_{2^{n}-1}\), where \(A = v_{1}+ v_{2}g_{2^{n}-1}\), \(v_{i} \in L_{0, n-1}(\mathbb {R})\), \(i=1,2\).
Note that the changes of signs of the terms of the conjugate of A in the above definition are not the same as those of the conjugate of elements obtained by the Cayley–Dickson construction.
Example 2.4
Let \(A = (a_{1}g_{1}+a_{2}g_{2})+(b_{1}g_{1}+b_{2}g_{2})g_{3}\in L_{0, 2}(\mathbb {R})\), \(a_{i}, b_{i} \in \mathbb {R}\), \(i=1,2\). Then,
The definition of conjugate yields the following properties.
Theorem 2.5
Let \(A, B \in L_{0, n}(\mathbb {R})\) and \(\alpha \in \mathbb {R}\). Then,
- (1)
\(\overline{\overline{A}}=A\).
- (2)
\(\overline{A+B}=\overline{A} + \overline{B}\).
- (3)
\(\overline{\alpha A}=\alpha \overline{A}\).
Proof
For \(n=0\), it is obvious. Now, assume that \(n\ge 1\). Let \(A=v_{1}+v_{2}g\) and \(B=w_{1}+w_{2}g\) for some \(v_{i}, w_{i}\in L_{0,n-1}(\mathbb {R})\), \(i=1,2\), where \(g=g_{2}\) for \(n=1\) and \(g=g_{2^{n}-1}\) for \(n\ge 2\). We will prove the theorem by the mathematical induction.
- (1)
If \(n=1\), then \(\overline{\overline{v_{i}}}=v_{i}\), \(i=1,2\) and
$$\begin{aligned} \overline{\overline{A}}=\overline{\overline{v_{1}+ v_{2}g_{2}}} = \overline{\overline{v_{1}}- \overline{v_{2}}g_{2}} = \overline{\overline{v_{1}}} + \overline{\overline{v_{2}}}g_{2} = v_{1}+ v_{2}g_{2}=A. \end{aligned}$$Now, assume that \(\overline{\overline{A}}=A\) for all \(A\in L_{0,k}(\mathbb {R})\). Then, for \(n=k+1\), \(A=v_{1}+ v_{2}g_{2^{k+1}-1}\) for some \(v_{1}, v_{2}\in L_{0,k}{(\mathbb R)}\) and by the mathematical induction hypothesis \(\overline{\overline{v_{i}}}=v_{i}\), \(i=1,2\). Thus,
$$\begin{aligned} \begin{array}{lll} \overline{\overline{A}}&{}=&{}\overline{\overline{v_{1}+ v_{2}g_{2^{k+1}-1}}} \,=\, \overline{\overline{v_{1}}- \overline{v_{2}}g_{2^{k+1}-1}} \,= \,\overline{\overline{v_{1}}} + \overline{\overline{v_{2}}}g_{2^{k+1}-1} \\ &{}=&{} v_{1}+ v_{2}g_{2^{k+1}-1}\,=\,A. \end{array} \end{aligned}$$ - (2)
If \(n=1\), then \(\overline{v_{i}+w_{i}}=\overline{v_{i}}+\overline{w_{i}}\), \(i=1,2\) and
$$\begin{aligned} \begin{array}{lll} \overline{A+B}&{}=&{}\overline{(v_{1}+w_{1})+ (v_{2}+w_{2})g_{2}} = \overline{(v_{1}+w_{1})}- \overline{(v_{2}+w_{2})}g_{2} \\ &{}=&{} (\overline{v_{1}}+\overline{w_{1}}) - (\overline{v_{2}}+\overline{w_{2}})g_{2} = (\overline{v_{1}}-\overline{v_{2}}g_{2}) + (\overline{w_{1}}-\overline{w_{2}}g_{2}) \\ &{}=&{}\overline{A}+\overline{B}. \end{array} \end{aligned}$$Assume that \(\overline{A+B}=\overline{A}+\overline{B}\) for \(A, B\in L_{0,k}(\mathbb {R})\). For \(n=k+1\), \(A=v_{1}+v_{2}g_{2^{k+1}-1}\) and \(B=w_{1}+w_{2}g_{2^{k+1}-1}\) for some \(v_{i}, w_{i}\in L_{0,k}(\mathbb {R})\), \(i=1,2\). Then, \(\overline{v_{i}+w_{i}}=\overline{v_{i}}+\overline{w_{i}}\), \(i=1,2\) by the mathematical induction hypothesis. Thus,
$$\begin{aligned} \begin{array}{lll} \overline{A+B}&{}=&{}\overline{(v_{1}+w_{1})+ (v_{2}+w_{2})g_{2^{k+1}-1}} = \overline{(v_{1}+w_{1})}- \overline{(v_{2}+w_{2})}g_{2^{k+1}-1} \\ &{}=&{} (\overline{v_{1}}+\overline{w_{1}}) - (\overline{v_{2}}+\overline{w_{2}})g_{2^{k+1}-1} = (\overline{v_{1}}-\overline{v_{2}}g_{2^{k+1}-1}) + (\overline{w_{1}}-\overline{w_{2}}g_{2^{k+1}-1}) \\ &{}=&{}\overline{A}+\overline{B}. \end{array} \end{aligned}$$ - (3)
If \(n=1\), then \(\overline{\alpha v_{i}}=\alpha \overline{v_{i}}\), \(i=1,2\) and
$$\begin{aligned} \overline{\alpha A}=\overline{\alpha v_{1}+ \alpha v_{2}g_{2}} = \overline{\alpha v_{1}}- \overline{\alpha v_{2}}g_{2} = \alpha \overline{v_{1}} - \alpha \overline{v_{2}}g_{2}=\alpha \overline{A}. \end{aligned}$$Now, assume that \(\overline{\alpha A}=\alpha \overline{A}\) for all \(A\in L_{0,k}(\mathbb {R})\). For \(n=k+1\), \(A=v_{1}+ v_{2}g_{2^{k+1}-1}\) for some \(v_{1}, v_{2}\in L_{0,k}{(\mathbb R)}\). Then, \(\overline{\alpha v_{i}}=\alpha \overline{v_{i}}\), \(i=1,2\) by the mathematical induction hypothesis and hence
$$\begin{aligned} \begin{array}{lll} \overline{\alpha A}&{}=&{}\overline{\alpha v_{1}+ \alpha v_{2}g_{2^{k+1}-1}} = \overline{\alpha v_{1}}- \overline{\alpha v_{2}}g_{2^{k+1}-1} = \alpha \overline{v_{1}} - \alpha \overline{v_{2}}g_{2^{k+1}-1} \\ &{}=&{} \alpha (\overline{v_{1}}-\overline{v_{2}}g_{2^{k+1}-1})=\alpha \overline{A}. \end{array} \end{aligned}$$
\(\Box \)
Theorem 2.6
If \(w=ag_{1}\) for some \(a\in \mathbb {R}\), then \(g_{2}w=\overline{w}g_{2}\). Furthermore, if \(w\in L_{0,n-1}(\mathbb {R})\), then \(g_{2^{n}-1}w=\overline{w}g_{2^{n}-1}\) for \(n\ge 2\).
Proof
If \(w=ag_{1}\), then \(\overline{w}=\overline{ag_{1}}=ag_{1}\) and so \(g_{2}w=\overline{w}g_{2}\). For the case of \(n\ge 2\), we will prove by the mathematical induction. For \(n=2\), \(w=u_{1}+u_{2}g_{2}\) for some \(u_{1}=ag_{1}\) and \(u_{2}=bg_{1}\), \(a, b \in \mathbb {R}\). Thus, \(g_{3}w=g_{3}(u_{1}+u_{2}g_{2})=u_{1}g_{3}+u_{2}g_{3}g_{2}=(u_{1}-u_{2}g_{2})g_{3}= \overline{w}g_{3}\) and the theorem is true for \(n=2\). Now, assume that \(g_{2^{k}-1}{w}=\overline{w}g_{2^{k}-1}\) for \(w \in L_{0, k-1}(\mathbb {R})\). Note that \(g_{2^{k+1}-1}{u}=\overline{u}g_{2^{k+1}-1}\) for \(u \in L_{0, k-1}(\mathbb {R})\) since \(g_{i}g_{j}=-g_{j}g_{i}\) for \(i\ne j \ge 2\). If \(w = u_{1}+ u_{2}g_{2^{k}-1}\in L_{0, k}(\mathbb {R})\) for some \(u_{1}, u_{2} \in L_{0, k-1}(\mathbb {R})\), then \(\overline{w} = \overline{u_{1}} - \overline{u_{2}}g_{2^{k}-1}\) and
Thus, it satisfies for the case of \(n=k+1\) and the theorem is proved. \(\Box \)
The multiplication AB of A and B in \(L_{0, n}(\mathbb {R})\) can be expressed by using the conjugate as follows:
Corollary 2.7
Let \(A, B \in L_{0, n}(\mathbb {R})\) and let \(A = v_{1}+ v_{2}g_{2^{n}-1}\), \(B = w_{1}+ w_{2}g_{2^{n}-1}\) for some \(v_{i}, w_{i} \in L_{0, n-1}(\mathbb {R})\), \(i=1,2\). Then, \(AB=(v_{1}w_{1}-v_{2}\overline{{w}_{2}})+(v_{1}w_{2} +v_{2}\overline{{w}_{1}})g_{2^{n}-1}\).
Proof
Note that
But,
and so \(AB=(v_{1}w_{1}-v_{2}\overline{w}_{2})+ (v_{1}w_{2}+v_{2}\overline{w}_{1})g_{2^{n}-1}.\)\(\Box \)
Corollary 2.8
Let \(A, B \in L_{0, n}(\mathbb {R})\). Then, \(\overline{AB}=\overline{A}\,\, \overline{B}\).
Proof
For \(n=0\), it is obvious. Now, assume that \(n\ge 1\). Let \(A=v_{1}+v_{2}g\) and \(B=w_{1}+w_{2}g\) for some \(v_{i}, w_{i}\in L_{0,n-1}(\mathbb {R})\), \(i=1,2\), where \(g=g_{2}\) for \(n=1\) and \(g=g_{2^{n}-1}\) for \(n\ge 2\). We will prove the theorem by the mathematical induction.
If \(n=1\), then \(\overline{v_{i}}=v_{i}, \overline{w_{i}}=w_{i}\), \(\overline{v_{i}w_{j}}=v_{i}w_{j}\), \(i=1,2,\,\,j=1,2\) and so
Now, assume that \(\overline{AB}=\overline{A}\,\, \overline{B}\) for all \(A, B\in L_{0,k}(\mathbb {R})\).
For \(n=k+1\), \(A=v_{1}+v_{2}g_{2^{k+1}-1}\) and \(B=w_{1}+w_{2}g_{2^{k+1}-1}\) for some \(v_{i}, w_{i}\in L_{0,k}(\mathbb {R})\), \(i=1,2\). Then, \(\overline{v_{i}w_{j}}=\overline{v_{i}}\,\, \overline{w_{j}}\), \(i=1,2,\,\,j=1,2\) by the mathematical induction hypothesis and so
\(\Box \)
Remark 2.9
Note that, if we identify \(A = v_{1}+ v_{2}g_{2^{n}-1}\) and \(B = w_{1}+ w_{2}g_{2^{n}-1}\) with \(A = (v_{1},v_{2})\) and \(B = (w_{1}, w_{2})\), respectively, then we may write
Obviously, the multiplication is basically different from the multiplication in the Cayley–Dickson construction.
It is well known that \(\mathbb {R} \,\,\rightarrow \,\, \mathbb {C} \,\,\rightarrow \,\, \mathbb {H} \,\,\rightarrow \,\, \cdots \) is an extension of number system through the Cayley–Dickson construction. Since
the following sequence of algebra extension is considered as another extension of number system obtained naturally:
If we let \(e_{1}, e_{2}, \ldots , e_{n}\) be the vector generators of \(C\ell _{0, n}\), then we may assume that
since \(L_{0, n}(\mathbb {R})\cong C\ell _{0, n}\).
Theorem 2.10
The Clifford algebra \(C\ell _{0, n}\) can be constructed from \(C\ell _{0, n-1}\) for all n. Thus, the following sequence of Clifford algebras can be considered as an extension of number system naturally.
3 Solutions for Some Linear Equations in \(C\ell _{0,3}\)
In this section, we will consider the linear equations such as \(ax=0, \,\,xa=0,\,\, ax=xa,\,\,ax-xb=c,\,\,axa=a\) over the Clifford algebra \(C\ell _{0,3}\). Recall that \(C\ell _{0,3}\) is isomorphic to the matrix algebra \(L_{0, 3}(\mathbb {R})\) and so it is enough to establish the existence of solutions of the linear equations over the matrix algebra \(L_{0, 3}(\mathbb {R})\).
Let \(A\in L_{0, 3}(\mathbb {R})\). Then, \(A=v_{1}+v_{2}g_{7}\) for some \(v_{1}, v_{2}\in L_{0, 2}(\mathbb {R})\). Now, let \(v_{1}=a_{1}g_{1}+a_{2}g_{2}+a_{3}g_{3}+a_{4}g_{2}g_{3},\,\, v_{2}=b_{1}g_{1}+b_{2}g_{2}+b_{3}g_{3}+b_{4}g_{2}g_{3}\) for some \(a_{i}, b_{i}\in \mathbb {R}\). Define \(\overrightarrow{A}\) by
Moreover, define \(\phi _{i}(v_{j})\) and \(\psi _{i}(A)\) as follows.
By the straightforward calculations, the following proposition is obtained.
Proposition 3.1
Let \(A=v_{1}+v_{2}g_{7}\) and \(X \in L_{0,3}(\mathbb {R})\), where \(v_{1}=a_{1}g_{1}+a_{2}g_{2}+a_{3}g_{3}+a_{4}g_{2}g_{3},\,\, v_{2}=b_{1}g_{1}+b_{2}g_{2}+b_{3}g_{3}+b_{4}g_{2}g_{3}\) for some \(a_{i}, b_{i}\in \mathbb {R}\). Then,
- (1)
\(\overrightarrow{AX}=\psi _{1}(A)\overrightarrow{X}\)
- (2)
\(\overrightarrow{XA}=\psi _{2}(A)\overrightarrow{X}\).
By Proposition 3.1, we have the following properties.
Theorem 3.2
Let \(A, B \in L_{0,3}(\mathbb {R})\). Then,
- (1)
If \(A=g_{1}\), then \(\psi _{i}(A)=A\), \(i=1,2\).
- (2)
\(A=B\) if and only if \(\psi _{i}(A)=\psi _{i}(B)\), \(i=1,2\).
- (3)
A is invertible if and only if \(\psi _{i}(A)\) is invertible, \(i=1,2\). Furthermore,
\(\psi _{i}(A^{-1})=(\psi _{i}(A))^{-1}\).
Proof
Let \(X \in L_{0,3}(\mathbb {R})\). Then, (1) \(\overrightarrow{AX}=\psi _{1}(A)\overrightarrow{X}\) and \(\overrightarrow{AX}=\overrightarrow{X}=A\overrightarrow{X}\). Thus, \(\psi _{1}(A)\overrightarrow{X}=A\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and so \(\psi _{1}(A)=A\). Also, \(\overrightarrow{XA}=\psi _{2}(A)\overrightarrow{X}\) and \(\overrightarrow{XA}=\overrightarrow{X}=A\overrightarrow{X}\). Thus, \(\psi _{2}(A)\overrightarrow{X}=A\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and so \(\psi _{2}(A)=A\). (2) Obviously, \(A=B\) implies that \(\psi _{i}(A)=\psi _{i}(B)\), \(i=1,2\). Conversely, if \(\psi _{1}(A)=\psi _{1}(B)\), then \(\psi _{1}(A)\overrightarrow{X}=\psi _{1}(B)\overrightarrow{X}\) and so \(\overrightarrow{AX}=\overrightarrow{BX}\). Thus, \(\overrightarrow{(A-B)X}=O\) for all \(X \in L_{0,3}(\mathbb {R})\) and hence \(A=B\). Also, if \(\psi _{2}(A)=\psi _{2}(B)\), then \(\psi _{2}(A)\overrightarrow{X}=\psi _{2}(B)\overrightarrow{X}\) and so \(\overrightarrow{XA}=\overrightarrow{XB}\). Thus, \(\overrightarrow{X(A-B)}=O\) for all \(X \in L_{0,3}(\mathbb {R})\) and hence \(A=B\). (3) \(\psi _{1}(AA^{-1})=\psi _{1}(A)\psi _{1}(A^{-1})\) and \(\psi _{2}(AA^{-1})=\psi _{2}(A^{-1})\psi _{2}(A)\). But, \(\psi _{i}(AA^{-1})=\psi _{i}(g_{1})=g_{1}\), \(i=1,2\) and hence \(\psi _{i}(A^{-1})=\psi _{i}(A)^{-1}\), \(i=1,2\). \(\Box \)
Theorem 3.3
Let \(A, B \in L_{0,3}(\mathbb {R})\) and \(a\in \mathbb {R}\). Then,
- (1)
\(\psi _{i}(aA)=a\psi _{i}(A)\), \(i=1,2\).
- (2)
\(\psi _{i}(A+B)=\psi _{i}(A)+\psi _{i}(B)\), \(i=1,2\).
- (3)
\(\psi _{1}(AB)=\psi _{1}(A)\psi _{1}(B)\), \(\psi _{2}(AB)=\psi _{2}(B)\psi _{2}(A)\).
- (4)
\(\psi _{1}(A)\psi _{2}(B)=\psi _{2}(B)\psi _{1}(A)\).
- (5)
\(\psi _{i}\) preserves the similar matrix. That is, if \(B=CAC^{-1}\) for some invertible \(C\in L_{0,3}(\mathbb {R})\), then \(\psi _{1}(B)=\psi _{1}(C)\psi _{1}(A)\psi _{1}(C)^{-1}\) and \(\psi _{2}(B)=\psi _{2}(C)^{-1}\psi _{2}(A)\psi _{2}(C)\).
Proof
Let \(X \in L_{0,3}(\mathbb {R})\). Then,
- (1)
According to the result of Proposition 3.1, \(\overrightarrow{aAX}=\psi _{1}(aA)\overrightarrow{X}\) and \(a\overrightarrow{AX}=a\psi _{1}(A)\overrightarrow{X}\). Since \(\overrightarrow{aAX}=a\overrightarrow{AX}\), \(\psi _{1}(aA)\overrightarrow{X}=a\psi _{1}(A)\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and so \(\psi _{1}(aA)=a\psi _{1}(A)\). Also, \(\overrightarrow{X(aA)}=\psi _{2}(aA)\overrightarrow{X}\) and \(a\overrightarrow{XA}=a\psi _{2}(A)\overrightarrow{X}\). Since \(\overrightarrow{X(aA)}=a\overrightarrow{XA}\), \(\psi _{2}(aA)\overrightarrow{X}=a\psi _{2}(A)\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and so \(\psi _{2}(aA)=a\psi _{2}(A)\).
- (2)
\(\overrightarrow{(A+B)X}=\psi _{1}(A+B)\overrightarrow{X}\) and \(\overrightarrow{AX}+\overrightarrow{BX}= (\psi _{1}(A)+\psi _{1}(B))\overrightarrow{X}\). Since \(\overrightarrow{(A+B)X}=\overrightarrow{AX}+\overrightarrow{BX}\), \(\psi _{1}(A+B)\overrightarrow{X}=(\psi _{1}(A)+\psi _{1}(B))\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and so \(\psi _{1}(A+B)=\psi _{1}(A)+\psi _{1}(B)\). Also, \(\overrightarrow{X(A+B)}=\psi _{2}(A+B)\overrightarrow{X}\) and \(\overrightarrow{XA}+\overrightarrow{XB}=(\psi _{2}(A)+\psi _{2}(B))\overrightarrow{X}\). Since \(\overrightarrow{X(A+B)}=\overrightarrow{XA}+\overrightarrow{XB}\), \(\psi _{2}(A+B)\overrightarrow{X}=(\psi _{2}(A)+\psi _{2}(B))\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and so \(\psi _{2}(A+B)=\psi _{2}(A)+\psi _{2}(B)\).
- (3)
\(\overrightarrow{(AB)X}=\psi _{1}(AB)\overrightarrow{X}\) and \(\psi _{1}(A)\overrightarrow{BX}=\psi _{1}(A)\psi _{1}(B)\overrightarrow{X}\). Since \(\overrightarrow{(AB)X}=\psi _{1}(A)\overrightarrow{BX}\), \(\psi _{1}(AB)\overrightarrow{X}=\psi _{1}(A)\psi _{1}(B)\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\). Hence \(\psi _{1}(AB)=\psi _{1}(A)\psi _{1}(B)\). Also, \(\overrightarrow{X(AB)}=\psi _{2}(AB)\overrightarrow{X}\) and \(\psi _{2}(B)\overrightarrow{XA}=\psi _{2}(B)\psi _{2}(A)\overrightarrow{X}\). Since \(\overrightarrow{X(AB)}=\psi _{2}(B)\overrightarrow{XA}\), \(\psi _{2}(AB)\overrightarrow{X}=\psi _{2}(B)\psi _{2}(A)\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and thus, \(\psi _{2}(AB)=\psi _{2}(B)\psi _{2}(A)\).
- (4)
\(\overrightarrow{AXB}=\psi _{1}(A)\overrightarrow{XB}= \psi _{1}(A)\psi _{2}(B)\overrightarrow{X}\) and \(\overrightarrow{AXB}=\psi _{2}(B)\overrightarrow{AX}= \psi _{2}(B)\psi _{1}(A)\overrightarrow{X}\). Thus, \(\psi _{1}(A)\psi _{2}(B)\overrightarrow{X}=\psi _{2}(B)\psi _{1}(A)\overrightarrow{X}\) for all \(X \in L_{0,3}(\mathbb {R})\) and thus, \(\psi _{1}(A)\psi _{2}(B)=\psi _{2}(B)\psi _{1}(A)\).
- (5)
\(\psi _{1}(B)=\psi _{1}(CAC^{-1})=\psi _{1}(C)\psi _{1}(A)\psi _{1}(C)^{-1}\) and \(\psi _{2}(B)=\psi _{2}(CAC^{-1})=\psi _{2}(C)^{-1}\psi _{2}(A)\psi _{2}(C)\). \(\Box \)
Using the properties, we investigate the solutions of the following linear equations. The zero square matrix of any size will be denoted by O for the convenience of expression.
Theorem 3.4
Let \(A=(a_{1}g_{1}+a_{2}g_{2}+a_{3}g_{3}+a_{4}g_{2}g_{3})+ (b_{1}g_{1}+b_{2}g_{2}+b_{3}g_{3}+b_{4}g_{2}g_{3})g_{7} \in L_{0,3}(\mathbb {R})\), \(a_{i}, b_{i}\in \mathbb {R}\), \(i=1,2,3,4\). Then, the equation \(AX=O\) has a nontrivial solution over \(L_{0,3}(\mathbb {R})\) if and only if \((a_{1}=-b_{4}, a_{2}=b_{3}, a_{3}=-b_{2}, a_{4}=b_{2})\) or \((a_{1}=b_{4}, a_{2}=-b_{3}, a_{3}=b_{2}, a_{4}=-b_{2})\).
Proof
Since \(\overrightarrow{AX}=\overrightarrow{O}\) and \(\overrightarrow{AX}=\psi _{1}(A)\overrightarrow{X}\),
By the straightforward calculations,
where \(c=\sum _{i=1}^{i=4}(a_{i}^{2}+b_{i}^{2}), \,\,\, d=2\sum _{i=1}^{i=4}(-1)^{i+1}a_{i}b_{5-i}.\) Thus, if we let
then
Note that \(\hbox {det}(\psi _{1}(A))=0\) is equivalent to \(c^{2}-d^{2}=0\) which is equivalent to \((a_{1}=-b_{4}, a_{2}=b_{3}, a_{3}=-b_{2}, a_{4}=b_{2})\) or \((a_{1}=b_{4}, a_{2}=-b_{3}, a_{3}=b_{2}, a_{4}=-b_{2})\). Thus, the theorem is proved. \(\Box \)
Example 3.5
Let \(A=3g_{2}-g_{3}+g_{2}g_{3}-g_{7}+g_{2}g_{3}g_{7} \in L_{0,3}(\mathbb {R})\). Then, \(v_{1}=3g_{2}-g_{3}+g_{2}g_{3},\,\,\,v_{2}= -g_{1}+g_{2}g_{3}\). Since \(a_{1}=0,\,\,b_{4}=1\) and so the equation \(AX=O\) has only the trivial solution \(X=O\) over \(L_{0,3}(\mathbb {R})\).
Theorem 3.6
Let \(A=(a_{1}g_{1}+a_{2}g_{2}+a_{3}g_{3}+a_{4}g_{2}g_{3})+ (b_{1}g_{1}+b_{2}g_{2}+b_{3}g_{3}+b_{4}g_{2}g_{3})g_{7} \in L_{0,3}(\mathbb {R})\), \(a_{i}, b_{i}\in \mathbb {R}\), \(i=1,2,3,4\). Then, the equation \(XA=O\) has a nontrivial solution over \(L_{0,3}(\mathbb {R})\) if and only if \((a_{1}=-b_{4}, a_{2}=b_{3}, a_{3}=-b_{2}, a_{4}=b_{2})\) or \((a_{1}=b_{4}, a_{2}=-b_{3}, a_{3}=b_{2}, a_{4}=-b_{2})\).
Proof
Since \(\overrightarrow{XA}=\overrightarrow{O}\) and \(\overrightarrow{XA}=\psi _{2}(A)\overrightarrow{X}\),
By the straightforward calculations,
where \(c=\sum _{i=1}^{i=4}(a_{i}^{2}+b_{i}^{2})\) and \(d=2\sum _{i=1}^{i=4}(-1)^{i+1}a_{i}b_{5-i}.\) Using the similar process of the proof of previous theorem, the equation \(XA=O\) has a nontrivial solution if and only if \(\hbox {det}(\psi _{2}(A))=0\) which is equivalent to \((a_{1}=-b_{4}, a_{2}=b_{3}, a_{3}=-b_{2}, a_{4}=b_{2})\) or \((a_{1}=b_{4}, a_{2}=-b_{3}, a_{3}=b_{2}, a_{4}=-b_{2})\). \(\Box \)
According to Theorems 3.4 and 3.6, we know that the linear equation \(AX=O\) has a nontrivial solution if and only if \(XA=O\) has a nontrivial solution.
Theorem 3.7
Let \(A=v_{1}+v_{2}g_{7} \in L_{0,3}(\mathbb {R})\) for some \(v_{i}\in L_{0,2}(\mathbb {R}), \,\,i=1,2\). Then the equation \(AX=XA\) has a nontrivial solution \(X\in L_{0,3}(\mathbb {R})\).
Proof
Since \(\overrightarrow{AX}=\psi _{1}(A)\overrightarrow{X}\) and \(\overrightarrow{XA}=\psi _{2}(A)\overrightarrow{X}\), the equation \(AX=XA\) implies that \((\psi _{1}(A)-\psi _{2}(A))\overrightarrow{X}=O.\) But,
Note that all the entries of the first column and eighth column of \(\psi _{1}(A)-\psi _{2}(A)\) are zeros. If we let \(\overrightarrow{X}=(x_{1},x_{2},\ldots ,x_{8})^{\mathrm{T}}\), then \(x_{1}\) and \(x_{8}\) can be chosen arbitrary and hence \(AX=XA\) has a nontrivial solution. \(\Box \)
Example 3.8
Let \(\alpha , \beta , \gamma \in \mathbb {R}\). Then, obviously \(X=\alpha \, g_{1}+\beta A+\gamma \,g_{2}g_{3}g_{7}\) is a solution for the equation \(AX=XA\).
Theorem 3.9
Let \(A, B, C \in L_{0,3}(\mathbb {R})\) and \(A=v_{1}+v_{2}g_{7}, \,\,B=w_{1}+w_{2}g_{7}\) for some \(v_{i}, w_{i}\in L_{0,2}(\mathbb {R}), \,\,i=1,2\). Assume also that \(\hbox {det}(\phi _{1}(v_{1})-\phi _{3}(w_{1}))\ne 0\). Then the equation \(AX-XB=C\) has a unique solution \(X\in L_{0,3}(\mathbb {R})\) if \(\hbox {det}(\phi _{1}(v_{1})-\phi _{3}(w_{1}))\hbox {det}(D)\ne 0\), where \(D=\phi _{1}(v_{1})-\phi _{4}(w_{1})- (\phi _{2}(v_{2})-\phi _{3}(w_{2}))(\phi _{1}(v_{1})- \phi _{3}(w_{1}))^{-1}(-\phi _{2}(v_{2})+\phi _{4}(w_{2}))\).
Proof
Note that \(\overrightarrow{AX}=\psi _{1}(A)\overrightarrow{X}\) and \(\overrightarrow{XB}=\psi _{2}(B)\overrightarrow{X}\), where
Thus, \(AX-XB=C\) implies that \((\psi _{1}(A)-\psi _{2}(B))\overrightarrow{X}=\overrightarrow{C}\). Since
and \(\hbox {det}(\phi _{1}(v_{1})-\phi _{3}(w_{1}))\ne 0\), we have
where \(D=\phi _{1}(v_{1})-\phi _{4}(w_{1})- (\phi _{2}(v_{2})-\phi _{3}(w_{2}))(\phi _{1}(v_{1})- \phi _{3}(w_{1}))^{-1}(-\phi _{2}(v_{2})+\phi _{4}(w_{2}))\). Thus, the linear equation \(AX-XB=C\) has a unique solution if \(\hbox {det}(\phi _{1}(v_{1})-\phi _{3}(w_{1}))\hbox {det}(D)\ne 0\). \(\Box \)
Example 3.10
Let \(A=g_{1}-g_{3} \in L_{0,3}(\mathbb {R})\) and \(B=g_{2}+g_{3} \in L_{0,3}(\mathbb {R})\). Then, \(v_{1}=g_{1}-g_{3}, \,\,w_{1}=g_{2}+g_{3},\,\,v_{2}=O=w_{2}\) and so
Thus,
and
Therefore, for any \(C \in L_{0,3}(\mathbb {R})\), the linear equation \(AX-XB=C\) has a unique solution.
Theorem 3.11
Let \(A \in L_{0,3}(\mathbb {R})\). If \(\hbox {det}(\psi _{1}(A)\psi _{2}(A))\ne 0\), then the equation \(AXA=A\) has a unique solution over \(L_{0,3}(\mathbb {R})\).
Proof
Since \(\overrightarrow{A(XA-I_{n})}=\overrightarrow{O}\) and \(\overrightarrow{A(XA-I_{n})}=\psi _{1}(A)\overrightarrow{XA-I_{n}}=\psi _{1}(A)(\psi _{2}(A)\overrightarrow{X}-(1,0,0,\cdots , 0)^{\mathrm{T}})\), we have \(\psi _{1}(A)\psi _{2}(A)\overrightarrow{X}=Col_{1}(\psi _{1}(A))\), where \(Col_{1}(\psi _{1}(A))\) is the first column of \(\psi _{1}(A)\). By the assumption, \(\overrightarrow{X}=(\psi _{1}(A)\psi _{2}(A))^{-1}Col_{1}(\psi _{1}(A))\) and the solution X exists uniquely. \(\Box \)
References
Flaut, C., Shpakivskyi, V.: Some identities in algebras obtained by the Cayley–Dickson process. Adv. Appl. Clifford Algebras 23, 63–76 (2013)
Lee, D., Song, Y.: Explicit matrix realization of Clifford algebras. Adv. Appl. Clifford Algebras 23, 441–451 (2013)
Smith, T.: Decomposition of generalized Clifford algebras. Q. J. Math. 42, 105–112 (1991)
Song, Y., Lee, D.: A construction of matrix representation of Clifford algebras. Adv. Appl. Clifford Algebras 23, 719–731 (2015)
Song, Y., Lee, D.: Vector generators of the real Clifford algebras. Linear Multilinear Algebra 63, 2448–2460 (2015)
Tian, Y.: Universal factorization equalities for quaternion matrices and their applications. Math. J. Okayama Univ. 41, 45–62 (1999)
Tian, Y.: Similarity and consimilarity of elements in real Cayley–Dickson algebras. Adv. Appl. Clifford Algebras 9, 61–76 (1999)
Trautman, A.: Clifford algebras and their representations. Encycl. Math. Phys. 1, 518–530 (2006)
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Fuad Kittaneh.
The present research has been conducted by the Research Grant of Kwangwoon University in 2017.
Rights and permissions
About this article
Cite this article
Song, Y. Number Extensions and Some Equations in Clifford Algebras. Bull. Malays. Math. Sci. Soc. 43, 1–13 (2020). https://doi.org/10.1007/s40840-018-0660-7
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40840-018-0660-7