Abstract
For \(\eta \in \{{\mathbf {i}},{\mathbf {j}},{\mathbf {k}}\}\), a real quaternion matrix A is said to be \(\eta \)-Hermitian if \(A=A^{\eta *},\) where \(A^{\eta *}=-\eta A^{*}\eta \), and \(A^{*}\) stands for the conjugate transpose of A. In this paper, we present some practical necessary and sufficient conditions for the existence of an \(\eta \)-Hermitian solution to a system of constrained two-sided coupled real quaternion matrix equations and provide the general \(\eta \)-Hermitian solution to the system when it is solvable. Moreover, we present an algorithm and a numerical example to illustrate our results.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The definition of \(\eta \)-Hermitian quaternion matrix was first proposed by Took et al. [38] (see Definition 2.1). Nowadays \(\eta \)-Hermitian quaternion matrices play an important role in widely linear model and statistical signal processing [35,36,37,38].
Since Took et al. [38] first studied the \(\eta \)-Hermitian quaternion matrix in 2011, there have been some papers to discuss the topics related to \(\eta \)-Hermitian quaternion matrix. These papers mainly considered the \(\eta \)-Hermitian solutions to quaternion matrix equations (e.g. [3, 31, 43,44,45,46]) and decompositions of quaternion matrices involving \(\eta \)-Hermicity (e.g. [15, 17]). For instance, He and Wang [13] provided some necessary and sufficient conditions for the existence of a solution (X, Y, Z) to the quaternion matrix equation
where Y and Z are required to be \(\eta \)-Hermitian. They also presented the general solution to equation (1.1) when it is solvable. Liu [29] considered the \(\eta \)-anti-Hermitian solutions to some classic quaternion matrix equations. Zhang and Wang [47] derived the exact solution to a system of quaternion matrix equations involving \(\eta \)-Hermicity. Many authors investigated the solutions to quaternion matrix equations (e.g. [1, 2, 4,5,6,7,8,9,10,11,12, 16, 18,19,20,21,22,23,24,25,26,27,28, 32,33,34, 39,40,41, 45]).
In 2014, He and Wang [14] gave the solvability conditions and general \(\eta \)-Hermitian solution to the system of quaternion matrix equations with two unknowns
However, to our knowledge, there has been little information on the generalization of system (1.2). Motivated by the work mentioned above and keeping the wide application of \(\eta \)-Hermitian quaternion matrices, we consider the \(\eta \)-Hermitian solution to the system of quaternion matrix equations with three unknowns
where \(A_{i},B_{i},C_{i},D_{i},P=P^{\eta *}\) and \(Q=Q^{\eta *}\) are given matrices \((i=1,2,3,4)\).
The remainder of the paper is organized as follows. In Sect. 2, we review the definition and properties of \(\eta \)-Hermitian quaternion matrix. We also discuss some quaternion matrix equations, which will be used in proving our main result. In Sect. 3, we explore the consistency and \(\eta \)-Hermitian solutions to system (1.3). In Sect. 4, we provide an algorithm and a numerical example to illustrate our results. Finally, a brief conclusion is given in Sect. 5.
Throughout this paper, let \({\mathbb {R}}\) and \({\mathbb {H}}^{m\times n}\) stand for the real number field and the set of all \(m\times n\) matrices over the real quaternion algebra
respectively. The symbol \(A^{*}\) stands for the conjugate transpose of matrix A. I and 0 denotes the identity matrix and zero matrix with appropriate sizes, respectively. The rank of \(A\in {\mathbb {H}}^{m\times n}\) is defined as the (quaternion) dimension of
the range of A (Definition 3.2.3 in [32]). If \(A\in {\mathbb {H}}^{m\times n},\) then there exist nonsingular matrices P and Q such that
where the symbol r(A) is the rank of A. The Moore-Penrose inverse \(A^{\dag }\) of a quaternion matrix A, is defined to be the unique matrix \(A^{\dag },\) such that
Furthermore, \(L_{A}\) and \(R_{A}\) stand for the projectors \(L_{A} =I-A^{\dag }A\) and \(R_{A}=I-AA^{\dag }\) induced by A, respectively.
2 Preliminaries
In this section, we first review the definition of \(\eta \)-Hermitian quaternion matrix.
Definition 2.1
(\(\eta \)-Hermitian quaternion matrix) [38] For \(\eta \in \{{\mathbf {i}},{\mathbf {j}},{\mathbf {k}}\}\), a square real quaternion matrix A is said to be \(\eta \)-Hermitian if \(A=A^{\eta *},\) where \(A^{\eta *}=-\eta A^{*}\eta \).
Some basic properties of \(\eta \)-Hermitian matrix are given in the following lemma.
Lemma 2.1
[13] Let \(A\in {\mathbb {H}}^{m\times n}\) be given. Then,
-
(1)
\((A^{\eta *})^{\dag }=(A^{\dag })^{\eta *}\).
-
(2)
\(r(A)=r(A^{\eta *})\).
-
(3)
\((L_{A})^{\eta *}=R_{A^{\eta *}}\).
-
(4)
\((R_{A})^{\eta *}=L_{A^{\eta *}}\).
Now, we discuss some quaternion matrix equations, which will be used for proving the main result of the paper.
Lemma 2.2
[14] The system of quaternion matrix equations
has an \(\eta \)-Hermitian solution if and only if \(R_{E_{1}}F_{1}=0\) and \(E_{1}F_{1}^{\eta *}=F_{1}E_{1}^{\eta *}\), where
In this case, the general \(\eta \)-Hermitian solution to (2.1) can be expressed as
where W is an arbitrary \(\eta \)-Hermitian matrix over \({\mathbb {H}}\) with appropriate size.
Lemma 2.3
[13] Consider the quaternion matrix equation (1.1). Set
Then, the following statements are equivalent:
-
(1)
Equation (1.1) has a solution (X, Y, Z), where Y and Z are \(\eta \)-Hermitian.
-
(2)
$$\begin{aligned} R_{M_{1}}R_{A}C=0,\qquad R_{A}CL_{B^{\eta *}}=0. \end{aligned}$$
-
(3)
$$\begin{aligned} r\begin{bmatrix}D_{1} &{} C_{1} &{} B_{1} &{} A_{1}\\ A_{1}^{\eta *} &{} 0 &{} 0 &{} 0\end{bmatrix}=r[C_{1},~B_{1},~A_{1}]+r(A_{1}), \end{aligned}$$$$\begin{aligned} r\begin{bmatrix}D_{1} &{} B_{1} &{} A_{1}\\ A_{1}^{\eta *} &{} 0 &{} 0\\ C_{1}^{\eta *} &{} 0 &{} 0\end{bmatrix}=r[B_{1},~A_{1}]+r[A_{1},~C_{1}]. \end{aligned}$$
If matrix equation (1.1) is consistent, then the general solution to (1.1) is
$$\begin{aligned} X=&\,A_{1}^{\dag }(D_{1}-B_{1}YB_{1}^{\eta *}-C_{1}ZC_{1}^{\eta *}) -\frac{1}{2}A_{1}^{\dag }(D_{1}-B_{1}YB_{1}^{\eta *} -C_{1}ZC_{1}^{\eta *})(A_{1}^{\dag })^{\eta *}A_{1}^{\eta *}\\&+L_{A_{1}}U_{1}+U_{2}^{\eta *}(A_{1}^{\dag })^{\eta *}A_{1}^{\eta *} -A_{1}^{\dag }U_{2}A_{1}^{\eta *},\\ Y=&\,Y^{\eta *}=A^{\dag }C(A^{\dag })^{\eta *}-\frac{1}{2}A^{\dag }BM_{1}^{\dag }C[I+(B^{\dag })^{\eta *}S_{1}^{\eta *}](A^{\dag })^{\eta *}\\&-\frac{1}{2}A^{\dag }(I+S_{1}B^{\dag })C(M_{1}^{\dag })^{\eta *} B^{\eta *}(A^{\dag })^{\eta *}\\&-A^{\dag }S_{1}VS_{1}^{\eta *}(A^{\dag })^{\eta *}+L_{A}U_{3}+(L_{A}U_{3})^{\eta *},\\ Z=&\,Z^{\eta *}=\frac{1}{2}M_{1}^{\dag }C(B^{\dag })^{\eta *}[I+(S_{1}^{\dag }S_{1})^{\eta *}]+\frac{1}{2}(I+S_{1}^{\dag }S_{1})B^{\dag }C(M_{1}^{\dag })^{\eta *} +L_{M_{1}}V(L_{M_{1}})^{\eta *}\\&+L_{B}U_{4}+(L_{B}U_{4})^{\eta *} +L_{M_{1}}L_{S_{1}}U_{5}+(L_{M_{1}}L_{S_{1}}U_{5})^{\eta *}, \end{aligned}$$where \(U_{1},U_{2},U_{3},U_{4},U_{5},V=V^{\eta *}\) are arbitrary matrices over \({\mathbb {H}}\) with appropriate sizes.
The following lemma can be easily generalized to \({\mathbb {H}}\).
Lemma 2.4
[30] Let \(A\in {\mathbb {H}}^{m\times n},B\in {\mathbb {H}}^{m\times k},C\in {\mathbb {H}}^{l\times n}\) be given. Then,
-
(1)
\(r(A)+r(R_{A}B)=r(B)+r(R_{B}A)=r [ A,~ B ]\).
-
(2)
\(r(A)+r(CL_{A})=r(C)+r(AL_{C})=r \begin{bmatrix} A\\ C \end{bmatrix}\).
3 The General \(\eta \)-Hermitian Solution to System (1.3)
In this section, we derive some practical necessary and sufficient conditions for the existence of an \(\eta \)-Hermitian solution (X, Y, Z) to the system of quaternion matrix equations
Moreover, we also give the expression of the general \(\eta \)-Hermitian solution to system (3.1) when it is solvable. For simplicity, we denote
Next, we establish the main result of the paper.
Theorem 3.1
Let \(A_{i},B_{i},C_{i},D_{i},P=P^{\eta *}\) and \(Q=Q^{\eta *}\) be given over \({\mathbb {H}}\), \((\eta \in \{{\mathbf {i}},{\mathbf {j}},{\mathbf {k}}\},i=1,\ldots ,4)\). Then, the following statements are equivalent:
-
(1)
System (3.1) has an \(\eta \)-Hermitian solution (X, Y, Z).
-
(2)
$$\begin{aligned} R_{E_{j}}F_{j}=&\,0,~E_{j}F_{j}^{\eta *}=F_{j}E_{j}^{\eta *},~(j=1,2,3), \end{aligned}$$(3.10)$$\begin{aligned} R_{M_{1}}R_{A_{5}}P_{1}=&\,0,~R_{A_{5}}P_{1}(R_{C_{5}})^{\eta *}=0,~ R_{M_{2}}R_{B_{5}}Q_{1}=0,~R_{B_{5}}Q_{1}(R_{D_{5}})^{\eta *}=0, \end{aligned}$$(3.11)$$\begin{aligned} R_{M}R_{A}C=&\,0,~R_{A}C(R_{B})^{\eta *}=0. \end{aligned}$$(3.12)
-
(3)
$$\begin{aligned}&r\begin{bmatrix}A_{j}&{}C_{j}\\ B_{1}^{\eta *}&{}D_{1}^{\eta *}\end{bmatrix} =r\begin{bmatrix}A_{j}\\ B_{1}^{\eta *}\end{bmatrix},~ \begin{bmatrix} A_{j}C_{j}^{\eta *}&{}A_{j}D_{j}\\ B_{j}^{\eta *}C_{j}^{\eta *}&{}B_{j}^{\eta *}D_{j} \end{bmatrix}\nonumber \\&\quad =\begin{bmatrix} C_{j}A_{j}^{\eta *}&{}C_{j}B_{j}\\ D_{j}^{\eta *}A_{j}^{\eta *}&{}D_{j}^{\eta *}B_{j} \end{bmatrix}, \quad (j=1,2,3), \end{aligned}$$(3.13)$$\begin{aligned}&r\begin{bmatrix} P&{}A_{4}&{}C_{4}\\ C_{1}A_{4}^{\eta *}&{}A_{1}&{}0\\ D_{1}^{\eta *}A_{4}^{\eta *}&{}B_{1}^{\eta *}&{}0\\ C_{2}C_{4}^{\eta *}&{}0&{}A_{2}\\ D_{2}^{\eta *}C_{4}^{\eta *}&{}0&{}B_{2}^{\eta *}\end{bmatrix}= r\begin{bmatrix} A_{4}&{}C_{4}\\ A_{1}&{}0\\ B_{1}^{\eta *}&{}0\\ 0&{}A_{2}\\ 0&{}B_{2}^{\eta *}\end{bmatrix}, \end{aligned}$$(3.14)$$\begin{aligned}&r\begin{bmatrix} P&{}A_{4}&{}C_{4}C_{2}^{\eta *}&{}C_{4}D_{2}\\ C_{4}^{\eta *}&{}0&{}A_{2}^{\eta *}&{}B_{2}\\ C_{1}A_{4}^{\eta *}&{}A_{1}&{}0&{}0\\ D_{1}^{\eta *}A_{4}^{\eta *}&{}B_{1}^{\eta *}&{}0&{}0\end{bmatrix}= r\begin{bmatrix} A_{1}\\ B_{1}^{\eta *}\\ A_{4}\end{bmatrix}+r\begin{bmatrix} A_{2}\\ B_{2}^{\eta *}\\ C_{4}\end{bmatrix}, \end{aligned}$$(3.15)$$\begin{aligned}&r\begin{bmatrix} Q&{}B_{4}&{}D_{4}\\ C_{3}B_{4}^{\eta *}&{}A_{3}&{}0\\ D_{3}^{\eta *}B_{4}^{\eta *}&{}B_{3}^{\eta *}&{}0\\ C_{2}D_{4}^{\eta *}&{}0&{}A_{2}\\ D_{2}^{\eta *}D_{4}^{\eta *}&{}0&{}B_{2}^{\eta *}\end{bmatrix}= r\begin{bmatrix} B_{4}&{}D_{4}\\ A_{3}&{}0\\ B_{3}^{\eta *}&{}0\\ 0&{}A_{2}\\ 0&{}B_{2}^{\eta *}\end{bmatrix}, \end{aligned}$$(3.16)$$\begin{aligned}&r\begin{bmatrix} Q&{}B_{4}&{}D_{4}C_{2}^{\eta *}&{}D_{4}D_{2}\\ D_{4}^{\eta *}&{}0&{}A_{2}^{\eta *}&{}B_{2}\\ C_{3}B_{4}^{\eta *}&{}A_{3}&{}0&{}0\\ D_{3}^{\eta *}B_{4}^{\eta *}&{}B_{3}^{\eta *}&{}0&{}0\end{bmatrix}= r\begin{bmatrix} A_{3}\\ B_{3}^{\eta *}\\ B_{4}\end{bmatrix}+r\begin{bmatrix} A_{2}\\ B_{2}^{\eta *}\\ D_{4}\end{bmatrix}, \end{aligned}$$(3.17)$$\begin{aligned}&r\begin{bmatrix} 0&{}-C_{4}^{\eta *}&{}D_{4}^{\eta *}&{}0&{}0&{}A_{2}^{\eta *}&{}B_{2}\\ C_{4}&{}P&{}0&{}A_{4}&{}0&{}0&{}0\\ D_{4}&{}0&{}Q&{}0&{}B_{4}&{}D_{4}C_{2}^{\eta *}&{}D_{4}D_{2}\\ A_{2}&{}C_{2}C_{4}^{\eta *}&{}0&{}0&{}0&{}0&{}0\\ B_{2}^{\eta *}&{}D_{2}^{\eta *}C_{4}^{\eta *}&{}0&{}0&{}0&{}0&{}0\\ 0&{}C_{1}A_{4}^{\eta *}&{}0&{}A_{1}&{}0&{}0&{}0\\ 0&{}D_{1}^{\eta *}A_{4}^{\eta *}&{}0&{}B_{1}^{\eta *}&{}0&{}0&{}0\\ 0&{}0&{}C_{3}B_{4}^{\eta *}&{}0&{}A_{3}&{}0&{}0\\ 0&{}0&{}D_{3}^{\eta *}B_{4}^{\eta *}&{}0&{}B_{3}^{\eta *}&{}0&{}0 \end{bmatrix}=r\begin{bmatrix}C_{4}\\ D_{4}\\ A_{2}\\ B_{2}^{\eta *} \end{bmatrix}\nonumber \\&\qquad + r\begin{bmatrix}C_{4}&{}A_{4}&{}0\\ D_{4}&{}0&{}B_{4}\\ A_{2}&{}0&{}0\\ B_{2}^{\eta *}&{}0&{}0\\ 0&{}A_{1}&{}0\\ 0&{}B_{1}^{\eta *}&{}0\\ 0&{}0&{}A_{3}\\ 0&{}0&{}B_{3}^{\eta *}\end{bmatrix}, \end{aligned}$$(3.18)$$\begin{aligned}&r\begin{bmatrix} 0&{}-C_{4}^{\eta *}&{}D_{4}^{\eta *}&{}0&{}A_{2}^{\eta *}&{}B_{2}&{}0&{}0\\ C_{4}&{}P&{}0&{}A_{4}&{}0&{}0&{}0&{}0\\ D_{4}&{}0&{}Q&{}0&{}D_{4}C_{2}^{\eta *}&{}D_{4}D_{2}&{}B_{4}C_{3}^{\eta *}&{}B_{4}D_{3}\\ A_{2}&{}C_{2}C_{4}^{\eta *}&{}0&{}0&{}0&{}0&{}0&{}0\\ B_{2}^{\eta *}&{}D_{2}^{\eta *}C_{4}^{\eta *}&{}0&{}0&{}0&{}0&{}0&{}0\\ 0&{}C_{1}A_{4}^{\eta *}&{}0&{}A_{1}&{}0&{}0&{}0&{}0\\ 0&{}D_{1}^{\eta *}A_{4}^{\eta *}&{}0&{}B_{1}^{\eta *}&{}0&{}0&{}0&{}0\\ 0&{}0&{}B_{4}^{\eta *}&{}0&{}0&{}0&{}A_{3}^{\eta *}&{}B_{3} \end{bmatrix}= r\begin{bmatrix}C_{4}&{}A_{4}\\ D_{4}&{}0\\ A_{2}&{}0\\ B_{2}^{\eta *}&{}0\\ 0&{}A_{1}\\ 0&{}B_{1}^{\eta *}\end{bmatrix}\nonumber \\&\qquad + r\begin{bmatrix} C_{4}&{}0\\ D_{4}&{}B_{4}\\ A_{2}&{}0\\ B_{2}^{\eta *}&{}0\\ 0&{}A_{3}\\ 0&{}B_{3}^{\eta *}\end{bmatrix}. \end{aligned}$$(3.19)
If system (3.1) is consistent, the general \(\eta \)-Hermitian solution to (3.1) can be expressed as
where
or
and \(U_{1},U_{4},W_{1},W_{2},W_{3}=W_{3}^{\eta *},W_{4},W_{5},W_{6}\) are arbitrary matrices over \({\mathbb {H}}\) with appropriate sizes, m is the column number of \(A_{2}\).
Proof
\((1)\Longleftrightarrow (2):\) It follows from Lemma 2.2 that the systems
and
have an \(\eta \)-Hermitian solution, respectively, if and only if
where \(E_{j}\) and \(F_{j}\) are defined in (3.2). In this case, the general \(\eta \)-Hermitian solutions to systems (3.23)–(3.25) can be expressed as
where \(X_{1},Y_{1}\) and \(Z_{1}\) are arbitrary \(\eta \)-Hermitian matrices over \({\mathbb {H}}\) with appropriate sizes. Substituting (3.26)–(3.28) into
yields
where \(A_{5},B_{5},C_{5},D_{5},P_{1}\) and \(Q_{1}\) are defined in (3.3)–(3.5).
Now, we want to find the solvability conditions and general \(\eta \)-Hermitian solution to system (3.29). From Lemma 2.3, it follows that the first equation in system (3.29) has an \(\eta \)-Hermitian solution if and only if
where \(M_{1}\) is defined in (3.6). The general \(\eta \)-Hermitian solution to the first equation in system (3.29) can be expressed as
where \(S_{1}\) is defined in (3.6), \(U_{1},U_{2},U_{3}\) and \(V_{1}=V_{1}^{\eta *}\) are arbitrary matrices over \({\mathbb {H}}\) with appropriate size.
Similarly, the second equation in system (3.29) has an \(\eta \)-Hermitian solution if and only if
where \(M_{2}\) is defined in (3.6). The general \(\eta \)-Hermitian solution to the second equation in system (3.29) can be expressed as
where \(S_{2}\) is defined in (3.6), \(U_{4},U_{5},U_{6}\) and \(V_{2}=V_{2}^{\eta *}\) are arbitrary matrices over \({\mathbb {H}}\) with appropriate sizes.
Equating (3.30) and (3.31), we have
Then, it can be rewritten as
where \(V_{1}\) and \(V_{2}\) are \(\eta \)-Hermitian, \(A_{6}\) and \(P_{2}\) are defined in (3.7) and (3.8), separately. We want to find the general solution to equation (3.32), where \(V_{1}\) and \(V_{2}\) are \(\eta \)-Hermitian. By Lemma 2.3, we know that equation (3.32) has a solution if and only if
where A, B, C, M are defined in (3.9). In this case, the general solution to equation (3.32) is given as follows:
where \(W_{1},W_{2},W_{3}=W_{3}^{\eta *},W_{4},W_{5},W_{6}\) are arbitrary matrices over \({\mathbb {H}}\) with appropriate sizes.
\((2)\Longleftrightarrow (3):\) By Lemma 2.4 and block Gaussian elimination, we obtain
i.e. (3.10) \(\Longleftrightarrow \) (3.13). It follows from Lemma 2.2 that systems (3.23)–(3.25) have an \(\eta \)-Hermitian solution, respectively, under the conditions (3.10). Note that
and
are special \(\eta \)-Hermitian solutions to systems (3.23)–(3.25). Then, \(P_{1}\) and \(Q_{1}\) can be rewritten as follows:
Now, we will use Lemma 2.4, (3.33), (3.34) and block Gaussian elimination to prove (3.11) \(\Longleftrightarrow \) (3.14)–(3.17). Applying Lemma 2.4 to \(R_{M_{1}}R_{A_{5}}P_{1}=0\) yields
Similarly, it can be shown that
Next, we prove that (3.12) \(\Longleftrightarrow \) (3.18) and (3.19). Note that
and
are special solutions to the first two equations in system (3.29), separately, under the conditions (3.11). Then, \(P_{2}\) can be rewritten as follows
We next use Lemma 2.4, together with (3.29), (3.33), (3.34), (3.35) and block Gaussian elimination to show \(R_{M}R_{A}C=0\) \(\Longleftrightarrow \) (3.18). Applying Lemma 2.4 to \(R_{M}R_{A}C=0\) gives
Similarly, we can prove that \(R_{A}C(R_{B})^{\eta *}=0\) \(\Longleftrightarrow \) (3.19). \(\square \)
As a special case of system (3.1), we obtain a result about the system of quaternion matrix equations
which we state as follows:
Corollary 3.2
Let \(A_{i},B_{i}\) and \(E_{i}=E_{i}^{\eta *}\) be given over \({\mathbb {H}}\), \((\eta \in \{{\mathbf {i}},{\mathbf {j}},{\mathbf {k}}\},i=1,2)\) and
Then, the following statements are equivalent:
-
(1)
System (3.36) has an \(\eta \)-Hermitian solution (X, Y, Z).
-
(2)
$$\begin{aligned} R_{M_{i}}R_{A_{i}}E_{i}=0, ~ R_{A_{i}}E_{i}L_{D_{i}}=0, ~ R_{M}R_{A}C=0,~R_{A}C(R_{B})^{\eta *}=0. \end{aligned}$$
-
(3)
$$\begin{aligned} r[E_{i},~A_{i},~B_{i}] =&\,r[A_{i},~B_{i}], ~ r\begin{bmatrix} E_{i} &{} A_{i}\\ B_{i}^{\eta *} &{} 0 \end{bmatrix} =r(A_{i})+r(B_{i}) ,\\ r\begin{bmatrix} 0&{}B_{1}^{\eta *}&{}B_{2}^{\eta *}&{}0&{}0\\ B_{1}&{}E_{1}&{}0&{}A_{1}&{}0\\ B_{2}&{}0&{}-E_{2}&{}0&{}A_{2} \end{bmatrix} =&\,r \begin{bmatrix} B_{1} &{} A_{1}&{}0\\ B_{2}&{}0&{}A_{2} \end{bmatrix}+r\begin{bmatrix}B_{1}\\ B_{2} \end{bmatrix},\\ r\begin{bmatrix} 0&{}B_{1}^{\eta *}&{}B_{2}^{\eta *}&{}0\\ B_{1}&{}-E_{1}&{}0&{}A_{1}\\ B_{2}&{}0&{}E_{2}&{}0\\ 0&{}0&{}A_{2}^{\eta *}&{}0 \end{bmatrix} =&\,r \begin{bmatrix} B_{1} &{} A_{1}\\ B_{2}&{}0 \end{bmatrix}+r\begin{bmatrix}B_{1}&{}0\\ B_{2}&{}A_{2}\end{bmatrix}. \end{aligned}$$
In this case, the general \(\eta \)-Hermitian solution to system (3.36) can be expressed as
or
where
and \(U_{1},U_{4},W_{1},W_{2},W_{3}=W_{3}^{\eta *},W_{4},W_{5},W_{6}\) are arbitrary matrices over \({\mathbb {H}}\) with appropriate sizes, m is the column number of \(B_{1}\).
4 An Algorithm and an Example
In this section, we give an algorithm and an example to illustrate the main result. The algorithm for finding the \(\eta \)-Hermitian solution to the system of real quaternion matrix equations (3.1) is presented as follows.
Algorithm 4.1
The \(\eta \)-Hermitian solution to the system of quaternion matrix equations (3.1).
Input: The system of quaternion matrix equations (3.1).
-
(i)
Compute \(E_{j},F_{j},(j=1,2,3)\) by (3.2).
-
(ii)
Compute \(A_{5},B_{5},C_{5},D_{5},P_{1},Q_{1},M_{1},S_{1},M_{2},S_{2},A_{6},P_{2},A,B,C,M,S\) by (3.3)–(3.9).
-
(iii)
Compute the Moore–Penrose inverses conditions in Theorem 3.1 (2).
-
(iv)
Compute the rank conditions in Theorem 3.1 (3).
-
(iv)
If any of equations in Theorem 3.1 (2) or (3) fails, return
“No \(\eta \)-Hermitian solution”.
-
(v)
Else compute the Moore–Penrose inverses in Theorem 3.1 (3.20)–(3.22).
Output: \(\eta \)-Hermitian solution (X, Y, Z) in Theorem 3.1.
Here is an example to illustrate Theorem 3.1.
Example 1
Given the matrices:
Now, we consider the \({\mathbf {j}}\)-Hermitian solution to system (3.1), where \(\eta ={\mathbf {j}}.\) Note that the rank of a quaternion matrix A can be quickly calculated by using the following property (see [42])
where \(A=A_{1}+A_{2}{\mathbf {j}},\) \(A_{1}\) and \(A_{2}\) are complex matrices, \(\overline{A_{1}}\) means the conjugate of the matrix \(A_{1}\). The Moore–Penrose inverse of a quaternion matrix A can be calculated in MATLAB by using SVD, i.e. \(A^{\dag }=V^{*}\begin{bmatrix}D_{r}^{-1}&{}0\\ 0&{}0\end{bmatrix}U^{*},\) where U and V are unitary quaternion matrices, \(D_{r}=\text{ diag }\{d_{1},\ldots ,d_{r}\}\) and the \(d^{,}\)s are the positive singular values of A. Note that
All equalities in (3.13)–(3.19) hold. By Theorem 3.1, we can conclude that system (3.1) has a \({\mathbf {j}}\)-Hermitian solution (X, Y, Z), where
5 Conclusions
We have derived some practical necessary and sufficient conditions for the existence of an \(\eta \)-Hermitian solution (X, Y, Z) to the system of constrained two-sided coupled real quaternion matrix equations (1.3). We have also given the general \(\eta \)-Hermitian solution to system (1.3) when the solvability conditions are satisfied. An algorithm and a numerical example have been provided to illustrate the main result.
References
Futorny, V., Klymchuk, T., Sergeichuk, V.V.: Roth’s solvability criteria for the matrix equations \(AX-{\widehat{X}}B=C\) and \(X-A{\widehat{X}}B=C\) over the skew field of quaternions with an involutive automorphism \(q\rightarrow {\hat{q}}\). Linear Algebra Appl. 510, 246–258 (2016)
He, Z.H.: Structure, properties and applications of some simultaneous decompositions for quaternion matrices involving \(\phi \)-skew-Hermicity. Adv. Appl. Clifford Algebras 29, 6 (2019)
He, Z.H.: The general solution to a system of coupled Sylvester-type quaternion tensor equations involving \(\eta \)-Hermicity. Bull. Iran. Math. Soc. 45, 1407–1430 (2019)
He, Z.H.: Pure PSVD approach to Sylvester-type quaternion matrix equations. Electron. J. Linear Algebra 35, 266–284 (2019)
He, Z.H.: A system of coupled quaternion matrix equations with seven unknowns and its applications. Adv. Appl. Clifford Algebras 29, Article ID 38 (2019)
He, Z.H.: Some quaternion matrix equations involving \(\phi \)-Hermicity. Filomat 33, 5097–5112 (2019)
He, Z.H.: Some new results on a system of Sylvester-type quaternion matrix equations. Linear Multilinear Algebra (2019). https://doi.org/10.1080/03081087.2019.1704213
He, Z.H., Liu, J., Tam, T.Y.: The general \(\phi \)-Hermitian solution to mixed pairs of quaternion matrix Sylvester equations. Electron. J. Linear Algebra 32, 475–499 (2017)
He, Z.H., Wang, Q.W.: A system of periodic discrete-time coupled Sylvester quaternion matrix equations. Algebra Colloq. 24, 169–180 (2017)
He, Z.H., Wang, Q.W.: The general solutions to some systems of matrix equations. Linear Multilinear Algebra 63, 2017–2032 (2015)
He, Z.H., Wang, Q.W., Zhang, Y.: The complete equivalence canonical form of four matrices over an arbitrary division ring. Linear Multilinear Algebra 66, 74–95 (2018)
He, Z.H., Wang, Q.W., Zhang, Y.: A simultaneous decomposition for seven matrices with applications. J. Comput. Appl. Math. 349, 93–113 (2019)
He, Z.H., Wang, Q.W.: A real quaternion matrix equation with applications. Linear Multilinear Algebra 61, 725–740 (2013)
He, Z.H., Wang, Q.W.: The \(\eta \)-bihermitian solution to a system of real quaternion matrix equations. Linear Multilinear Algebra 62, 1509–1528 (2014)
He, Z.H., Wang, Q.W., Zhang, Y.: Simultaneous decomposition of quaternion matrices involving \(\eta \)-Hermicity with applications. Appl. Math. Comput. 298, 13–35 (2017)
He, Z.H., Wang, Q.W., Zhang, Y.: A system of quaternary coupled Sylvester-type real quaternion matrix equations. Automatica 87, 25–31 (2018)
Horn, R.A., Zhang, F.Z.: A generalization of the complex Autonne–Takagi factorization to quaternion matrices. Linear Multilinear Algebra 60(11–12), 1239–1244 (2012)
Kyrchei, I.: Determinantal representations of solutions to systems of quaternion matrix equations. Adv. Appl. Clifford Algebras 28, 23 (2018)
Kyrchei, I.: Determinantal representations of general and (skew-)Hermitian solutions to the generalized Sylvester-Type quaternion matrix equation. Abstr. Appl. Anal. 2019, Article ID 5926832
Kyrchei, I.: Determinantal representations of solutions and Hermitian solutions to some system of two-sided quaternion matrix equations. J. Math. 2018, Article ID 6294672
Kyrchei, I.: Cramer’s rules for Sylvester quaternion matrix equation and its special cases. Adv. Appl. Clifford Algebras 28, 90 (2018)
Kyrchei, I.: Cramer’s Rules of \(\eta \)-(skew-) Hermitian solutions to the quaternion Sylvester-type matrix equations. Adv. Appl. Clifford Algebras 29, 56 (2019)
Kyrchei, I.: Explicit representation formulas for the minimum norm least squares solutions of some quaternion matrix equations. Linear Algebra Appl. 438, 136–152 (2018)
Kyrchei, I.: Determinantal representations of the Drazin inverse over the quaternion skew field with applications to some matrix equations. Appl. Math. Comput. 238, 193–207 (2014)
Kyrchei, I.: Explicit determinantal representation formulas of W-weighted Drazin inverse solutions of some matrix equations over the quaternion skew field. Math. Probl. Eng. Article ID 8673809 (2016)
Kyrchei, I.: Cramer’s rules for the system of quaternion matrix equations \(\eta \)-Hermicity. 4open 2, 24 (2019)
Liu, X., He, Z.H.: On the split quaternion matrix equation \(AX=B\). Banach J. Math. Anal. 14, 228–248 (2020)
Liu, X., Huang, H., He, Z.H.: Real representation approach to quaternion matrix equation involving \(\phi \)-Hermicity. Math. Probl. Eng. Article ID 3258349 (2019)
Liu, X.: The \(\eta \)-anti-Hermitian solution to some classic matrix equations. Appl. Math. Comput. 320, 264–270 (2018)
Marsaglia, G., Styan, G.P.H.: Equalities and inequalities for ranks of matrices. Linear Multilinear Algebra 2, 269–292 (1974)
Rehman, A., Wang, Q.W., He, Z.H.: Solution to a system of real quaternion matrix equations encompassing \(\eta \)-Hermicity. Appl. Math. Comput. 265, 945–957 (2015)
Rodman, L.: Topics in Quaternion Linear Algebra. Princeton University Press, Princeton (2014)
Song, C., Chen, G., Zhang, X.: An iterative solution to coupled quaternion matrix equations. Filomat 26(4), 809–826 (2012)
Song, C., Chen, G.: On solutions of matrix equations \(XF-AX=C\) and \(XF-A{\widetilde{X}}=C\) over quaternion field. J. Appl. Math. Comput. 37, 57–88 (2011)
Took, C.C., Mandic, D.P.: Augmented second-order statistics of quaternion random signals. Signal Process. 91, 214–224 (2011)
Took, C.C., Mandic, D.P.: The quaternion LMS algorithm for adaptive filtering of hypercomplex real world processes. IEEE Trans. Signal Process. 57, 1316–1327 (2009)
Took, C.C., Mandic, D.P.: Quaternion-valued stochastic gradient-based adaptive IIR filtering. IEEE Trans. Signal Process. 58(7), 3895–3901 (2010)
Took, C.C., Mandic, D.P., Zhang, F.Z.: On the unitary diagonalization of a special class of quaternion matrices. Appl. Math. Lett. 24, 1806–1809 (2011)
Wang, Q.W.: Bisymmetric and centrosymmetric solutions to system of real quaternion matrix equations. Comput. Math. Appl. 49, 641–650 (2005)
Wang, Q.W., He, Z.H., Zhang, Y.: Constrained two-sided coupled Sylvester-type quaternion matrix equations. Automatica 101, 207–213 (2019)
Wang, Q.W., van der Woude, J.W., Chang, H.X.: A system of real quaternion matrix equations with applications. Linear Algebra Appl. 431, 2291–2303 (2009)
Wolf, L.A.: Similarity of matrices in which the elements are real quaternions. Bull. Am. Math. Soc. 42, 737–743 (1936)
Yuan, S.F., Wang, Q.W.: Two special kinds of least squares solutions for the quaternion matrix equation \(AXB+CXD=E\). Electron. J. Linear Algebra 23, 257–274 (2012)
Yuan, S.F., Wang, Q.W., Xiong, Z.P.: The least squares \(\eta \)-Hermitian problems of quaternion matrix equation \(A^{H}XA+ B^{H}YB= C\). Filomat 28, 1153–1165 (2014)
Zhang, F., Wei, M., Li, Y., Zhao, J.: Special least squares solutions of the quaternion matrix equation \(AXB+CXD=E\). Comput. Math. Appl. 72, 1426–1435 (2016)
Zhang, X.: A system of generalized Sylvester quaternion matrix equations and its applications. Appl. Math. Comput. 273, 74–81 (2016)
Zhang, Y., Wang, R.H.: The exact solution of a system of quaternion matrix equations involving \(\eta \)-Hermicity. Appl. Math. Comput. 222, 201–209 (2013)
Acknowledgements
The authors thank the anonymous referees for their valuable suggestions that improved the exposition of this article.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Fuad Kittaneh.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research was supported by the National Natural Science Foundation of China (Grant Nos. 11801354 and 11971294), the Science and Technology Development Fund, Macau SAR (Grant No. 185/2017/A3), and National Natural Science Foundation for the Youth of China (Grant No. 11701598).
Rights and permissions
About this article
Cite this article
Liu, X., He, ZH. \(\eta \)-Hermitian Solution to a System of Quaternion Matrix Equations. Bull. Malays. Math. Sci. Soc. 43, 4007–4027 (2020). https://doi.org/10.1007/s40840-020-00907-w
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40840-020-00907-w