Abstract
Our main result is that we describe the solutions \(g,f:S\rightarrow \mathbb {C}\) of the functional equation
where S is a semigroup, \(\alpha \in \mathbb {C}\) is a fixed constant and \(\sigma :S\rightarrow S\) an involutive automorphism.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The cosine addition formula, cosine subtraction formula, and sine addition formula on any semigroup S for unknown functions \(g,f:S\rightarrow \mathbb {C}\) are, respectively, the functional equations
for all \(x,y\in S\), where \(\sigma :S\rightarrow S\) is an involutive automorphism. That is \(\sigma (xy)=\sigma (x)\sigma (y)\) and \(\sigma (\sigma (x))=x\) for all \(x,y\in S\). These functional equations have been investigated by many authors. In Vincze [11] obtained the solutions of (1.2) on abelian groups with \(\sigma =id\), and it was solved on general groups by Chung et al. [3] with \(\sigma =id\). The results were extended to the case of topological groups by Poulsen and Stetkær [7] and to semigroups generated by their squares by Ajebbar and Elqorachi [1], both papers with \(\sigma \) an involutive automorphism. Also Stetkær [10, Theorem 6.1] gave a description of the solution of (1.1) with \(\sigma =id\) on a general semigroup in terms of the solutions of (1.3) with \(\sigma =id\). The most recent result about (1.3) with \(\sigma =id\) was obtained by Ebanks [5, Theorem 2.1] and [6, Theorem 3.1] on semigroups. Ebanks [5, Theorem 4.1] gives the solution of (1.2) on monoids. Recently the authors [2] solved (1.1), (1.2) and (1.3) on semigroups.
Stetkær [10, Theorem 3.1] solved the functional equation
on a semigroup S, where \(\alpha \in \mathbb {C}\) is a fixed constant. He expressed the solutions in terms of multiplitive functions on S and solutions \(h:S\rightarrow \mathbb {C}\) of the special case of the sine addition law
in which \(\chi :S\rightarrow \mathbb {C}\) is a multiplicative function. Equation (1.4) generalizes both the cosine addition formula (1.1) and the cosine subtraction formula (1.2) with \(\sigma =id\).
As a continuation and a generalization of these investigations we determine the complex valued solutions of the functional equation
on semigroups, where \(\sigma :S\rightarrow S\) is an involutive automorphism. We obtain explicit formulas for the solutions expressed in terms of multiplicative, additive and sometimes arbitrary functions. The paper concludes with some examples.
Our notation is decribed in the following set up.
2 Set up and notation
Throughout this paper S denotes a semigroup and \(\sigma :S\rightarrow S\) an involutive automorphism.
A function \(a: S\rightarrow \mathbb {C}\) is additive if
A function \(\chi : S\rightarrow \mathbb {C}\) is multiplicative if
A function \(f: S\rightarrow \mathbb {C}\) is central if \(f(xy) = f(yx)\) for all \(x, y\in S\), and f is abelian if f is central and \(f(xyz)=f(xzy)\) for all \(x,y,z\in S\).
For any subset \(T\subseteq S\) we define the set \(T^2:=\lbrace xy\ \vert \ x, y\in T\rbrace \), so \(T^2\) consists of all products of two (or more) elements of T.
The following remark follows easily from Stetkær [8, Corollary 3.19] and will be used throughout the paper without explicit mentioning.
Remark 2.1
If \(\chi _1\) and \(\chi _2\) are multiplicative functions on S such that \(\chi _1=-\chi _2\), then \(\chi _1=\chi _2=0\).
If \(\chi : S \rightarrow \mathbb {C}\) is a non-zero multiplicative function, we define the sets
For any function \(f: S \rightarrow \mathbb {C}\) we define the function
We call \(f^{e}:=\frac{f+f^{*}}{2}\) the even part of f and \(f^{\circ }:=\frac{f-f^{*}}{2}\) its odd part. The function f is said to be even if \(f=f^{*}\), and f is said to be odd if \(f=-f^{*}\). In the following lemma we give some properties of the set \(P_\chi \).
Lemma 2.2
-
(a)
If \(u\in S\backslash I_\chi \) and \(p\in P_\chi \), then \(up,pu\in P_\chi \).
-
(b)
\(\sigma (P_\chi )=P_{\chi ^*}\). Note in particular that \(\sigma (P_\chi )=P_\chi \), if \(\chi =\chi ^*\).
Proof
-
(a)
follows directly from the definition of \(P_\chi \).
-
(b)
is easy to verify using that \(\sigma :S\rightarrow S\) is a bijection (See [4, Lemma 4.1]).
\(\square \)
For a topological semigroup S let C(S) denote the algebra of continuous functions from S into \(\mathbb {C}\).
3 The main result
The following lemmas will be used later.
Lemma 3.1
Let \(f,g:S\rightarrow \mathbb {C}\) be a solution of the functional equation
where \(g=0\) on \(S^2\) and \(\beta \in \mathbb {C}\backslash \lbrace 0\rbrace \) is a constant. Then f and g are linearly dependent.
Proof
If \(f=0\) or \(g=0\) then f and g are linearly dependent, so we may assume that \(f\ne 0\) and \(g\ne 0\). By applying (3.1) to the pair (xy, z) and taking into account that \(g=0\) on \(S^2\) we obtain
Now if we apply (3.1) to the pair \((x,\sigma (y)z)\) and take into account that \(g=0\) on \(S^2\) we get
So we deduce from (3.2) and (3.3) since \(\beta \ne 0\) that
Then we get since \(f\ne 0\) that \(f(xy)=f(x)\phi (\sigma (y))\) for some function \(\phi :S\rightarrow \mathbb {C}\). Substituting this into (3.1) we find that
This implies that
Choosing \(y_0\in S\) such that \(g(y_0)\ne 0\) we see that f and g are linearly dependent. This completes the proof of Lemma 3.1. \(\square \)
Lemma 3.2
Let \(g,f:S\rightarrow \mathbb {C}\) be two functions such that
where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions, \(g\ne 0\) and \(a_1,a_2\in \mathbb {C}\backslash \lbrace 0\rbrace \), \(b_1,b_2\in \mathbb {C}\) are constants. We have
-
(1)
If f is even and g is odd, then \(a_1=a_2\) and \(b_1+b_2=0\).
-
(2)
If f is odd and g is even, then \(a_1+a_2=0\) and \(b_1=b_2\).
Proof
(1) If f is even and g is odd, we get that
and
By the help of [10, Proposition A.2] we deduce from (3.4) that \(a_1\chi _1=a_1\chi _1^*\) or \(a_1\chi _1=a_2\chi _2^*\). If \(a_1\chi _1=a_1\chi _1^*\) then \(\chi _1=\chi _1^*\) since \(a_1\ne 0\), and then we get from (3.4) that \(\chi _2=\chi _2^*\) since \(a_2\ne 0\). In view of (3.5) we deduce that \(g=0\). This contradicts the fact that \(g\ne 0\). So \(a_1\chi _1=a_2\chi _2^*\) then since \(\chi _1,\chi _2\) are non-zero, \(a_1\ne 0\) and \(a_2\ne 0\), according to [8, Theorem 3.18 (a)] we deduce that \(\chi _1=\chi _2^*\) and then \(a_1=a_2\). Now (3.5) becomes \((b_1+b_2)(\chi _1+\chi _2)=0\) which implies that \(b_1+b_2=0\). This is case (1).
(2) We proceed similarly to case (1) to get the result. This completes the proof of Lemma 3.2. \(\square \)
In the following lemma we give some key properties of the solutions of Eq. (1.6).
Lemma 3.3
Let \(g,f:S\rightarrow \mathbb {C}\) be a solution of the functional Eq. (1.6), and define the function \(G:=g-\alpha f\). The following statements hold:
-
(1)
\(G(x\sigma (y))=G(y\sigma (x))\) for all \(x,y\in S\).
-
(2)
\(G(xyz)=G^*(xyz)\) for all \(x,y,z\in S\).
-
(3)
For all \(x,y,z\in S\)
$$\begin{aligned} g^e(x)g^{\circ }(yz)= & {} f^e(x)f^{\circ }(yz), \end{aligned}$$(3.6)$$\begin{aligned} g^e(yz)g^{\circ }(x)= & {} f^e(yz)f^{\circ }(x). \end{aligned}$$(3.7)
Proof
(1) The functional Eq. (1.6) is equivalent to
The right hand side of (3.8) is invariant under the interchange of x and y. That is
(2) Replacing y by \(\sigma (y)\) in (3.9), we get \(G(xy)=G^*(yx)\) for all \(x,y\in S\), and then we deduce that
(3) By applying the identity (3.8) to the pair \((x,\sigma (yz))\) we obtain
Then by using (2), we deduce from (3.10) that
Since \(k=k^e+k^{\circ }\) and \(k^*=k^e-k^{\circ }\) for any function \(k:S\rightarrow \mathbb {C}\), we get from (3.11) after some rearrangement that for all \(x,y,z\in S\)
In the identity (3.12) the left hand side is an odd function of x while the right is an even function of x, so
and
for all \(x,y,z\in S\). This completes the proof of Lemma 3.3. \(\square \)
Now we present the general solution of (1.6) on semigroups. Stetkær [10, Theorem 3.1] is Theorem 3.4 with \(\sigma =id\), and point (8) of Theorem 3.4 is new compared to [10].
Theorem 3.4
The solutions \(g,f:S\rightarrow \mathbb {C}\) of the functional Eq. (1.6) are the following families:
-
(1)
\(\alpha = \pm 1\), f is any non-zero function and \(g=\alpha f\).
-
(2)
\(\alpha \ne 1\), \(f=g\ne 0\) and \(g=0\) on \(S^2\).
-
(3)
\(\alpha \ne -1\), \(f=-g\ne 0\) and \(g=0\) on \(S^2\).
-
(4)
\(f=(q+\alpha )\dfrac{\chi }{2}\) and \(g=\left( 1\pm \sqrt{1+q^2-\alpha ^2} \right) \dfrac{\chi }{2} \), where \(q\in \mathbb {C}\) is a constant and \(\chi :S\rightarrow \mathbb {C}\) a non-zero even multiplicative function.
-
(5)
\(f=\alpha \dfrac{\chi _1+\chi _2}{2}+q\dfrac{\chi _1-\chi _2}{2}\) and \(g=\dfrac{\chi _1+\chi _2}{2}\pm \sqrt{1+q^2-\alpha ^2}\dfrac{\chi _1-\chi _2}{2}\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero even multiplicative functions and \(q\in \mathbb {C}\backslash \lbrace \pm \alpha \rbrace \) is a constant.
-
(6)
\(\alpha \ne 0\), \(f=\alpha \chi _1\) and \(g=\chi _2\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero even multiplicative functions.
-
(7)
\(f=\alpha \chi +h\) and \(g=\chi \pm h\), where \(\chi :S\rightarrow \mathbb {C}\) is a non-zero even multiplicative function and \(h:S\rightarrow \mathbb {C}\) is a non-zero even solution of the special sine addition law (1.5).
-
(8)
\(\alpha \ne \pm 1\), and
$$\begin{aligned} f=\dfrac{1+\alpha }{2}\chi -\dfrac{1-\alpha }{2}\chi ^*\quad \text {and}\quad g=\dfrac{1+\alpha }{2}\chi +\dfrac{1-\alpha }{2}\chi ^*, \end{aligned}$$where \(\chi :S\rightarrow \mathbb {C}\) is a multiplicative function such that \(\chi \ne \chi ^*\). Note that, off the exceptional case (1), g and f are abelian. Moreover, if S is a topological semigroup, \(f\in C(S)\) is such that \(f\ne \alpha \mu \) for any multiplicative function \(\mu \in C(S)\), then \(g\in C(S)\). In addition \(\chi ,\chi ^*\in C(S)\) in case (8).
Proof
It is easy to check that each of the pairs described in Theorem 3.4 is a solution of (1.6). So assume that the pair (g, f) is a solution of Eq. (1.6). If \(g=0\), then (1.6) can be written as
If \(\alpha =0\) then \(f=0\). This occurs in case (4) with \(q=-\alpha \). Now if \(\alpha \ne 0\), then
Then according to [2, Lemma 4.1], \(f=\alpha \chi \), where \(\chi :S\rightarrow \mathbb {C}\) is an even multiplicative function. This occurs in case (4) with \(q=\alpha \). From now on we assume that \(g\ne 0\) and we discuss two cases according to whether g and f are linearly dependent or not.
First case: g and f are linearly dependent. There exists a constant \(c\in \mathbb {C}\) such that \(f=cg\). So (1.6) becomes
Then we obtain by proceeding exactly as in the proof of [10, Lemma4.3] and using [2, Lemma 4.1] the cases (1), (2), (3) and (4).
Second case: g and f are linearly independent.
Subcase A: \(g^{\circ }=0\) and \(f^{\circ }=0\). That is g and f are even, so if we replace y by \(\sigma (y)\), the functional Eq. (1.6) can be written as
According to [10, Theorem 3.1] and taking into account that f and g are linearly independent and \(g\ne 0\) we have the following possibilities:
(i) \(f=\alpha \dfrac{\chi _1+\chi _2}{2}+q\dfrac{\chi _1-\chi _2}{2}\) and \(g=\dfrac{\chi _1+\chi _2}{2}\pm \sqrt{1+q^2-\alpha ^2}\dfrac{\chi _1-\chi _2}{2}\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions and \(q\in \mathbb {C}\backslash \lbrace \pm \alpha \rbrace \) is a constant. So
Since f and g are even, we see that
On the other hand the fact that g and f are linearly independent implies that \(f-\alpha g\ne 0\), and so \(\dfrac{q\pm \alpha \sqrt{1+q^2-\alpha ^2}}{2}\ne 0\). So (3.15) reduces to
Then since g is even, we get that
So, we deduce from the last two identities that \(\chi _1=\chi _1^*\) and \(\chi _2=\chi _2^*\). This occurs in case (5).
(ii) \(\alpha \ne 0\), \(f=\alpha \chi _1\) and \(g=\chi _2\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions. In addition since f and g are even, we deduce that \(\chi _1^*=\chi _1\) and \(\chi _2^*=\chi _2\). This is case (6).
(iii) \(f=\alpha \chi +h\) and \(g=\chi \pm h\), where \(h,\chi :S\rightarrow \mathbb {C}\) is a solution of the sine addition law
such that \(\chi \ne 0\) is multiplicative and \(h\ne 0\). So \(f-\alpha g=(1\mp \alpha )h\), and \(1\pm \alpha \ne 0\) since f and g are linearly independent. Since f and g are even, we get that \(h^*=h\) and then \(\chi ^*=\chi \). This occurs in case (7).
Subcase B: \(g^{\circ }=0\) and \(f^{\circ }\ne 0\). In this case we deduce from (3.6) and (3.7) respectively that \(f^e(x)f^{\circ }(yz)=0\) and \(f^e(yz)=0\) for all \(x,y,z\in S\).
Subcase B.1: \(f^{\circ }=0\) on \(S^2\). That is \(f=0\) on \(S^2\) since \(f^e=0\) on \(S^2\). The functional Eq. (1.6) can be written as
Since \(f\ne 0\) (\(f^{\circ }\ne 0\)) we get according to Lemma 3.1 that f and g are linearly dependent. This is a contradiction. This case does not occur.
Subcase B.2: \(f^{\circ }\ne 0\) on \(S^2\). This implies that \(f^e=0\), so if we replace y by \(\sigma (y)\), the functional Eq. (1.6) can be written as
This means that the pair (g, if) satisfies (1.4), where \(\alpha \) is replaced by \(-i\alpha \). According to [10, Theorem 3.1] and taking into account that g and f are linearly independent and \(g\ne 0\) we have the following possibilities:
-
(i)
\(if=-i\alpha \dfrac{\chi _1+\chi _2}{2}+q\dfrac{\chi _1-\chi _2}{2}\) and \(g=\dfrac{\chi _1+\chi _2}{2}\pm \sqrt{1+q^2+\alpha ^2}\dfrac{\chi _1-\chi _2}{2}\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions and \(q\in \mathbb {C}\backslash \lbrace \pm i\alpha \rbrace \) is a constant. So
$$\begin{aligned} f=\left( \dfrac{-\alpha -iq}{2} \right) \chi _1+\left( \dfrac{iq-\alpha }{2} \right) \chi _2, \end{aligned}$$and
$$\begin{aligned} g=\left( \dfrac{1+\sqrt{1+q^2+\alpha ^2}}{2} \right) \chi _1+\left( \dfrac{1-\sqrt{1+q^2+\alpha ^2}}{2} \right) \chi _2, \end{aligned}$$or
$$\begin{aligned} g=\left( \dfrac{1-\sqrt{1+q^2+\alpha ^2}}{2} \right) \chi _1+\left( \dfrac{1+\sqrt{1+q^2+\alpha ^2}}{2} \right) \chi _2. \end{aligned}$$Since \(q\ne \pm i\alpha \), we see that \(\alpha \ne \pm iq\). Then since f is odd and g is even we get according to Lemma 3.2 (2) that
$$\begin{aligned} \dfrac{-\alpha -iq}{2} +\dfrac{iq-\alpha }{2}=0\end{aligned}$$and
$$\begin{aligned} \dfrac{1-\sqrt{1+q^2+\alpha ^2}}{2}=\dfrac{1+\sqrt{1+q^2+\alpha ^2}}{2}. \end{aligned}$$Then \(\alpha =0\) and \(\sqrt{1+q^2+\alpha ^2}=0\). This implies that \(iq=\pm 1\), so
$$\begin{aligned} f=\pm \left( \dfrac{\chi _2-\chi _1}{2} \right) , \end{aligned}$$and
$$\begin{aligned} g= \left( \dfrac{\chi _1+\chi _2}{2}\right) . \end{aligned}$$This occurs in case (8) with \(\alpha =0\), \(\chi :=\chi _1\) and \(\chi ^*:=\chi _2\) or \(\chi :=\chi _2\) and \(\chi ^*:=\chi _1\).
-
(ii)
\(\alpha \ne 0\), \(if=-i\alpha \chi _1\) and \(g=\chi _2\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions. Then \(f=-\alpha \chi _1\). In addition since f is odd, we deduce that \(\chi _1^*=-\chi _1\). Then \(\chi _1=0\), contradicting \(\chi _1\ne 0\). So case (ii) does not occur.
-
(iii)
\(if=-i\alpha \chi +h\) and \(g=\chi \pm h\), where \(\chi ,h:S\rightarrow \mathbb {C}\) is a solution of (1.5). Then \(f=-\alpha \chi -ih\). If \(g=\chi +h\), we get that
and then since f is odd and g is even
By adding the last two identities we obtain \(g=(i-\alpha )\dfrac{\chi +\chi ^*}{2i}\), and then \(f=(i-\alpha )\dfrac{\chi -\chi ^*}{2}\). The fact that \(f\ne 0\) implies \(\chi \ne \chi ^*\) and \(\alpha \ne i\). Substituting the forms of f and g in (1.6) we get after reduction that
for all \(x,y\in S\). Then since \(\chi \ne \chi ^*\) we deduce that \(1+i\alpha =0\). That is \(\alpha =i\), so this case does not occur since \(\alpha \ne i\). For the case \(g=\chi -h\) we proceed in the same way to get that \(g=(i+\alpha )\dfrac{\chi +\chi ^*}{2i}\) and \(f=(i+\alpha )\dfrac{\chi -\chi ^*}{2}\) with \(\chi \ne \chi ^*\) and \(\alpha \ne -i\). Similarly to the previous case we get by substitution that \(\alpha =-i\). This case does not occur.
Subcase C: \(g^{\circ } \ne 0\) and \(f^{\circ } =0\). We deduce from (3.6) and (3.7) respectively that \(g^e(x)g^{\circ }(yz)=0\) and \(g^e(yz)=0\) for all \(x,y,z\in S\).
Subcase C.1: \(g^{\circ } = 0\) on \(S^2\). Then \(g=0\) on \(S^2\) since \(g^e=0\) on \(S^2\), so the functional Eq. (1.6) becomes
That is
Since f and g are linearly independent, we see that \(\alpha \ne 0\), so for all \(x,y\in S\)
Then according to Lemma 3.1 we get that f and g are linearly dependent. This is a contradiction. This case does not occur.
Subcase C.2: \(g^{\circ } \ne 0\) on \(S^2\). In this case we get that \(g^e=0\), so if we replace y by \(\sigma (y)\) the functional Eq. (1.6) can be written as
This means that the pair \((-g,if)\) satisfies the functional Eq. (1.4), where \(\alpha \) is replaced by \(i\alpha \). According to [10, Theorem 3.1] and taking into account that g and f are linearly independent we have the following cases
(i) \(if=i\alpha \dfrac{\chi _1+\chi _2}{2}+q\dfrac{\chi _1-\chi _2}{2}\) and \(-g=\dfrac{\chi _1+\chi _2}{2}\pm \sqrt{1+q^2+\alpha ^2}\dfrac{\chi _1-\chi _2}{2}\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions and \(q\in \mathbb {C}\backslash \lbrace \pm i\alpha \rbrace \) is a constant. Then
and
or
Since \(q\ne \pm i\alpha \), we see that \(\alpha \ne \pm iq\). Then since f is even and g is odd we get according to the statement \(b_1+b_2=0\) in Lemma 3.2 (1) that
This case does not occur.
(ii) \(\alpha \ne 0\), \(if=i\alpha \chi _1\) and \(-g=\chi _2\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions. Then \(f=\alpha \chi _1\) and \(g=-\chi _2\). In addition since g is odd and f is even, we deduce that \(\chi _1^*=\chi _1\) and \(\chi _2^*=-\chi _2\). This means that \(\chi _2=0\). This case does not occur.
(iii) \(if=i\alpha \chi +h\) and \(-g=\chi \pm h\), where \(\chi ,h:S\rightarrow \mathbb {C}\) is a solution of (1.5). So \(f=\alpha \chi -ih\) and \(g=-\chi \pm h\). Suppose that \(g=-\chi -h\), to get that
then since f is even and g is odd we deduce that
By adding the last two identities, we get that
and then
The fact that \(g\ne 0\) implies that \(\chi \ne \chi ^*\) and \(\alpha \ne -i\). Substituting the forms of g and f in (1.6) we get that for all \(x,y\in S\)
Then since \(\chi \ne \chi ^*\), we deduce that \(1-i\alpha =0\). That is \(\alpha =-i\). This contradicts the fact that \(\alpha \ne -i\), so this case does not occur.
Now if \(g=-\chi +h\), we get in the same way that \(f=\dfrac{\alpha -i}{2}\left[ \chi +\chi ^* \right] \) and \(g=\dfrac{\alpha -i}{2i}\left[ \chi -\chi ^* \right] \), where \(\chi \ne \chi ^*\) and \(\alpha \ne i\), and then by substitution that \(\alpha =i\). This case does not occur.
Subcase D: \(g^{\circ }\ne 0\) and \(f^{\circ }\ne 0\).
Subcase D.1: \(g^{\circ }= 0\) on \(S^2\) and \(f^{\circ }= 0\) on \(S^2\). That is \(g=g^e\) on \(S^2\) and \(f=f^e\) on \(S^2\), and from (3.7) we deduce that \(f^e=cg^e\) on \(S^2\) for some constant \(c\in \mathbb {C}\). That is \(f=cg\) on \(S^2\), so the functional Eq. (1.6) can be written as
Since g and f are linearly independent, we see that \(\beta :=1-c\alpha \ne 0\). Then
This means that the pair \(\left( \dfrac{1}{\beta }g,\dfrac{i}{\beta }f \right) \) satisfies the functional Eq. (1.2), so taking into account that g and f are linearly independent we get from [2, Theorem 4.2] that \(\dfrac{1}{\beta }g\) is even. This is a contradiction since \(g^{\circ }\ne 0\). This case does not occur.
Subcase D.2: \(g^{\circ }= 0\) on \(S^2\) and \(f^{\circ }\ne 0\) on \(S^2\). We get from (3.6) that \(f^e=0\), then from (3.7) that \(g^e=0\) on \(S^2\). That is \(g=0\) on \(S^2\). The functional Eq. (1.6) becomes
Since g and f are linearly independent, we see that \(\alpha \ne 0\), so
So we get according to Lemma 3.1 that f and g are linearly dependent. This is a contradiction. This case does not occur.
Subcase D.3: \(g^{\circ }\ne 0\) on \(S^2\) and \(f^{\circ }= 0\) on \(S^2\). We deduce from (3.6) that \(g^e=0\), and then by using (3.7) we get that \(f^e=0\) on \(S^2\). This implies that \(f=0\) on \(S^2\), and (1.6) can be written as
Since \(f\ne 0\) (\(f^{\circ }\ne 0\)) we get according to Lemma 3.1 that f and g are linearly dependent. This case does not occur.
Subcase D.4: \(g^{\circ }\ne 0\) on \(S^2\) and \(f^{\circ }\ne 0\) on \(S^2\). By using (3.6) we get that
for some constant \(\delta \in \mathbb {C}\).
Subcase D.4.1: \(f^e= 0\). So \(g^e=0\), and then if we replace y by \(\sigma (y)\) the functional Eq. (1.6) becomes
This means that the pair \(\left( -g,f \right) \) satisfies (1.4), where \(\alpha \) is replaced by \(-\alpha \). According to [10, Theorem 3.1] and taking into account that g and f are linearly independent we have the following cases
(i) \(f=-\alpha \dfrac{\chi _1+\chi _2}{2}+q\dfrac{\chi _1-\chi _2}{2}\) and \(-g=\dfrac{\chi _1+\chi _2}{2}\pm \sqrt{1+q^2-\alpha ^2}\dfrac{\chi _1-\chi _2}{2}\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions and \(q\in \mathbb {C}\backslash \lbrace \pm \alpha \rbrace \) is a constant. Then \(g=-\dfrac{\chi _1+\chi _2}{2}\pm \sqrt{1+q^2-\alpha ^2}\dfrac{\chi _1-\chi _2}{2}\).
So \(f-\alpha g=\dfrac{q\pm \alpha \sqrt{1+q^2-\alpha ^2}}{2}\left( \chi _1-\chi _2 \right) \), and \(\dfrac{q\pm \alpha \sqrt{1+q^2-\alpha ^2}}{2}\ne 0\) since g and f are linearly independent. The fact that f and g are odd implies that
This implies that
Then, since g is odd we get that
So we deduce from the last two identities that \(\chi _1=-\chi _1^*\) and \(\chi _2=-\chi _2^*\). This implies that \(\chi _1=\chi _2=0\). This case does not occur.
(ii) \(\alpha \ne 0\), \(f=-\alpha \chi _1\) and \(-g=\chi _2\), where \(\chi _1,\chi _2:S\rightarrow \mathbb {C}\) are two different non-zero multiplicative functions. Then \(g=-\chi _2\). In addition since g and f are odd, we deduce that \(\chi _1^*=-\chi _1\) and \(\chi _2^*=-\chi _2\). That is \(\chi _1=\chi _2=0\). This case does not occur.
(iii) \(f=-\alpha \chi +h\) and \(-g=\chi \pm h\), where \(\chi :S\rightarrow \mathbb {C}\) is a non-zero multiplicative function and \(h:S\rightarrow \mathbb {C}\) is a solution of (1.5). So \(g=-\chi \mp h\). Then \(f-\alpha g=(1\pm \alpha )h\) and \(1\pm \alpha \ne 0\) since g and f are linearly independent. Since g and f are odd, we see that \(h=-h^*\), and then \(\chi =-\chi ^*\). This implies that \(\chi =0\). This case does not occur.
Subcase D.4.2: \(f^e\ne 0\). We get from (3.6) that \(f^{\circ }=bg^{\circ }\) on \(S^2\) for some constant \(b\in \mathbb {C}\) and \(\delta \ne 0\).
Subcase D.4.2.1: \(f^e= 0\) on \(S^2\). So \(g^e=0\) on \(S^2\) by (3.16). That is \(f=bg\) on \(S^2\) since \(f^e=g^e=0\) on \(S^2\). The functional Eq. (1.6) can be written as
The fact that g and f are linearly independent implies that \(\gamma :=1-b\alpha \ne 0\), so we get
This means that the pair \(\left( \dfrac{1}{\gamma } g,\dfrac{i}{\gamma }f\right) \) satisfies (1.2), then since f and g are linearly independent we get from [2, Theorem 4.2] that \(\dfrac{1}{\gamma } g\) is even. This contradicts the fact that \(g^{\circ }\ne 0\) here in subcase D. This case does not occur.
Subcase D.4.2.2: \(f^e\ne 0\) on \(S^2\). In this case we get from (3.7) that \(f^{\circ }=ag^{\circ }\) for some constant \(a\in \mathbb {C}\), and \(a\ne 0\) since \(f^{\circ }\ne 0\). So
So with (3.16), Eq. (3.6) becomes
for all \(x,y,z\in S\). This implies that \(\dfrac{\delta }{a}=1\), and then \(\delta =a\ne 0\), so we get that
By adding (3.17) to (3.16) we get that
Since f and g are linearly independent, we see from (3.18) that \(\delta \ne \pm 1\). So in view of (3.18) the functional Eq. (1.6) becomes
If we apply (3.19) first to the pair \((x,\sigma (y))\) and then to the pair \((\sigma (x),y)\) we get respectively the following two identities
By adding (3.20) to (3.21) and taking into account that \(k^*=k^e-k^{\circ }\) and \(k=k^e+k^{\circ }\) for any function \(k:S\rightarrow \mathbb {C}\) we get
Now by using (3.16) and (3.17), the identitiy (3.22) becomes
Since \(\delta \ne \pm 1\) and \(f^e\ne 0\), we see from (3.23) that \(\alpha \ne \delta \), and if we apply (3.23) to the pair \((\sigma (y),x)\) we get
By adding (3.24) to (3.23) we get since \(\alpha \ne \delta \)
According to [9, Theorem 2.1] we deduce from (3.25) that
where \(\chi :S\rightarrow \mathbb {C}\) is multiplicative. \(\chi \) is non-zero since \(f^e\ne 0\). Substituting (3.26) in (3.23) and taking into account that \(f^{\circ }\ne 0\) we obtain
and \(\chi \ne \chi ^*\), so we deduce from (3.26) and (3.27) that
or
and then by using (3.16) and (3.17) that
or
So we deduce that
or
By substituting (3.28) and (3.29) in (1.6) we deduce after reduction that \(\delta \alpha =1\). That is \(\delta =\dfrac{1}{\alpha }\), and then we get that
or
where \(\alpha \ne \pm 1\). This occurs in case (8).
Finally, if S is a topological semigroup, we get the continuity of g in cases (1), (2), (3), (4), (5) and (7) from the continuity of f by proceeding like in the proof of [10, Proposition 5.1]. Case (6) contains the excluded form of f. For case (8) we get according to [8, Theorem 3.18(d)] from
that \((1+\alpha )\chi ,(1-\alpha )\chi ^*\in C(S)\). So \(\alpha \ne \mp 1\) implies that \(\chi ,\chi ^*\in C(S)\), then \(g\in C(S)\) as a linear combination of two continuous functions. This completes the proof of Theorem 3.4. \(\square \)
Remark 3.1
Theorem 3.4(7) can by the help of Ebanks [6, Theorem 3.1(B)] be formulated in more details as follows:
\(f=\alpha \chi +h\) and \(g=\chi \pm h\) such that
where \(\chi :S\rightarrow \mathbb {C}\) is an even non-zero multiplicative function, \(A:S\backslash I_{\chi }\rightarrow \mathbb {C}\) an additive function such that \(A\circ \sigma =A\), and \(\rho :P_\chi \rightarrow \mathbb {C}\) a function such that \(\rho \circ \sigma =\rho \). In addition we have the following conditions.
-
(I):
If \(x\in \lbrace up,pv,upv\rbrace \) for \(p\in P_{\chi }\) and \(u,v\in S\backslash I_{\chi }\), then we have respectively \(\rho (x)=\rho (p)\chi (u)\), \(\rho (x)=\rho (p)\chi (v)\), or \(\rho (x)=\rho (p)\chi (uv)\).
-
(II):
\(h(xy)=h(yx)=0\) for all \(x\in I_{\chi }\backslash P_{\chi }\) and \(y\in S\backslash I_{\chi }\).
4 Examples
In this section we give some examples of solutions of the functional Eq. (1.6).
Example 4.1
Let \(S=(\mathbb {R},+)\) under the usual topology, and let \(\sigma :\mathbb {R}\rightarrow \mathbb {R}\) be the involution defined by \(\sigma (x)=-x\) for all \(x\in \mathbb {R}\). The functional Eq. (1.6) can be written as
where \(f,g:\mathbb {R}\rightarrow \mathbb {C}\). We determine the continuous solutions of (4.1). The case \(\alpha =0\) is [8, Example 4.18] where the function f is replaced by if. The continuous non-zero multiplicative functions on S are the functions
where \(\lambda \in \mathbb {C}\). The only even, non-zero multiplicative function \(\chi \) on S is \(\chi =1\). The only even, additive fonction a on S is \(a=0\). The solutions \(f,g\in C(S)\) of (4.1) are the following:
-
(a)
\(\alpha =\pm 1\), f is any non-zero continuous function on S and \(g=\alpha f\).
-
(b)
\(f=\dfrac{q+\alpha }{2}\) and \(g=\dfrac{1\pm \sqrt{1+q^2-\alpha ^2}}{2}\), where \(q\in \mathbb {C}\) is a constant.
-
(c)
\(f(x)=\dfrac{1+\alpha }{2}e^{i\lambda x}- \dfrac{1-\alpha }{2}e^{-i\lambda x}\) and \(g(x)=\dfrac{1+\alpha }{2}e^{i\lambda x}+\dfrac{1-\alpha }{2}e^{-i\lambda x}\), where \(\lambda \in \mathbb {C}\backslash \lbrace 0\rbrace \) and \(\alpha \ne \mp 1\).
Example 4.2
Let \(S=H_3\) be the Heisenberg group defined by
and let
for all \(x,y,z\in \mathbb {R}\). We consider the following involution
According to [8, Example 3.14], the continuous non-zero multiplicative functions on S have the form
where \(a,b\in \mathbb {C}\). The only even, non-zero multiplicative function \(\chi \) on S is \(\chi =1\). So the continuous solutions of Eq. (1.6) are the following three types:
-
(1)
\(\alpha =\pm 1\), f is any non-zero continuous function on S and \(g=\alpha f\).
-
(2)
\(f(X)=\dfrac{q+\alpha }{2}\) and \(g(X)=\dfrac{1\pm \sqrt{1+q^2-\alpha ^2}}{2}\), where \(q\in \mathbb {C}\).
-
(3)
\(\alpha \ne \pm 1\), and
$$\begin{aligned} \begin{matrix} f(X)=\dfrac{1+\alpha }{2}e^{ax+by}-\dfrac{1-\alpha }{2}e^{-ax-by},\\ g(X)=\dfrac{1+\alpha }{2}e^{ax+by}+\dfrac{1-\alpha }{2}e^{-ax-by}, \\ \end{matrix} \end{aligned}$$where \((a,b)\ne (0,0)\).
In the following example we shall apply our theory to a semigroup S such that \(S^2\ne S\).
Example 4.3
Let \(S=\left( \mathbb {N}\backslash \lbrace 1\rbrace ,\cdot \right) \), and let \(\chi :S\rightarrow \mathbb {C}\) be the multiplicative function defined by
Then \(I_{\chi }=2\mathbb {N}\) and \(P_{\chi }=2\mathbb {N}\backslash 4\mathbb {N}\). We let \(\sigma :=id\) be the identity function. So \(\chi \) is even with respect to \(\sigma \). An additive function \(A:\mathbb {N}\backslash \left( 2\mathbb {N}\cup \lbrace 1\rbrace \right) \rightarrow \mathbb {C}\) is defined by
In Remark 3.1, the form of the function \(h:\mathbb {N}\backslash \lbrace 1\rbrace \rightarrow \mathbb {C}\) is
where \(\rho :2\mathbb {N}\backslash 4\mathbb {N}\rightarrow \mathbb {C}\) is a function satisfying condition (I). So
This implies that \(\rho (x)=c\) for all \(x\in 2\mathbb {N}\backslash 4\mathbb {N}\), where \(c:=\rho (2)\in \mathbb {C}\). That is
So condition (II) is satisfied. We get the solutions \(f,g:S\rightarrow \mathbb {C}\) of Eq. (1.6) by plugging the appropriate forms above into the formula of Theorem 3.4 and Remark 3.1.
For the semigroup S we have that \(S^2=S\backslash \mathbb {P}\), where \(\mathbb {P}\) denotes the set of all prime numbers, and also S is neither a monoid nor generated by its squares, so we can see that we can not avoid the cases (2), (3) and (7) from Theorem 3.4.
References
Ajebbar, O., Elqorachi, E.: Solutions and stability of trigonometric functional equations on an amenable group with an involutive automorphism. Commun. Korean Math. 34(1), 55–82 (2019). https://doi.org/10.4134/CKMS.C170487
Aserrar, Y., Elqorachi, E.: Five trigonometric addition laws on semigroups. arXiv:2210.06181
Chung, J.K., Kannappan, P., Ng, C.T.: A generalization of the Cosine-Sine functional equation on groups. Linear Algebra Appl. 66, 259–277 (1985)
Ebanks, B.: The sine addition and subtraction formulas on semigroups. Acta Math. Hung. 164(2), 533–555 (2021)
Ebanks, B.: The cosine and sine addition and subtraction formulas on semigroups. Acta Math. Hung. 2(165), 337–354 (2021)
Ebanks, B.: Around the sine addition law and d’Alembert’s equation on semigroups. Results Math. 77, 11 (2022). https://doi.org/10.1007/s00025-021-01548-6
Poulsen, T.A., Stetkær, H.: On the trigonometric subtraction and addition formulas. Aequat. Math. 59(1–2), 84–92 (2000)
Stetkær, H.: Functional Equations on Groups. World scientific Publishing Co, Singapore (2013)
Stetkær, H.: A variant of d’Alembert’s functional equation. Aequat. Math. 89(3), 657–662 (2015)
Stetkær, H.: The cosine addition law with an additional term. Aequat Math. 90(6), 1147–1168 (2016)
Vincze, E.: Eine allgemeinere Methode in der Theorie der Funktionalgleichungen II. Publ. Math. Debrecen 9, 314–323 (1962)
Funding
None.
Author information
Authors and Affiliations
Contributions
The authors confirm contribution to the paper as follows: study conception and design: YA, EE; data collection: YA; analysis and interpretation of results: YA, EE; draft manuscript preparation: YA. All authors reviewed the results and approved the final version of the manuscript.
Corresponding author
Ethics declarations
Ethical approval
Not applicable.
Conflict of interest
None.
Availability of data and materials
Not applicable.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Aserrar, Y., Elqorachi, E. A generalization of the cosine addition law on semigroups. Aequat. Math. 97, 787–804 (2023). https://doi.org/10.1007/s00010-023-00946-1
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00010-023-00946-1