Multiplicative spectral functions on some Banach function algebras Nahid Bayati and Fereshteh Sady Abstract . In this paper, we study multiplicative functions φ : A → ℂ \varphi\colon A\to\mathbb{C} on a natural Banach function algebra A A on a compact Hausdorff space X X , such that φ ⁡ ( f ) ∈ σ ⁡ ( f ) \varphi(f)\in\sigma(f) for all f ∈ A f\in A . It is shown that for certain natural Banach function algebras A A , either ker ⁡ ( φ ) \ker(\varphi) is a maximal ideal of A A or 1 ∈ span ⁡ ( ker ⁡ ( φ ) ) 1\in{\rm span}({\rm ker}(\varphi)) (that is 1 = f 1 + f 2 + ⋯ f n 1=f_{1}+f_{2}+\cdots f_{n} for some f 1 , … , f n ∈ ker ⁡ ( φ ) f_{1},...,f_{n}\in{\rm ker}(\varphi) ). Then we investigate for the linearity of φ \varphi in either of cases that φ \varphi is continuous or 1 ∉ span ⁡ ( ker ⁡ ( φ ) CLOSE 1\notin{\rm span}({\rm ker}(\varphi) . We show that, for some natural Banach function algebras A A , in either of these cases, there exists a point x 0 ∈ X x_{0}\in X such that φ ⁡ ( f ) = f ⁡ ( x 0 ) \varphi(f)=f(x_{0}) for some family of functions f ∈ A f\in A (including those functions f ∈ A f\in A that f ¯ ∈ A \overline{f}\in A ). In particular, such a multiplicative spectral function on some Banach algebras including C ⁡ ( X ) C(X) , Lipschitz algebras, Banach algebras of absolutely continuous functions on [ 0 , 1 ] [0,1] and C 1 ​ ( [ 0 , 1 ] ) C^{1}([0,1]) is linear and hence it is a character. 2020 Mathematics Subject Classification Primary 46J10 and 46J20, Secondary 47A11 1. Introduction For a unital complex Banach algebra A , let G(A) be the group of all invertible elements of A , G_{1}(A) be the principle component of G(A) and {\rm exp}(A) be the subset \{e^{a}:a\in A\} of G_{1}(A) . Given a\in A , let \sigma(a) denote its spectrum. We say that a function (not necessarily linear) \varphi\colon A\to\mathbb{C} on a Banach algebra A is spectral, if \varphi(a)\in\sigma(a) for all a\in A . By the classical Gleason-Kahane-Żelazko theorem, any complex-linear spectral function \varphi\colon A\to\mathbb{C} on a unital Banach algebra A is a character on A . The Kowalski-Słodowski theorem [ 6 ] provides an additive version of the Gleason-Kahane-Żelazko theorem by showing that any complex function \varphi\colon A\to\mathbb{C} that satisfies \varphi(a)+\varphi(b)\in\sigma(a+b) for all a,b\in A , is a character on A . Without the linearity assumption, the following result has been proven by A. Maouche for multiplicative spectral functions on a unital Banach algebra. Theorem 1.1 . [ 8 ] Let A be a (complex) unital Banach algebra and \varphi\colon A\to\mathbb{C} be a multiplicative function such that \varphi(a)\in\sigma(a) for all a\in A . Then there exists a unique character \psi on A such that \psi(a)=\varphi(a) for all a\in G_{1}(A) . The example given in [ 8 ] shows that the above multiplicative spectral function \varphi is not necessarily linear. However, under the continuity assumption, there are some recent results concerning the linearity of such multiplicative functions. For a compact Hausdorff space X , it was shown in [ 9 ] any continuous multiplicative spectral function on the Banach algebra C(X) of all continuous complex-valued functions on X , is a character and, in particular, it is linear. More generally, in [ 2 ] the same result has been obtained for continuous multiplicative spectral functions on arbitrary C^{*} -algebras. Finally, it was shown in [ 7 ] , that if A is a Hermitian Banach algebra, then any continuous multiplicative spectral function \varphi\colon A\to\mathbb{C} is a character, as well. We note that such results can be considered as multiplicative versions of the Gleason-Kahane-Żelazko theorem. We also refer to [ 3 ] , in which a multiplicative Kowalski-Słodkowski theorem has been proven for a Hermitian algebra A , i.e. it was shown that if \varphi:A\to\mathbb{C} is a continuous function satisfying \varphi(x)\varphi(y)\in\sigma(xy) for all x,y\in A , then either \varphi or -\varphi is a character on A . We should note that for a compact Hausdorff space X , it was shown in [ 1 ] that for a multiplicative spectral function \varphi:C(X)\to\mathbb{C} (not assumed to be continuous), either {\rm ker}(\varphi) is a maximal ideal or 1\in{\rm span}({\rm ker}(\varphi)) and \varphi is linear in either of cases that \varphi is continuous or 1\notin{\rm span}({\rm ker}(\varphi)) . In this paper, we improve the results of [ 1 ] . We consider the case that A is a natural Banach function algebra on a compact Hausdorff space X and study multiplicative spectral functions \varphi\colon A\to\mathbb{C} . In the lack of continuity assumption, we first show that for certain natural Banach function algebras A , either \ker(\varphi) is a maximal ideal of A or it spans A , i.e. 1=f_{1}+\cdots f_{n} for some f_{1},...,f_{n}\in\ker(\varphi) . Moreover, in the uniform algebra case, there exists a point x_{0}\in X such that \varphi(f)=f(x_{0}) for all f\in G(A) (Proposition 3.1 ). Then we investigate for the linearity of \varphi whenever either 1\notin\ker(\varphi) or \varphi is continuous. Indeed, we prove two theorems (Theorems 3.7 and 3.10 ) providing a point x_{0}\in X such that \varphi(f)=f(x_{0}) for some family of functions f\in A , including those functions f\in A with \overline{f}\in A . The first theorem concludes a result concerning the linearity of multiplicative spectral functions on the Banach algebras such as C(X) , Lipschitz algebras and also the Banach algebra of all absolutely continuous functions on [0,1] . The second theorem provides another linearity result for some Banach function algebras such as C^{1}([0,1]) . Our approach is based on using peaking functions, strong boundary points and their properties in uniform algebras and Banach function algebras. 2. Preliminaries For a compact Hausdorff space X , we denote the Banach algebra of all continuous complex-valued functions on X by C(X) . The supremum norm of a function f\in C(X) will be denoted by \|f\|_{X} . A point separating subalgebra A of C(X) which contains the constant functions is called a Banach function algebra on X if it is a Banach algebra under some norm \|\cdot\| . In the case that the norm of a Banach function algebra A on X is the same supremum norm, we call it a uniform algebra on X . For a Banach function algebra A on X , \overline{A} is the uniform closure of A . It is easy to see that the norm \|\cdot\| of a Banach function algebra on X satisfies the inequality \|\cdot\|\geq\|\cdot\|_{X} . A Banach function algebra A on X is called natural if each character on A is an evaluation homomorphism at some point of X . For a subspace A of C(X) , let A^{*} denote its dual space (with respect to the supremum norm). For each x\in X , \varphi_{x} is the evaluation functional on A at x and the Choquet boundary {\rm ch}(A) of A consists of all points x\in X such that \varphi_{x} is an extreme point of the closed unit ball of A^{*} . It is well-known that {\rm ch}(A) is a boundary for A , that is for each f\in A there exists a point x_{0}\in{\rm ch}(A) such that |f(x_{0})|=\|f\|_{X} . A point x_{0}\in X is called a strong boundary point of A if for each neighborhood U of x_{0} and \varepsilon>0 there exists a function f\in A such that f(x_{0})=1=\|f\|_{X} and |f(x)|\leq\varepsilon for all x\in X\setminus U . We denote the set of all strong boundary points of A by \Theta(A) . It should be noted that \Theta(A)\subseteq{\rm ch}(A) and in the uniform algebra case we have \Theta(A)={\rm ch}(A) [ 4 , Theorems 2.2.1, 2.3.4] . Clearly for a compact Hausdorff space X , the uniform algebra C(X) is natural and its Choquet boundary is the same X . For an example of a nontrivial natural uniform algebra A on a compact (metric) space X with {\rm ch}(A)=X we can refer to the Cole example (see [ 4 , Page 255] ). Here are some examples of natural Banach function algebras A with \Theta(A)={\rm ch}(A)=X : (i) The Banach algebra {\rm Lip}(X) consisting of all complex Lipschitz functions on a compact metric space X endowed with the norm \|f\|=\|f\|_{X}+p(f) where p(f) is the Lipschitz constant of f\in{\rm Lip}(X) . (ii) The Banach algebra C^{1}([0,1]) of all continuously differentiable functions on the unit interval [0,1] endowed with the norm \|f\|=\|f\|_{[0,1]}+\|f^{\prime}\|_{[0,1]} . (iii) The Banach algebra {\rm AC}([0,1]) of all absolutely continuous functions on [0,1] under the norm \|f\|=\|f\|_{[0,1]}+{\rm var}(f) where {\rm var}(f) is the total variation of f\in{\rm AC}([0,1]) . Let A be a subspace of C(X) . A function f\in A is called a peaking function of A , if 1\in f(X)\subseteq{\rm int}(\mathbb{D})\cup\{1\} where \mathbb{D} is the closed unit disk in the complex plane. For x_{0}\in A , we use the notation P_{x_{0}}(A) for the set of all peaking functions f\in A with f(x_{0})=1 . Such a function f is called a peaking function at x_{0} . Hence for a peaking function f\in A at x_{0} , its maximum set M(f)=\{x\in X:|f(x)|=\|f\|_{X}\} is a closed subset of X containing x_{0} such that |f(y)|<1 for all y\in X\setminus M(f) . A point x_{0}\in X is called a peak point of A if there exists a peaking function f\in P_{x_{0}}(A) with M(f)=\{x_{0}\} . We should note that if the subspace A of C(X) contains the constant function 1, then for each \varepsilon>0 , x_{0}\in\Theta(A) and open neighborhood U of x_{0} , replacing f by \frac{1+f}{2} , we can find a function f\in P_{x_{0}}(A) such that |f|<\varepsilon on X\setminus U . The following lemma is easily verified. For the sake of completeness we state its proof. Lemma 2.1 . Let A be a Banach function algebra on X . Then for each x_{0}\in{\rm ch}(A) and f\in A we have |f(x_{0})|=\inf\{\|fh\|_{X}:h\in P_{x_{0}}(\overline{A})\cap{\rm exp}(\overline{A})\} (2.1) Proof. The equality is obvious if f=0 . So we assume that f is nonzero. Clearly for each h\in P_{x_{0}}(\overline{A})\cap{\rm exp}(\overline{A}) we have |f(x_{0})|\leq\|fh\|_{X} . Now assume that x_{0}\in{\rm ch}(A) . Given \epsilon\in(0,1) , we set U=\{x\in X:|f(x)|<|f(x_{0})|+\epsilon\} . Clearly U is a neighborhood of x_{0} . We note that {\rm ch}(A)={\rm ch}(\overline{A}) and since \overline{A} is a uniform algebra on X we have {\rm ch}(\overline{A})=\Theta(\overline{A}) . Hence there exists a function u\in P_{x_{0}}(\overline{A}) such that |u|<\epsilon on X\setminus U . Then for any n\in\mathbb{N} with e^{-n(1-\varepsilon)}<\frac{\epsilon}{\|f\|_{X}} , the function h=e^{-n(1-u)} is an element of P_{x_{0}}(\overline{A})\cap{\rm exp}(\overline{A}) such that |h|<\frac{\epsilon}{\|f\|_{X}} on X\setminus U . An easy verification shows that \|fh\|_{X}\leq|f(x_{0})|+\epsilon , which proves the desired equality. ∎ The following lemma, called multiplicative Bishop’s lemma, has been proven in [ 5 ] . Lemma 2.2 . Let A be a uniform algebra on a compact Hausdorff space X and x_{0}\in{\rm ch}(A) . Then for each f\in A with f(x_{0})\neq 0 there exists g\in P_{x_{0}}(A) such that \frac{1}{f(x_{0})}fg\in P_{x_{0}}(A) . Remark 2.3 . i) By [ 5 , Page 283] , we may choose the desired function g in the above lemma in such a way that g is an element of {\rm exp}(A) . ii) We should note that the closedness of A in C(X) in the above lemma has an important role in its proof. However, as the next lemma shows, in some Banach function algebras like the Banach algebra of Lipschitz functions on a compact metric space or absolutely continuous functions on the unit interval, we may obtain directly the function g with the desired properties. Lemma 2.4 . Let A be the Banach function algebra {\rm Lip}(X) for some compact metric space (X,d) . Then for each point x_{0}\in X and f\in A with f(x_{0})\neq 0 there exists a nonnegative peaking function g\in P_{x_{0}}(A) such that \frac{1}{f(x_{0})}fg\in P_{x_{0}}(A) . The same conclusion holds for A={\rm AC}(X) where X=[0,1] . Proof. First consider the case that A={\rm Lip}(X) . We may assume that f(x_{0})=1 . Given x_{0}\in X , the function h_{x_{0}}\colon X\to[0,1] defined by h_{x_{0}}(x)=1-\frac{d(x,x_{0})}{{\rm diam}(X)} is a peaking function of A with M(h_{x_{0}})=\{x_{0}\} , that is x_{0} is a peak point of A . We can choose easily a function g_{0}\in A with values in [0,1] such that g_{0}=1 on the set \{x\in X:|f(x)|\leq 1\} and g_{0}=0 on \{x:|f(x)|\geq 2\} . Now let g:X\to[0,1] be defined by g={\rm min}(2-|f(x)|,1)h_{x_{0}}g_{0} . Then an easy verification shows that g is a peaking function in A satisfying the desired properties. The same proof works for A={\rm AC}([0,1]) , since any Lipschitz function is absolutely continuous. ∎ 3. Main Result Throughout this section, unless otherwise is stated, we assume that A is a natural Banach function algebra on a compact Hausdorff space X . We put A_{+}=\{f\in A:f\geq 0\} . As we noted before, by a spectral function on A we mean a (not necessarily continuous) function \varphi\colon A\to\mathbb{C} satisfying \varphi(f)\in\sigma(f) for all f\in A . Clearly in this case we have \sigma(f)=f(X) for all f\in A . By Theorem 1.1 for each multiplicative spectral function \varphi\colon A\to\mathbb{C} there exists a unique point x_{0}\in X such that \varphi(e^{f})=e^{f(x_{0})} for all f\in A . We call x_{0} the associated point of \varphi . We begin with the following proposition concerning the kernel of a multiplicative spectral function on A . Proposition 3.1 . Let {\rm ch}(A)=X and \varphi:A\to\mathbb{C} be a (not necessarily continuous) multiplicative spectral function. Then the following statements hold. (i) There exists a unique point x_{0}\in X such that |\varphi(f)|\leq|f(x_{0})| for all f\in A . In particular, |\varphi(f)|=|f(x_{0})| for all f\in G(A) . (ii) Either \ker(\varphi) is a maximal ideal of A or 1=f_{1}+\cdots+f_{n} for some f_{1},...,f_{n}\in\ker(\varphi) . (iii) In the case that A is a uniform algebra, we have \varphi(f)=f(x_{0}) for all f\in G(A) . Proof. i) Let x_{0}\in X be the associated point of \varphi , i.e. \varphi=\varphi_{x_{0}} on {\rm exp}(A) . We show that |\varphi(f)|\leq|f(x_{0})| for all f\in A . Given f\in A and \epsilon>0 , by Lemma 2.2 , there exists a function h\in P_{x_{0}}(\overline{A})\cap{\rm exp}(\overline{A}) such that \|fh\|_{X}<|f(x_{0})|+\epsilon . Choose a sequence \{h_{n}\} in {\rm exp}(A) which uniformly converges to h . Then we have \|fh_{n}\|_{X}<|f(x_{0})|+\epsilon for sufficiently large n\in\mathbb{N} . Clearly we have \varphi(h_{n})=h_{n}(x_{0})\to 1 . Hence, using the spectral condition once again, we have |\varphi(f)h_{n}(x_{0})|=|\varphi(fh_{n})|\leq\|fh_{n}\|_{X}<|f(x_{0})|+\epsilon, for sufficiently large n\in\mathbb{N} . Tending n\to\infty , since \epsilon>0 is arbitrary we get |\varphi(f)|\leq|f(x_{0})| , as desired. Now assume that f\in G(A) . Without loss of generality we assume that f(x_{0})=1 . Since |\varphi(f)|\leq|f(x_{0})|=1 , |\varphi(f^{-1})|\leq|f^{-1}(x_{0})|=1 and \varphi(ff^{-1})=1 we get |\varphi(f)|=1=|f(x_{0})| . The uniqueness property is immediate from the equality |\varphi(e^{f})|=e^{{\rm Re}(f(x_{0}))} for all f\in A , since A separates the points of X . We note that, by the above argument, x_{0} is indeed, the same associated point to \varphi , that is \varphi(e^{f})=e^{f(x_{0})} for all f\in A . ii) Clearly, by part (i), we have \varphi(f)=f(x_{0}) for all f\in A with f(x_{0})=0 , that is {\rm ker}(\varphi_{x_{0}})\subseteq\ker(\varphi) . Being \varphi multiplicative, it follows that {\rm span}(\ker(\varphi)) is an ideal of A . Hence we have either 1\in{\rm span}(\ker(\varphi)) or {\rm span}(\ker(\varphi))=\ker(\varphi_{x_{0}}) . In the latter case, we have clearly \ker(\varphi)=\ker(\varphi_{x_{0}}) , i.e. \ker(\varphi) is a maximal ideal of A . iii) Let A be a uniform algebra and f\in G(A) . We again assume that f(x_{0})=1 . Then by Lemma 2.2 and Remark 2.3 , there exists a function h\in P_{x_{0}}(A)\cap{\rm exp}(A) such that fh\in P_{x_{0}}(A) . Clearly \varphi(h)=h(x_{0})=1 . Being fh invertible in A , we have |\varphi(fh)|=|fh(x_{0})|=1 and since \varphi(fh)\in fh(X)\subseteq\mathbb{\rm int}(\mathbb{D})\cup\{1\} we get \varphi(fh)=1 . Thus \varphi(f)=\varphi(fh)\varphi(h^{-1})=1=f(x_{0}) . ∎ Next corollary gives a generalization of part (iii) in the above proposition. Corollary 3.2 . Under the assumptions of Proposition 3.1 (iii), we have \varphi(f)\in\{0,f(x_{0})\} for each f\in A such that 0 is not a limit point of f(X) . Proof. We first show that \varphi(h)\in\{0,h(x_{0})\} if h\in P_{x_{0}}(A) and 0 is not a limit point of h(X) . To do this, we note that for each n\in\mathbb{N} we have \varphi(e^{n(1-h)})=1 and consequently, by hypotheses, we have \varphi(h)=\varphi\big(h\,e^{n(1-h)}\big)=h(x_{n})e^{n(1-h(x_{n}))}\qquad(n\in\mathbb{N}), for some sequence \{x_{n}\} in X . Passing through a subsequence, we assume that h(x_{n})\to c for some c\in h(X)\subseteq{\rm int}(\mathbb{D})\cup\{1\} . If c=0 , then since 0 is not a limit point of h(X) , we have h(x_{n})=0 for all sufficiently large n\in\mathbb{N} . In particular, \varphi(h)=0 in this case. So assume that c\neq 0 . Clearly, in this case the sequence \{e^{n(1-h(x_{n}))}\} is convergent. If c\neq 1 (equivalently {\rm Re}(c)\neq 1 ), then there exists r>0 such that 1-{\rm Re}(h(x_{n}))>r and |h(x_{n})|>r for all sufficiently large n\in\mathbb{N} . Thus |\varphi(h)|=|h(x_{n})|e^{n\big(1-{\rm Re}(h(x_{n}))\big)}\geq re^{nr} for all sufficiently large n\in\mathbb{N} , which is impossible. Hence c=1 and consequently |\varphi(h)|=\lim_{n\to\infty}e^{n\big(1-{\rm Re}(h(x_{n}))\big)}\geq 1 . Therefore |\varphi(h)|=1 which concludes that \varphi(h)=1 , since h is a peaking function. Now assume that f\in A such that 0 is not a limit point of f(X) . Since |\varphi(f)|\leq|f(x_{0})| , by Proposition 3.1 , we have \varphi(f)=0 if f(x_{0})=0 . So we may assume that f(x_{0})=1 . Then, by Lemma 2.2 , there exists a function k\in P_{x_{0}}(A)\cap{\rm exp}(A) such that fk\in P_{x_{0}}(A) . Obviously we have \varphi(f)=\varphi(fk) . Set h=fk . We claim that 0 is not a limit point of h(X) , as well. Indeed, assume that there exists a sequence \{h(x_{n})\} of distinct points in h(X) converging to 0 . Passing through a subsequence we may assume that \{f(x_{n})\} converges to a point c\in f(X) . If c\neq 0 , then k(x_{n})\to 0 , that is 0\in k(X) which is impossible, since k\in{\rm exp}(A) . Hence f(x_{n})\to 0 and since 0 is not a limit point of f(X) we get f(x_{n})=0 for sufficiently large n\in\mathbb{N} . Thus h(x_{n})=0 for sufficiently large n\in\mathbb{N} . This shows that 0 is not a limit point of h(X) . Therefore, by the above argument, we have \varphi(f)=\varphi(fk)\in\{0,fk(x_{0})\}=\{0,1\} , as desired. ∎ Lemma 3.3 . Let \varphi\colon A\to\mathbb{C} be a multiplicative spectral function on the natural Banach function algebra A and x_{0}\in X be its associated point. Then for each real-valued f\in A we have \varphi(f)\in\{0,f(x_{0})\} . Proof. Similar to [ 10 ] we can show that for each real-valued f\in A with \varphi(f)\neq 0 we have \varphi(1+if)=1+i\varphi(f) and \varphi(e^{f})=e^{\varphi(f)} . To do this, first we can choose by hypotheses \alpha,\beta\in f(X) such that \varphi(f)(1+i\alpha)=\varphi(f)\varphi(1+if)=\varphi(f(1+if))=\beta(1+i\beta). Hence \varphi(f)=\beta and \varphi(f)\alpha=\beta^{2} , and consequently \varphi(f)=\alpha . Thus \varphi(1+if)=(1+i\varphi(f)) . On the other hand, we can also choose a scalar \gamma\in f(X) such that e^{\gamma}(1+i\gamma)=\varphi(e^{f}(1+if))=\varphi(e^{f})(1+i\varphi(f)) which concludes that \varphi(e^{f})=e^{\gamma} and \varphi(f)=\gamma , that is \varphi(e^{f})=e^{\varphi(f)} . Since \varphi(e^{f})=e^{f(x_{0})} and f is real-valued we have \varphi(f)=f(x_{0}) . ∎ Remark 3.4 . If A is a natural uniform algebra on X , then we can use Lemma 2.2 to see that a multiplicative spectral function \varphi\colon A\to\mathbb{C} is linear (equivalently \varphi=\varphi_{x_{0}} on A , where x_{0}\in X is the associated point of \varphi ) if and only if |\varphi(h)|=1 for all h\in P_{x_{0}}(A) . The authors do not aware if (under the continuity assumption on \varphi ) this equality holds for all peaking functions or not. However, for some family of peaking functions h\in P_{x_{0}}(A) , we may have \varphi(h)\in\{0,1\} (see Corollary 3.2 and Lemma 3.3 ). For another example, it is easy to see that if h\in P_{x_{0}}(A) such that 1 is not a limit point of h(X) , then \varphi(h)\in\{0,1\} . Fo this, let n\in\mathbb{N} . Then since 1+\frac{n-1}{n}h\in{\rm exp}(A) we can choose a point x_{n}\in X such that \varphi(h)\Big(1+\frac{n-1}{n}\Big)^{n}=\varphi\Big(h\big(1+\frac{n-1}{n}h\big)^{n}\Big)=h(x_{n})\Big(1+\frac{n-1}{n}h(x_{n})\Big)^{n}. Hence for each n\in\mathbb{N} we have \varphi(h)=h(x_{n})\Big(\frac{n+(n-1)h(x_{n})}{2n-1}\Big)^{n}. Passing through a subsequence, there exists a point c\in h(X) such that h(x_{n})\to c . If c=1 , then since 1 is not a limit point of h(X) we get h(x_{n})=1 for all sufficiently large n\in\mathbb{N} . In particular \varphi(h)=1 . If c\neq 1 , then there exists r\in(0,1) such that |h(x_{n})|\leq r for sufficiently large n\in\mathbb{N} . Choosing s\in(\frac{r+1}{2},1) it is easy to see that \frac{n+(n-1)r}{2n-1}0 , the function h_{0}=\alpha\sum_{n=1}^{\infty}\frac{f_{n}}{2^{n}\|f_{n}\|} is a (positive) peaking function of A at the point x_{0} . This implies that each point in X is a peak point of A and consequently {\rm ch}(A)=X . Using Proposition 3.1 we obtain a point x_{0}\in X such that \varphi(e^{f})=e^{f(x_{0})} for all f\in A , |\varphi(f)|\leq|f(x_{0})| for all f\in A and |\varphi(f)|=|f(x_{0})| whenever f\in G(A) . We also conclude that either {\rm ker}(\varphi) is a maximal ideal of A or 1\in{\rm span}({\rm ker}(\varphi)) . Clearly if f\in A and f(x_{0})=0 , then \varphi(f)=0=f(x_{0}) . As in the proof of Lemma 3.6 we have \varphi(g)=g(x_{0}) for all g\in A_{+} . Hence for each f\in A with \overline{f}\in A , we have again |\varphi(f)|=|f(x_{0})| . Indeed, assuming that f(x_{0})=1 , since f\overline{f}\in A_{+} we have 1=f(x_{0})\overline{f(x_{0})}=\varphi(f\overline{f})=\varphi(f)\varphi(\overline{f}) and consequently |\varphi(f)|=1 . In the sequel we show that for each f\in A with |\varphi(f)|=|f(x_{0})| we have \varphi(f)=f(x_{0}) , which clearly concludes the desired equality in both cases. For this let f\in A and |\varphi(f)|=|f(x_{0})| . We may assume that f(x_{0})=1 . Then \varphi(f)=e^{i\theta} for some \theta\in[0,2\pi) . Now for each n\in\mathbb{N} , choose, by assumption, h_{n}\in A with values in [0,1] such that h_{n}=1 on \{x\in X\colon|f(x)|\leq 1\} and h_{n}=0 on {\{x\in X\colon|f(x)|\geq 1+\frac{1}{n}\}} . By the above argument, x_{0} is a peak point of A and the corresponding peaking function h_{0} can be chosen to be nonnegative. In particular, we have \varphi(h_{0})=h_{0}(x_{0})=1 . Setting g_{n}=h_{0}h_{n} for n\in\mathbb{N} , we have \varphi(g_{n})=g_{n}(x_{0})=1 , since g_{n} is also nonnegative. Now it follows from the hypotheses that e^{i\theta}=\varphi(g_{n})\varphi(f)=\varphi(g_{n}f)=g_{n}(x_{n})f(x_{n}) for some x_{n}\in X . Thus e^{i\theta}=g_{n}(x_{n})f(x_{n})\qquad(n\in\mathbb{N}). (3.1) In the case that there exists n_{0}\in\mathbb{N} such that |f(x_{n_{0}})|\leq 1 , we get e^{i\theta}=g_{n_{0}}(x_{n_{0}})f(x_{n_{0}})=h_{0}(x_{n_{0}})f(x_{n_{0}}), which conclude that |h_{0}(x_{n_{0}})f(x_{n_{0}})|=1 , that is |h_{0}(x_{n_{0}})|=1=|f(x_{n_{0}})| . In particular, x_{n_{0}}=x_{0} , since x_{0} is a peak point of A with the peaking function h_{0} . Thus in this case we have \varphi(f)=e^{i\theta}=g_{n_{0}}(x_{n_{0}})f(x_{n_{0}})=g_{n_{0}}(x_{0})f(x_{0})=1, as desired. So we assume that |f(x_{n})|>1 for all n\in\mathbb{N} . Clearly ( 3.1 ) shows that for each n\in\mathbb{N} we have 1<|f(x_{n})|<1+\frac{1}{n} . Passing through a subsequence we may assume that \{x_{n}\} converges to a point x\in X . Since e^{i\theta}=h_{0}(x_{n})h_{n}(x_{n})f(x_{n}) for each n\in\mathbb{N} , we have 1=|h_{0}(x_{n})|\,|h_{n}(x_{n})|\,|f(x_{n})|\leq|h_{0}(x_{n})|(1+\frac{1}{n}). Tending n to infinity it follows that 1\leq|h_{0}(x)| , that is |h_{0}(x)|=1 and consequently x=x_{0} . Hence, by ( 3.1 ) we get \lim g_{n}(x_{n})=e^{i\theta} and since for each n\in\mathbb{N} we have g_{n}\geq 0 , it follows that \theta=0 . Thus \varphi(f)=1=f(x_{0}) , as desired. ∎ Corollary 3.11 . Let \varphi\colon C^{1}([0,1])\to\mathbb{C} be a multiplicative spectral function. Then the following statements are equivalent: (i) \varphi is a character. (ii) \varphi is continuous. (iii) 1\notin{\rm span}({\rm ker}(\varphi) . Acknowledgement. This work is based upon research funded by Iran National Science Foundation (INSF) under project No. 4031504.