Submitted:
27 October 2026
Posted:
28 October 2026
You are already at the latest version
Abstract
Let A be a unital JB*-algebra with unit 1, Jordan product ∘ and involution \( * \). Since the commutator formed from the Jordan product vanishes, we measure deviations by the Jordan associator {x, w} := (x ∘ w) ∘ w − x ∘ (w ∘ w). Let S(A) consist of those z ∈ A for which {z, a} = 0 for every a ∈ A. A mapping δ : A→ A is called an S(A)-extended Jordan \( * \)-derivation if additivity, the Jordan derivation identity and the involution condition hold modulo S(A). We assume that A ≠ C1, S(A) = C1, and the associator controls the distance to S(A); these conditions hold for Mn(C), n ≥ 2. Under dyadic summability and decay assumptions on the errors in a three-variable functional inequality and in the product and involution conditions, we obtain a unique S(A)-extended Jordan \( * \)-derivation L(x) = limn→∞2nδ(x/2n); an analogous result holds for limn→∞2−nδ(2nx). These limits exist in norm, whereas the quantitative approximation estimates control the associator error and the distance of L(x) − δ(x) to C1. The limit is additive and exactly \( * \)-preserving. An additional one-variable decay condition yields δ = L (superstability), and a unimodular version of the functional inequality yields C-linearity. Matrix examples demonstrate a nonzero limit and the distinction between associator control and full norm control.
Keywords:
JB*-algebra
; Jordan associator
; S(A)-extended Jordan ∗-derivation
; Hyers–Ulam stability
; superstability
; functional inequality
MSC: 17C65; 39B72; 39B82; 46K70
1. Introduction
The stability problem for functional equations originates in a question posed by Ulam [1]: is every approximately additive map on a Banach space close to an exactly additive map? Hyers [2] answered affirmatively, initiating a vast literature; see [3,4,5] and the references therein. The related phenomenon of superstability, where every approximate solution is automatically an exact solution, was first observed by Baker, Lawrence and Zorzitto [6] for the exponential equation and later established for derivation-type maps by Šemrl [7] and Badora [8]; see also [9].
Centrally-extended derivations on rings were introduced by Bell and Daif [10], and their involutive analogues by El-Deken and Nabiel [11]. The Hyers–Ulam stability and superstability of centrally-extended derivations on Banach algebras and Banach *-algebras were studied in [12,13], where the functional inequality
is analyzed by means of the commutator .
The present paper extends this program to -algebras. Since the Jordan product is commutative, the commutator formed from that product cannot serve as a deviation measure. We replace it by the Jordan associator. For the Jordan product of a -algebra one has (Lemma 3), so the associator is a second-order commutator, and its kernel contains the center; for the matrix algebras it is exactly (Proposition 1).
Transferring the associative arguments to this setting requires three adjustments, which we now describe.
(a) The norm-control condition. A condition of the form cannot hold on a unital algebra, since for every a. In the spirit of Stampfli’s formula [15], the natural condition is to control the distance to :
This condition holds in every finite-dimensional -algebra (Proposition 2).
(b) Recovering the scalar part. The associator does not see elements of , so the inequality above only gives convergence of the approximating sequence modulo. Under and , the space is not a Jordan ideal: otherwise would belong to for every . Thus the quotient is a Banach space but does not inherit the Jordan product, so convergence in the quotient alone is not enough to pass to the limit in the product condition. We show that the product inequality itself forces the scalar part of the sequence to converge (Lemma 5). This gives genuine norm convergence and makes the limit unique. The quantitative estimates nevertheless control the associator error and distance to , not the full norm of ; see Remark 1.
(c) The form of the derivation identity. In the convention of [14], an additive map on a ring with involution is called a *-derivation if it satisfies
Its literal Jordan analogue
is not symmetric in , whereas is; we show in Proposition 3 that it forces for every self-adjoint h. We therefore use the Jordan derivation identity
together with the *-condition , each modulo (Definition 3).
The paper is organized as follows. Section 2 collects the preliminaries, including the structural lemmas that drive the proofs. Section 3 contains the Hyers–Ulam stability theorems (Theorems 1 and 2), and Section 4 the superstability results (Theorems 3 and 4, Corollary 1). Section 5 gives matrix examples illustrating the hypotheses and the scope of the estimates, and Section 6 summarizes the results.
2. Preliminaries
2.1. -Algebras and the Jordan Associator
A Jordan algebra over is a commutative algebra satisfying the Jordan identity . A -algebra is a complex Jordan algebra with a conjugate-linear involution * satisfying , which is a Banach space with and , where ; see [16,17,18]. Every -algebra is a -algebra under . In a -algebra one has , and if is unital with unit , then . In this case we write for the set of scalar multiples of the unit.
From now on, let be a -algebra with Jordan product ∘ and involution *. We write for all .
Definition 1.
For , the Jordan associator is , and
Lemma 1.
For all and :
- (i)
- ;
- (ii)
-
and ; in particular;
- (iii)
- ;
- (iv)
- .
Proof. (i) follows from the -bilinearity of ∘. (ii) follows from , and (i). (iii) follows from . (iv) follows from the bilinearity of ∘. □
Lemma 2.
is a closed, *-invariant, complex linear subspace of , and for all , . If is unital, then .
Proof.
Linearity and closedness follow from Lemma 1(i),(ii). If , then by Lemma 1(iii)
The identity follows from Lemma 1(i). Finally, . □
Thus is a continuous seminorm on whose kernel is exactly .
Lemma 3.
Let be a -algebra with the Jordan product . Then for all . Consequently, the center of is contained in .
Proof.
We have
hence for all . □
Proposition 1.
Let and let carry the operator norm and the Jordan product . Then .
Proof.
Let and let p be an orthogonal projection. By Lemma 3, we have
Multiplying by p on the left gives , and on the right gives ; hence . Since every matrix is a linear combination of orthogonal projections (spectral theorem applied to the real and imaginary parts), z commutes with all of , so . The reverse inclusion is Lemma 2. □
2.2. The Norm-Control Condition
Definition 2.
A -algebra satisfies the norm-control condition if there exists such that
Proposition 2.
Every finite-dimensional -algebra satisfies the norm-control condition (1).
Proof.
By Lemma 2, induces a seminorm on the quotient space , and if and only if . Thus is a norm on the finite-dimensional space , hence equivalent to the quotient norm . □
Assumption 1.
- (H1)
- and ;
- (H2)
- satisfies the norm-control condition (1).
By Propositions 1 and 2, with satisfies Assumption 1.
2.3. Structural Lemmas Under (H1)
Lemma 4.
Assume (H1). If satisfies for all , then .
Proof.
Taking gives , say . Then for every . Since , we get . □
By (H1) we may fix and, since , some with .
Lemma 5
(Scalar separation). Assume (H1).
- (i)
- If satisfies for all , then .
- (ii)
- Let be functions such that, for all , the sequenceconverges in . Then converges in for every .
Proof.
Put . (i) Taking , gives , so . Taking , then gives , so for every x.
(ii) Taking , shows that converges. Since
the sequence is Cauchy. Next, taking , , the sequence
converges, and the same argument shows that converges. □
Lemma 6.
Assume (H1) and (H2). Let be the dual space of . Fix with and (Hahn–Banach), and put . Then P is linear, and for all .
Proof.
For every we have , so . Taking the infimum over c and using (1) gives . □
2.4. -Extended Jordan *-Derivations
Definition 3.
A mapping is called an-extended Jordan *-derivation (-E Jordan *-derivation) if, for all ,
- (E1)
- ;
- (E2)
- ;
- (E3)
- .
The next proposition explains why does not appear in (E2).
Proposition 3.
Let be a unital -algebra and suppose that satisfies
Then for every self-adjoint .
Proof.
Since , subtracting the two relations obtained by interchanging a and b gives for all . For and this reads . □
Proposition 4
(Rigidity). Assume (H1). Every -E Jordan *-derivation is additive and satisfies
Proof.
Put . A direct computation gives
for all . The left-hand side and lie in , hence
and Lemma 4 gives . Thus L is additive and . Finally,
so by Lemma 4.
To prove exact preservation of the involution, put . By (E2) and the *-invariance of ,
By (E3), the map takes values in ; write . Subtracting the product identities modulo for and L gives
The first term is scalar, so applying the associator yields
Lemma 5(i) gives . Hence , which is equivalent to . □
2.5. The Functional Inequalities
Fix real numbers with . We consider mappings satisfying
for all , where , and . For the -linearity results we also use the unimodular version of (2):
Here and . Taking recovers (2). This formulation is compatible with nonzero -linear maps: if L is -linear and , then
and . Thus (6) holds for such a map with .
We write
and
for all .
Lemma 7.
Suppose that satisfies (5) and
for all . Then for all .
Proof.
Setting , and in (7), the right-hand side is in each case, so
Subtracting the first two identities from the third and using and Lemma 1(i), we get . Replacing x by and y by gives for all . □
Lemma 7 also holds for normed Jordan algebras.
Lemma 8.
If satisfies (2) and (5), then
Proof.
Setting equal to
in (2), the right-hand norm vanishes by (5) and we obtain
By the triangle inequality, Lemma 1(i) and ,
Replacing x by and y by gives (8). □
3. Hyers–Ulam Stability
Theorem 1.
Let satisfy Assumption 1. Assume that is finite-valued and that, for all , and ,
Suppose satisfies (2), (3), (4) and (5). Then the limit
exists in norm for every , L is an additive -E Jordan *-derivation with and for all , and
Moreover, L is the unique -E Jordan *-derivation satisfying (12). Setting
we also have, for every ,
Proof.
Step 2: convergence of the non-scalar part. With P and as in Lemma 6, write
By Lemma 6 and (13), , which tends to 0 as uniformly in , since it is bounded by a tail of the convergent series defining . Hence the limit
exists in norm for every .
where . Since and , Lemma 1(i) gives
The left-hand side tends to 0 by (10), and the first term on the right converges to
by Step 2 and Lemma 1(ii). Therefore converges for all , and Lemma 5(ii) shows that exists. Consequently
which proves (11).
Taking the supremum in (12) yields , and (H2) gives the asserted distance estimate.
The last term tends to 0 by (9), so letting shows that L satisfies (7). Moreover , so Lemma 7 gives (E1) for L. Letting in Step 3 gives
which is (E2). Finally, by (4) with ,
and since the involution is isometric, , which is (E3). Hence L is an -E Jordan *-derivation, and by Proposition 4 it is additive, vanishes at , and satisfies for every .
Thus takes values in ; write with . Subtracting (E2) for from (E2) for L gives , i.e., for all . By Lemma 5(i), , so . □
Each of the three terms of has the form for a fixed triple, e.g., for the first one. Hence, by (9), the series defining and converge.
- Step 1: Cauchy estimates. Put . Applying (8) at and multiplying by , we get . Summing over j, for ,
- Step 3: convergence of the scalar part. Fix . Applying (3) with , and multiplying by , we obtain
- Step 4: the estimates. Take in (13) and let , using and Lemma 1(ii), to obtain (12). Since the terms in the series are nonnegative,
- Step 5: L satisfies (E1)–(E3). Replacing by in (2) and multiplying by , we get
- Step 6: uniqueness. Let be another -E Jordan *-derivation satisfying (12). Both L and are additive (Proposition 4), hence and . Therefore
Theorem 2.
Let satisfy Assumption 1. Assume that is finite-valued and that, for all , and ,
Suppose satisfies (2), (3), (4) and (5). Then exists in norm for every , L is an additive -E Jordan *-derivation with and for all , and
Moreover, L is the unique -E Jordan *-derivation satisfying (16). Setting
we also have, for every ,
Proof.
Put . Then , so (8) at gives , and hence, for ,
The right-hand sides are tails of series that converge by (14). Applying (3) with , and dividing by gives with the same expression
as before. With these changes, Steps 2–5 of the proof of Theorem 1 apply verbatim, using (14) and (15) in place of (9) and (10). (In Step 5, (2) is applied at and divided by , and (4) is applied at .) For uniqueness, additivity gives and , so
and we conclude as in Step 6. □
Remark 1
(Meaning of the stability estimates). Theorems 1 and 2 give convergence of their approximating sequences in the original norm of , not merely in a quotient. Their quantitative error estimates, however, concern and the distance of to . Since for every , those estimates alone do not bound . Example 2 shows that the associator error can vanish identically while the full norm error is arbitrarily large.
4. Superstability
The following results show that, under an additional decay condition on the product control in which only the first variable is scaled, the mapping itself is an -E Jordan *-derivation. In fact it coincides with L.
Theorem 3.
Under the hypotheses of Theorem 1, assume in addition that
Then . In particular, δ is an additive -E Jordan *-derivation satisfying and for all .
Proof.
Applying (3) with , and multiplying by , we get
where . Letting and using (11), (17) and Lemma 1(ii), we obtain . Comparing this relation with (E2) for L yields
that is, for all . By Lemma 4, for every . □
Theorem 4.
Under the hypotheses of Theorem 2, assume in addition that
Then . In particular, δ is an additive -E Jordan *-derivation satisfying and for all .
Proof.
Apply (3) with , , divide by , let , and argue as in the proof of Theorem 3. □
Corollary 1.
Under the hypotheses of Theorem 1(resp. Theorem 2), suppose that δ satisfies (6) for all . Then L is -linear. If, moreover, (17)(resp. (18))holds, then is a -linear -E Jordan *-derivation.
Proof.
We treat the case of Theorem 1; the case of Theorem 2 follows by applying (6) at and dividing by . In the first case, replace by in (6), multiply by , and let . Norm convergence, (9), and give
Fix . Since L is additive, . Taking gives
The same identity with the unimodular parameter equal to 1 and x replaced by reads
Subtracting these identities and using , we obtain
Since is a complex subspace, (E2) gives . A direct computation yields
The first term on the right belongs to . Consequently for every , and Lemma 4 gives . Thus
Now let and choose an integer . Since , there exist such that . Additivity implies -linearity, hence
The case follows from . This proves -linearity without any continuity assumption on L. The final assertion follows from Theorem 3 (resp. Theorem 4). □
5. Examples
Example 1.
Let and let carry the operator norm and the Jordan product . By Propositions 1 and 2, satisfies Assumption 1. Fix self-adjoint matrices and a real number , and define
is a Jordan *-derivation. Since Δ is an associative derivation and ,
Inequality (2). Put and . Then
By Lemma 1(ii), , so (2) holds with
Here is positively homogeneous of degree , so both series in (9) are dominated by a multiple of .
The unimodular inequality (6). For , put
Since Δ is -linear and ,
Also . Applying the associator and using therefore proves (6) with the same control function Φ, uniformly for .
Inequality (3). We have . Using and Lemma 1(ii), (3) holds with
and .
Conditions (4) and (5). Since and , we have , so (4) holds with ; clearly .
By Theorem 1,
is the unique -E Jordan *-derivation satisfying (12). Indeed,
which is nonzero for suitable w when , because . Hence . Moreover, is nonzero since . Thus the unimodular hypothesis is compatible with a nonzero -linear limit.
The additional decay condition. For the chosen product control,
For nonzero this limit is nonzero. Thus the chosen control does not satisfy (17), consistently with .
Example 2
(Scalar errors are invisible to the associator). Let with its usual operator norm, Jordan product and involution. Fix and , and define
All values are scalar, so both (2) and (6) hold with . Moreover, and , so we may take . The product defect satisfies
Thus (3) holds with
For every ,
All hypotheses of Theorem 1 are therefore satisfied, and
Here and , while
For fixed , this norm becomes arbitrarily large as M increases, while the associator error and quotient distance vanish. Thus (12) does not imply a full norm estimate based on η alone.
6. Conclusions
We have established Hyers–Ulam stability (Theorems 1 and 2) and superstability (Theorems 3 and 4) for the functional inequality (2), coupled with approximate product and involution conditions, in unital -algebras with satisfying the norm-control condition (1). This extends the program of [12,13] from the associative to the Jordan setting. The main features are the following.
- (1)
- The Jordan associator replaces the commutator. For a -algebra equipped with its Jordan product, it is the second-order commutator , and its kernel contains the associative center.
- (2)
- The norm-control condition bounds rather than . It is therefore compatible with the unit, and it holds in every finite-dimensional -algebra.
- (3)
- The scalar part of the approximating sequence, which the associator cannot see, is recovered from the product inequality (Lemma 5). This yields norm convergence and the uniqueness of the limit.
- (4)
- Under (H1), every -E Jordan *-derivation is automatically additive, vanishes at the unit, and preserves the involution exactly (Proposition 4).
- (5)
- The quantitative estimates control the associator error and the distance to , not the full norm error. The unimodular inequality (6) gives -linearity and admits nonzero limits, as demonstrated in Example 1.
Author Contributions
Conceptualization, J.R.; Writing—original draft, J.R.; Writing—review & editing, J.R.; Supervision, J.R.; Funding acquisition, J.R.; The author has read and agreed to the published version of the manuscript.
Funding
This paper was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2021R1A2C109489611).
Data Availability Statement
No new data were created or analyzed in this study. Data sharing is not applicable to this article.
Conflicts of Interest
The author declares no conflicts of interest.
References
- Ulam, S.M. A Collection of Mathematical Problems; Interscience: New York, NY, USA, 1960.
- Hyers, D.H. On the stability of the linear functional equation. Proc. Natl. Acad. Sci. USA 1941, 27, 222–224. [CrossRef]
- Rassias, T.M. On the stability of the linear mapping in Banach spaces. Proc. Am. Math. Soc. 1978, 72, 297–300. [CrossRef]
- Găvruta, P. A generalization of the Hyers–Ulam–Rassias stability of approximately additive mappings. J. Math. Anal. Appl. 1994, 184, 431–436. [CrossRef]
- Gajda, Z. On stability of additive mappings. Int. J. Math. Math. Sci. 1991, 14, 431–434. [CrossRef]
- Baker, J.A.; Lawrence, J.; Zorzitto, F. The stability of the equation f(x+y)=f(x)f(y). Proc. Am. Math. Soc. 1979, 74, 242–246. [CrossRef]
- Šemrl, P. The functional equation of multiplicative derivation is superstable on standard operator algebras. Integr. Equ. Oper. Theory 1994, 18, 118–122. [CrossRef]
- Badora, R. On approximate derivations. Math. Inequal. Appl. 2006, 9, 167–173. [CrossRef]
- Chang, I.-S.; Kim, H.-M.; Roh, J. On superstability of derivations in Banach algebras. Open Math. 2025, 23, 1–9. [CrossRef]
- Bell, H.E.; Daif, M.N. On centrally-extended maps on rings. Beitr. Algebra Geom. 2016, 57, 129–136. [CrossRef]
- El-Deken, S.F.; Nabiel, H. Centrally-extended generalized *-derivations on rings with involution. Beitr. Algebra Geom. 2019, 60, 217–224. [CrossRef]
- Chang, I.-S.; Roh, J. Some centrally extended derivations on Banach algebras. Mathematics 2025, 13, 2271. [CrossRef]
- Chang, I.-S.; Roh, J. Stability of some inequalities in Banach *-algebras. Mathematics 2026, 14, 1407. [CrossRef]
- Brešar, M.; Vukman, J. On some additive mappings in rings with involution. Aeq. Math. 1989, 38, 178–185. [CrossRef]
- Stampfli, J.G. The norm of a derivation. Pac. J. Math. 1970, 33, 737–747. [CrossRef]
- Wright, J.D.M. Jordan C*-algebras. Michigan Math. J. 1977, 24, 291–302. [CrossRef]
- Hanche-Olsen, H.; Størmer, E. Jordan Operator Algebras; Pitman: Boston, MA, USA, 1984.
- McCrimmon, K. A Taste of Jordan Algebras; Springer: New York, NY, USA, 2004. [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.