1 Introduction

Let G be a locally compact abelian group with a fixed left Haar measure and let \(L^1(G)\) be the group algebra of G defined as in [4] equipped with the convolution product \(*\) and the norm \(\Vert .\Vert _1\). We denote by \(L_0^\infty (G)\) the subspace of all functions \(f\in L^\infty (G)\), the usual Lebesgue space as defined in [4] equipped with the essential supremum norm \(\Vert .\Vert _\infty \), that for each \(\varepsilon > 0\), there is a compact subset K of G for which

$$\begin{aligned} \Vert f\chi _{G\setminus K}\Vert _\infty < \varepsilon , \end{aligned}$$

where \(\chi _{G\setminus K}\) denotes the characteristic function \(G\setminus K\) on G. It is well-known from [6] that the subspace \(L_0^\infty (G)\) is a topologically introverted subspace of \(L^\infty (G)\), that is, for each \(n\in L_0^\infty (G)^*\) and \(f\in L_0^\infty (G)\), the function \(nf\in L_0^\infty (G)\), where

$$\begin{aligned} \langle nf, \phi \rangle = \langle n, f\phi \rangle ,\quad \hbox {in which}\quad \langle f\phi , \psi \rangle = \langle f, \phi *\psi \rangle \end{aligned}$$

for all \(\phi , \psi \in L^1(G)\). Hence \(L_0^\infty (G)^*\) is a Banach algebra with the first Arens product “\(\cdot \)” defined by the formula

$$\begin{aligned} \langle m\cdot n, f\rangle =\langle m, nf\rangle \end{aligned}$$

for all \(m, n\in L_0^\infty (G)^*\) and \(f\in L_0^\infty (G)\). Note that \(L^1(G)\) may be regarded as a subspace of \(L_0^\infty (G)^*\) and then \(L^1(G)\) is a closed ideal in \(L_0^\infty (G)^*\) with a bounded approximate identity [6]. Let \(\Lambda _0(G)\) denote the set of all weak\(^*\)-cluster points of an approximate identity in \(L^1(G)\) bounded by one. It is easy to see that if \(u\in \Lambda _0(G)\), then for every \(m\in L_0^\infty (G)^*\) and \(\phi \in L^1(G)\)

$$\begin{aligned} m\cdot u=m\quad \hbox {and}\quad u\cdot \phi =\phi . \end{aligned}$$

Let \(\pi \) denote the natural continuous operator that associates to any functional in \(L_0^\infty (G)^*\) its restriction to \(C_0(G)\), the space of all continuous functions on G vanishing at infinity. Then the restriction map \(\pi \) from \(L_0^\infty (G)^*\) into M(G), the measure algebra of G as defined in [4] endowed with the convolution product \(*\) and the total variation norm, is a homomorphism and

$$\begin{aligned} \pi _u:=\pi |_{u\cdot L_0^\infty (G)^*} \end{aligned}$$

is an isomorphism for all \(u\in \Lambda _0(G)\). Note that, for every \(f\in L_0^\infty (G)\) and \(\phi \in L^1(G)\), we have \(f\phi \in C_0(G)\). Hence for every \(n\in L_0^\infty (G)^*\) and \(f\in L_0^\infty (G)\), we may define the function \(\pi (n)f\in L^\infty (G)\) by

$$\begin{aligned} \langle \pi (n)f, \phi \rangle =\langle \pi (n), f\phi \rangle . \end{aligned}$$

Then

$$\begin{aligned} \pi (n)f=nf\in L_0^\infty (G). \end{aligned}$$

This enable us to define the functional \(m\cdot \pi (n)\in L_0^\infty (G)^*\) by

$$\begin{aligned} \langle m\cdot \pi (n), f\rangle =\langle m, \pi (n)f\rangle . \end{aligned}$$

It follows that

$$\begin{aligned} m\cdot \pi (n)=m\cdot n \end{aligned}$$

for all \(m, n\in L_0^\infty (G)^*\); see [6]. Let \(\hbox {Ann}_r(L_0^\infty (G)^*)\) denote the right annihilator of \(L_0^\infty (G)^*\); i.e. the set of all \(r\in L_0^\infty (G)^*\) such that \(m\cdot r=0\) for all \(m\in L_0^\infty (G)^*\). one can easily prove that

$$\begin{aligned} \hbox {Ann}_r(L_0^\infty (G)^*)= \hbox {ker}(\pi ). \end{aligned}$$

Furthermore, an easy application of the Hahn-Banach theorem shows that G is discrete if and only if

$$\begin{aligned} \hbox {Ann}_r(L_0^\infty (G)^*)=\{0\}. \end{aligned}$$

Let \(\mathfrak {A}\) be a Banach algebra; a linear mapping \(d: \mathfrak {A}\rightarrow \mathfrak {A}\) is called a derivation if

$$\begin{aligned} d(ab)=d(a)b+ad(b). \end{aligned}$$

A fundamental question for derivations concerns their image. Singer and Wermer [12] showed that the range of a continuous derivation on a commutative Banach algebra is contained in the radical of algebra. They conjectured that this result holds for discontinuous derivations. Thomas [13] proved this conjecture. Posner [10] gave a noncommutative version of the Singer-Wermer theorem for prime rings. He proved that the zero map is the only centralizing derivation on a noncommutative prime ring (Posner’s second theorem). These results have been extended in various directions by several authors; see for instance [1, 3, 5, 7, 8, 11, 14].

Can we apply the well-known results concerning derivations of commutative Banach algebras and derivations of prime rings to \(L_0^\infty (G)^*\)? This question seems natural, because \(L_0^\infty (G)^*\) is neither a commutative Banach algebra nor a prime ring, when G is a non-discrete group. In this paper, we investigate the truth of these results for \(L_0^\infty (G)^*\).

This paper is organized as follows: In Sect. 2, we investigate the Singer- Wermer conjecture and automatic continuity for \(L_0^\infty (G)^*\). We prove that the range of a derivation on the noncommutative Banach algebra \(L_0^\infty (G)^*\) is contained in the radical of \(L_0^\infty (G)^*\) and a derivation on \(L_0^\infty (G)^*\) is continuous if and only if its restriction to \(\hbox {Ann}_r(L_0^\infty (G)^*)\) is continuous. In Sect. 3, we investigate Posner’s second theorem and show that the zero map is the only centralizing derivation on \(L_0^\infty (G)^*\). In Sect. 4, we characterize the space of all inner derivations of \(L_0^\infty (G)^*\) and prove that G is discrete if and only if any inner derivation on \(L_0^\infty (G)^*\) is zero.

2 The Singer-Wermer conjecture for \(L_0^\infty (G)^*\)

We commence this section with the following result.

Theorem 1

Let G be a locally compact abelian group and d be a derivation on \(L_0^\infty (G)^*\). Then d has its image in the right annihilator of \(L_0^\infty (G)^*\).

Proof

Let \(u\in \Lambda _0(G)\). Define the function \(D: M(G)\rightarrow M(G)\) by

$$\begin{aligned} D(\mu )=\pi \circ \tilde{d}\circ \pi ^{-1}_u(\mu ), \end{aligned}$$

where \(\tilde{d}=d|_{u\cdot L_0^\infty (G)^*}\). It is routine to check that D is derivation on the commutative semisimple Banach algebra M(G). Hence D is zero. It follows that

$$\begin{aligned} \tilde{d}\circ \pi ^{-1}_u(M(G))\subseteq \hbox {ker}(\pi )=\hbox {Ann}_r(L_0^\infty (G)^*). \end{aligned}$$

Since \(\pi _u\) maps \(u\cdot L_0^\infty (G)^*\) onto M(G), we have

$$\begin{aligned} d(u\cdot L_0^\infty (G)^*)\subseteq \hbox {Ann}_r(L_0^\infty (G)^*). \end{aligned}$$

On the one hand,

$$\begin{aligned} m\cdot d(r)=d(m\cdot r)-d(m)\cdot r=0 \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\) and \(r\in \hbox {Ann}_r(L_0^\infty (G)^*)\). So

$$\begin{aligned} d(\hbox {Ann}_r(L_0^\infty (G)^*))\subseteq \hbox {Ann}_r(L_0^\infty (G)^*). \end{aligned}$$

Now, we only need to recall that \(L_0^\infty (G)^*\) is the Banach space direct sum of \(u\cdot L_0^\infty (G)^*\) and \(\hbox {Ann}_r(L_0^\infty (G)^*)\). \(\square \)

Before we give the following consequence of Theorem 1, let us recall that a linear mapping T on \(L_0^\infty (G)^*\) is called spectrally bounded if there is a non-negative number \(\alpha \) such that \(r(T(m))\le \alpha r(m)\) for all \(m\in L_0^\infty (G)^*\), where \(r(\cdot )\) stands for the spectral radius.

Corollary 1

Let G be a locally compact abelian group. Then the following statements hold.

  1. (i)

    Every derivation on \(L_0^\infty (G)^*\) maps it into its radical.

  2. (ii)

    Primitive ideals of \(L_0^\infty (G)^*\) are invariant under derivations on \(L_0^\infty (G)^*\).

  3. (iii)

    Every derivation on \(L_0^\infty (G)^*\) is spectrally bounded.

  4. (iv)

    The composition of two derivations on \(L_0^\infty (G)^*\) is always a derivation on \(L_0^\infty (G)^*\).

Proof

The statement (i) follows from Theorem 1 together with the fact that the set of nilpotent elements is contained in the radical of the algebra. The statement (ii) follows immediately from (i). For (iii), note that if d is a derivation on \(L_0^\infty (G)^*\), then

$$\begin{aligned} d(m)^i=0 \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\) and \(i\ge 2\). Finally, the statement (iv) follows from Theorem 1. \(\square \)

As an another consequence of Theorem 1, we have the following result.

Corollary 2

Let G be a locally compact abelian group. Then the following statements hold.

  1. (i)

    If d is a derivation on \(L_0^\infty (G)^*\), then \(d|_{L^1(G)}\) is zero.

  2. (ii)

    The zero map is the only \(\hbox {weak}^{*}-\hbox {weak}^{*}\) continuous derivation on \(L_0^\infty (G)^*\).

Proof

First note that

$$\begin{aligned} r\cdot \phi =\phi \cdot r=0 \end{aligned}$$

for all \(r\in \hbox {Ann}_r(L_0^\infty (G)^*)\) and \(\phi \in L^1(G)\). So if d is a derivation on \(L_0^\infty (G)^*\), then

$$\begin{aligned} d(\phi _1*\phi _2)=d(\phi _1)\cdot \phi _2+\phi _1\cdot d(\phi _2)=0 \end{aligned}$$

for all \(\phi _1,\phi _2\in L^1(G)\). In view of Cohen’s factorization theorem, \(d=0\) on \(L^1(G)\). So (i) holds. The statement (ii) follows from Goldstein’s theorem (see e.g. [2, chapter 5, Proposition 4.1]) and (i). \(\square \)

Theorem 2

Let G be a locally compact abelian group and d be a derivation on \(L_0^\infty (G)^*\). Then the following statements hold.

  1. (i)

    For every \(u\in \Lambda _0(G)\), \(d|_{u\cdot L_0^\infty (G)^*}\) is always continuous.

  2. (ii)

    d is continuous if and only if \(d|_{\hbox {Ann}_r(L_0^\infty (G)^*)}\) is continuous.

Proof

(i) Let \(u\in \Lambda _0(G)\) and \((u\cdot m_\alpha )_{\alpha \in A}\) be a net in \(L_0^\infty (G)^*\) such that \(u\cdot m_\alpha \rightarrow 0\). It follows from Theorem 1 that

$$\begin{aligned} \Vert d(u\cdot m_\alpha )\Vert= & {} \Vert d(u\cdot u\cdot m_\alpha )\Vert \\= & {} \Vert d(u)\cdot u\cdot m_\alpha \Vert \le \Vert d(u)\Vert \;\Vert u\cdot m_\alpha \Vert \end{aligned}$$

for all \(\alpha \in A\). Hence \(d(u\cdot m_\alpha )\rightarrow 0\). This shows that \(d|_{u\cdot L_0^\infty (G)^*}\) is continuous. (ii) Let \(m\in L_0^\infty (G)^*\) and \(u\in \Lambda _0(G)\). Then \(m-u\cdot m\) is an element of \(\hbox {Ann}_r(L_0^\infty (G)^*)\). If \(d|_{\hbox {Ann}_r(L_0^\infty (G)^*)}\) is continuous, then for some \(\alpha >0\)

$$\begin{aligned} \Vert d(m-u\cdot m)\Vert \le \alpha \Vert m-u\cdot m\Vert \le 2\;\alpha \Vert m\Vert . \end{aligned}$$

By (i) there exists \(\beta >0\) such that

$$\begin{aligned} \Vert d(u\cdot m)\Vert \le \beta \;\Vert u\cdot m\Vert \le \beta \;\Vert m\Vert . \end{aligned}$$

Thus

$$\begin{aligned} \Vert d(m)\Vert =\Vert d(u\cdot m)+d(m-u\cdot m)\Vert \le (2\alpha +\beta )\Vert m\Vert . \end{aligned}$$

It follows that d is continuous. \(\square \)

Our last result of this section is an immediate consequence of Theorem 2(ii).

Corollary 3

Let G be a discrete abelian locally compact group.Then every derivation on \(L_0^\infty (G)^*\) is continuous.

3 Posner’s second theorem for \(L_0^\infty (G)^*\)

Let \(\hbox {Z} (L_0^\infty (G)^*)\) denote the center of \(L_0^\infty (G)^*\); that is, the set of all \(m\in L_0^\infty (G)^*\) such that \(m\cdot n=n\cdot m\) for all \(n\in L_0^\infty (G)^*\).

Proposition 1

Let G be a locally compact abelian group. Then

$$\begin{aligned} Z (L_0^\infty (G)^*)=L^1(G). \end{aligned}$$

Proof

Let \(u\in \Lambda _0(G)\). Since \(L^1(G)\) is an ideal in \(L_0^\infty (G)^*\) and \(\pi \) is identity on \(L^1(G)\), we have

$$\begin{aligned} \phi \cdot m= & {} \pi (\phi \cdot m)=\pi (\phi )*\pi (m)\\= & {} \pi (m)*\pi (\phi )=\pi (m\cdot \phi )=m\cdot \phi \end{aligned}$$

for all \(\phi \in L^1(G)\) and \(m\in L_0^\infty (G)^*\). So \(L^1(G)\) is contained in \(\hbox {Z}(L_0^\infty (G)^*)\). For \(m\in \hbox {Z}(L_0^\infty (G)^*)\), we have

$$\begin{aligned} m=m\cdot u=u\cdot m. \end{aligned}$$

This shows that

$$\begin{aligned} m\in \cap _{u\in \Lambda _0(G)}u\cdot L_0^\infty (G)^*. \end{aligned}$$

Hence \(\hbox {Z} (L_0^\infty (G)^*)\) is contained in \(L^1(G)\); see Theorem 2.11 of [6]. \(\square \)

For any positive integer k, a mapping \(T: L_0^\infty (G)^*\rightarrow L_0^\infty (G)^*\) is called k -centralizing if

$$\begin{aligned}{}[T(m), m^k]\in Z(L_0^\infty (G)^*) \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\); in a special case when \([T(m),m^k]=0\) for all \(m\in L_0^\infty (G)^*\), T is called k-commuting, where \([m,n]:=m\cdot n-n\cdot m\) for all \(m,n\in L_0^\infty (G)^*\).

Theorem 3

Let G be a locally compact abelian group, d be a derivation on \( L_0^\infty (G)^*\) and k be a positive integer. Then the following assertions are equivalent.

  1. (a)

    \(d=0\).

  2. (b)

    d is k-centralizing.

  3. (c)

    d is k-commuting.

Proof

It is clear that (a) implies (b). If (b) holds, then by Theorem 1 and Proposition 1, we obtain

$$\begin{aligned}{}[d(m),m^k]= & {} d(m)\cdot m^k\\= & {} d(m^{k+1})\in \hbox {Ann}_r( L_0^\infty (G)^*)\cap L^1(G)=\{0\} \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\). Thus (c) holds. Now, let d be k-commuting. Choose \(u\in \Lambda _0(G)\). Then

$$\begin{aligned} d(u)=[d(u),u]=[d(u),u^k]=0. \end{aligned}$$
(1)

For every \(r\in \hbox {Ann}_r(L_0^\infty (G)^*)\), we have \( (r+u)=(r+u)^k. \) Hence

$$\begin{aligned} d(r)=[d(r+u),(r+u)]=[d(r+u),(r+u)^k]=0. \end{aligned}$$
(2)

From (1) and (2) we infer that

$$\begin{aligned} d(m)= & {} d(u\cdot m)+d(m-u\cdot m)\\= & {} d(u)\cdot m+d(m-u\cdot m)\\= & {} 0 \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\). Thus (c) implies that (a). \(\square \)

As an immediate consequence from Theorem 3, we have the following result.

Corollary 4

Let G be a locally compact abelian group. Then the zero map is the only centralizing derivation on \(L_0^\infty (G)^*\).

Let \([m,n]_1=[m, n]\) and \([m, n]_k=[[m, n]_{k-1}, n]\) for all \(m,n\in L_0^\infty (G)^*\) and all positive integers \(k>1\).

Corollary 5

Let G be a locally compact abelian group and d be a derivation on \(L_0^\infty (G)^*\). Then the following assertions are equivalent.

  1. (a)

    \(d=0\).

  2. (b)

    d is centralizing.

  3. (c)

    For every \(k\in {\mathbb {N}}\), d is k-centralizing.

  4. (d)

    There exists \(k\in {\mathbb {N}}\) such that d is k-centralizing.

  5. (e)

    There exist positive integers kl such that \(l\ge 2\) and \([d(m), n]_k=[m, n]^l\) for all \(m, n\in L_0^\infty (G)^*\)

Proof

This follows from Theorem 3 with the observation that for every \(m, n\in L_0^\infty (G)^*\), we have \([m, n]\in \hbox {Ann}_r(L_0^\infty (G)^*)\) and so \([m, n]^l=0\) for all \(l\ge 2\).\( \square \)

We conclude the section with the following result.

Theorem 4

Let G be a locally compact abelian group and d be a derivation on \(L_0^\infty (G)^*\). Then the following assertions are equivalent.

  1. (a)

    G is discrete.

  2. (b)

    \(L_0^\infty (G)^*\) is commutative.

  3. (c)

    There exist \(i, j, k\in {\mathbb {N}}\) such that \([d(m), n]_i^j=[m, n]_k\) for all \(m, n\in L_0^\infty (G)^*\).

In this case, \(d=0\).

Proof

If G is discrete, then by Proposition 3.1 of [9], we have \(L_0^\infty (G)^*=L^1(G)\). Since G is an abelian, \(L_0^\infty (G)^*\) is commutative. Thus (a) implies (b). It is clear that (b) implies (c) and (d). Now, let \(i, j, k\in {\mathbb {N}}\) and

$$\begin{aligned} d(m\cdot n^i)^j=[m, n]\cdot n^k. \end{aligned}$$

Then for every \(u\in \Lambda _0(G)\), we have

$$\begin{aligned} d(u)^j=d(u\cdot u^i)^j=[u, u]\cdot u^k=0. \end{aligned}$$

On the one hand, for every \(r\in \hbox {Ann}_r(L_0^\infty (G)^*)\), we get

$$\begin{aligned} d(u)^j= & {} d(u\cdot (u+r))^j=d(u\cdot (u+r)^i)^j\\= & {} [u, u+r]\cdot (u+r)^k=[u, u+r]\cdot (u+r)=-r. \end{aligned}$$

Hence

$$\begin{aligned} \hbox {Ann}_r(L_0^\infty (G)^*)=\{0\}, \end{aligned}$$

which implies that G is discrete. To complete the proof, it suffices to notice that the assertion (c) implies that \(d(m\cdot n^i)^j=[m, n]\cdot n^{k-1}\). \(\square \)

4 Inner derivations of \(L_0^\infty (G)^*\)

A derivation d on \(L_0^\infty (G)^*\) is said to be inner if there exists \(n_0\in L_0^\infty (G)^*\) such that \(d(m)=[m,n_0]\) for all \(m\in L_0^\infty (G)^*\).

Proposition 2

Let G be a locally compact abelian group and d be a derivation on \(L_0^\infty (G)^*\). Then the following assertions are equivalent.

  1. (a)

    d is inner.

  2. (b)

    There exists \(n_0\in L_0^\infty (G)^*\) such that for each \(k\in {\mathbb {N}}\) the mapping \(m\mapsto d(m)+n_0\cdot m\) is k-commuting.

  3. (c)

    There exists \(n_0\in L_0^\infty (G)^*\) and \(k\in {\mathbb {N}}\) such that the mapping \(m\mapsto d(m)+n_0\cdot m\) is k-commuting.

  4. (d)

    There exists \(n_0\in L_0^\infty (G)^*\) and \(k\in {\mathbb {N}}\) such that the mapping \(m\mapsto d(m)+n_0\cdot m\) is k-centralizing.

Proof

Let there exist \(n_0\in L_0^\infty (G)^*\) such that \(d(m)=[m,n_0]\) for all \(m\in L_0^\infty (G)^*\). For \(k\in {\mathbb {N}}\) and \(m\in L_0^\infty (G)^*\), we obtain

$$\begin{aligned} m^k\cdot n_0\cdot m= & {} m^k\cdot \pi (n_0\cdot m)=m^k\cdot \pi (n_0)*\pi (m)\\= & {} m^k\cdot \pi (m)*\pi (n_0)=m^{k+1}\cdot n_0. \end{aligned}$$

It follows that

$$\begin{aligned}{}[d(m)+n_0\cdot m, m^k]= & {} d(m)\cdot m^k+n_0\cdot m^{k+1}-m^k\cdot n_0\cdot m\\= & {} d(m^{k+1})+n_0\cdot m^{k+1}-m^{k+1}\cdot n_0\\= & {} 0. \end{aligned}$$

Hence (a) implies (b). It is obvious that (b)\(\Rightarrow \)(c)\(\Rightarrow \) (d). To complete the proof, let (d) hold. Define the function \(D:L_0^\infty (G)^*\rightarrow L_0^\infty (G)^*\) by

$$\begin{aligned} D(m)=d(m)-[m,n_0]. \end{aligned}$$

It is clear that D is a derivation on \(L_0^\infty (G)^*\). So

$$\begin{aligned}{}[D(m), m^k]=D(m)\cdot m^k=[d(m)+n_0 m, m^k]\in \hbox {Z} (L_0^\infty (G)^*). \end{aligned}$$

We now invoke Corollary 4 to conclude that \(D=0\). So, we obtain (a). \(\square \)

In the sequel, let \(\hbox {InnD}(L_0^\infty (G)^*)\) be the space of all inner derivations on \(L_0^\infty (G)^*\).

Theorem 5

Let G be a locally compact abelian group. Then \(InnD (L_0^\infty (G)^*)\) is continuously linearly isomorphic to \(L_0^\infty (G)^*/L^1(G)\).

Proof

We define the mapping \(\mathfrak {I}\) from \(L_0^\infty (G)^*/L^1(G)\) into \(\hbox {InnD}(L_0^\infty (G)^*)\) by

$$\begin{aligned} \mathfrak {I}(m+L^1(G))=\mathfrak {I}_m, \end{aligned}$$

where \(\mathfrak {I}_m(n)=[n, m]\) for all \(n\in L_0^\infty (G)^*\). By Proposition 1, the mapping \(\mathfrak {I}\) is well defined. Obviously, \(\mathfrak {I}\) is a linear map from \(L_0^\infty (G)^*/L^1(G)\) onto \(\hbox {InnD}(L_0^\infty (G)^*)\). To see that \(\mathfrak {I}\) is injective, let \(m\in L_0^\infty (G)^*\) and

$$\begin{aligned} \mathfrak {I} (m+L^1(G))=0. \end{aligned}$$

Then

$$\begin{aligned} \mathfrak {I}_m(n)=n\cdot m-m\cdot n=0 \end{aligned}$$

for all \(n\in L_0^\infty (G)^*\). It follows that

$$\begin{aligned} m\in \hbox {Z}(L_0^\infty (G)^*)=L^1(G). \end{aligned}$$

Hence \(m+L^1(G)=L^1(G)\). Consequently, \(\mathfrak {I}\) is an isomorphism. Now, let \(n\in L_0^\infty (G)^*\) and \(\phi \in L^1(G)\). Then

$$\begin{aligned} \Vert \mathfrak {I}_m(n)\Vert= & {} \Vert n\cdot m-m\cdot n\Vert \\\le & {} \Vert n\cdot m-\phi \cdot n\Vert +\Vert \phi \cdot n-m\cdot n\Vert \\\le & {} \Vert n\Vert \;\Vert m-\phi \Vert +\Vert \phi -m\Vert \;\Vert n\Vert \\= & {} 2\Vert n\Vert \;\Vert m-\phi \Vert \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\). This implies that

$$\begin{aligned} \Vert \mathfrak {I} (m+L^1(G))\Vert =\Vert \mathfrak {I}_m\Vert \le 2\Vert m-\phi \Vert \end{aligned}$$

for all \(m\in L_0^\infty (G)^*\) and \(\phi \in L^1(G)\). Hence

$$\begin{aligned} \Vert \mathfrak {I} (m+L^1(G))\Vert\le & {} 2\inf \{\Vert m-\phi \Vert : \phi \in L^1(G)\}\\= & {} 2\inf \{\Vert m+\phi \Vert : \phi \in L^1(G)\}=2\;\Vert m+L^1(G)\Vert . \end{aligned}$$

Therefore, \(\mathfrak {I}\) is continuous. \(\square \)

We finish the paper with following result.

Theorem 6

Let G be a locally compact abelian group. Then the following assertions are equivalent.

  1. (a)

    G is discrete.

  2. (b)

    Any derivation on \(L_0^\infty (G)^*\) is zero.

  3. (c)

    Any inner derivation on \(L_0^\infty (G)^*\) is zero.

Proof

If G is discrete, then \(\hbox {Ann}_r(L_0^\infty (G)^*)=\{ 0\}\). By Theorem 1,

$$\begin{aligned} d(L_0^\infty (G)^*)\subseteq \hbox {Ann}_r(L_0^\infty (G)^*)=\{ 0\}. \end{aligned}$$

Hence (a) implies (b). It is plain that (b) implies (c). Finally, if (c) holds, then \([m,n]=0\) for all \(m,n\in L_0^\infty (G)^*\). This implies that

$$\begin{aligned} \hbox {Z} (L_0^\infty (G)^*)=L_0^\infty (G)^*. \end{aligned}$$

So \(L^1(G)=L_0^\infty (G)^*\). This shows that G is discrete; see Proposition 3.1 of [9]. \(\square \)