---
title: Strongly Generalized Derivation
url: https://www.emergentmind.com/topics/strongly-generalized-derivation
type: topic
---

# Strongly Generalized Derivation

Strongly generalized derivation is a Leibniz-type operator notion whose most explicit formal definition in the recent literature is an order-\(n\) product identity of the form
\[
D_n(ab)=\sum_{k=1}^{n}\bigl[E_k(a)F_k(b)+G_k(a)H_k(b)\bigr],
\]
with suitable auxiliary families of maps. In order \(1\), this reduces to an \((E,F,G,H)\)-derivation,
\[
D(ab)=E(a)F(b)+G(a)H(b),
\]
which subsumes ordinary derivations, generalized derivations, generalized \((\sigma,\tau)\)-derivations, centralizers, homomorphisms, and ternary derivations. The phrase is not uniform across the broader derivation literature: many papers instead work with generalized derivations, quasiderivations, quasicentroids, or stronger constraints on generalized derivations, and these constitute the closest parallel frameworks in Banach, Lie, Hom, and \(n\)-ary settings [2308.16367] [2509.06634] [2201.06359].

## 1. Formal definition and order structure

Let \(\mathcal A\) and \(\mathcal B\) be algebras and let \(n\) be a positive integer. A linear mapping \(D:\mathcal A\to\mathcal B\) is called a strongly generalized derivation of order \(n\) if there exist families of linear mappings
\[
\{E_k:\mathcal A\to\mathcal B\}_{k=1}^n,\quad
\{F_k:\mathcal A\to\mathcal B\}_{k=1}^n,\quad
\{G_k:\mathcal A\to\mathcal B\}_{k=1}^n,\quad
\{H_k:\mathcal A\to\mathcal B\}_{k=1}^n
\]
such that
\[
D(ab)=\sum_{k=1}^{n}\bigl[E_k(a)F_k(b)+G_k(a)H_k(b)\bigr]
\qquad (a,b\in\mathcal A).
\]
A closely related formulation replaces the codomain algebra by a Banach \(\mathcal B\)-bimodule \(\mathcal M\): then a linear mapping \(D_n:\mathcal A\to\mathcal M\) is called a strongly generalized derivation of order \(n\) if there exist families
\[
\{E_k:\mathcal A\to\mathcal M\}_{k=1}^n,\quad
\{H_k:\mathcal A\to\mathcal M\}_{k=1}^n,\quad
\{F_k:\mathcal A\to\mathcal B\}_{k=1}^n,\quad
\{G_k:\mathcal A\to\mathcal B\}_{k=1}^n
\]
satisfying the same product identity [2308.16367] [2509.06634].

The order-one case is singled out in both papers:
\[
D(ab)=E(a)F(b)+G(a)H(b).
\]
This is called an \((E,F,G,H)\)-derivation. The order-one format is the main vehicle for automatic continuity results on Banach algebras and \(C^*\)-algebras, while the higher-order definition is designed to encompass finite sums of product-controlled terms [2308.16367] [2509.06634].

The order hierarchy contains many standard notions as exact special cases. If \(D=E=H\) and \(F=G=I\), then
\[
D(ab)=D(a)b+aD(b),
\]
so one recovers an ordinary derivation. If \(D=E\), \(F=G=I\), and \(H=d\), then
\[
D(ab)=D(a)b+a\,d(b),
\]
which is the usual generalized derivation form. If \(D=E\), \(F=\sigma\), \(G=\tau\), and \(H=d\), then
\[
D(ab)=D(a)\sigma(b)+\tau(a)d(b),
\]
so generalized \((\sigma,\tau)\)-derivations are included. If \(F=G=I\), then
\[
D(ab)=E(a)b+aH(b),
\]
which is the ternary derivation form. If \(D=E=F\) and either \(G=0\) or \(H=0\), then the identity collapses to homomorphism-like behavior [2308.16367] [2509.06634].

The higher-order definition is also nontrivial in concrete families. Every \((\delta,\varepsilon)\)-double derivation,
\[
d(ab)=d(a)b+ad(b)+\delta(a)\varepsilon(b)+\varepsilon(a)\delta(b),
\]
is a strongly generalized derivation of order \(2\). If \(\{d_n\}\) is a higher derivation satisfying
\[
d_n(ab)=\sum_{k=0}^n d_{n-k}(a)d_k(b),
\]
then each \(d_n\) is a strongly generalized derivation of suitable finite order \(m\), where
\[
m=
\begin{cases}
\frac{n+2}{2}, & n \text{ even},\\[4pt]
\frac{n+1}{2}, & n \text{ odd}.
\end{cases}
\]
These inclusions show that the order parameter organizes a wide class of “derivation-like” identities within one finite-sum scheme [2308.16367].

## 2. Relation to generalized derivations and earlier Leibniz-type frameworks

The strongest historical point of contact is the classical generalized derivation of Brešar type: a linear map \(u:\mathcal A\to\mathcal A\) for which there exists an ordinary derivation \(d\) such that
\[
u(ab)=u(a)b+a\,d(b).
\]
This identity is the algebraic starting point of the differential-geometric construction in “Generalized derivations and general relativity” [1301.0910]. In the Banach-algebra literature, the analogous notion is a pair \((\delta,d)\) with
\[
\delta(ab)=a\delta(b)+d(a)b,
\]
where \(d\) is a derivation and \(\delta\) is called a generalized derivation [2201.06359]. The order-one strongly generalized identity therefore enlarges two-sided Leibniz decompositions from one derivation companion to four auxiliary maps.

A second major comparison comes from the framework of \((o,T)\)-derivations and \((g_1,h_1,g_2,h_2)\)-derivations. An additive map \(d:A\to A\) is an \((o,T)\)-derivation if
\[
d(ab)=d(a)o(b)+T(a)d(b),
\]
and an additive map \(A:A\to A\) is a generalized \((o,T)\)-derivation corresponding to an additive \(d\) if
\[
A(ab)=A(a)o(b)+T(a)d(b).
\]
The most general class in that paper is the \((g_1,h_1,g_2,h_2)\)-derivation:
\[
f(ab)=g_1(a)h_1(b)+g_2(a)h_2(b).
\]
The paper explicitly notes that this framework includes ordinary derivations, generalized derivations, \((o,T)\)-derivations, generalized \((o,T)\)-derivations, ternary derivations, homomorphisms, and left/right centralizers [2209.11848]. In this sense, the order-one strongly generalized derivation is aligned with a broader tradition in which Leibniz-type rules are encoded by finitely many product components.

A third adjacent notion is the generalized \(\delta\)-derivation. A usual \(\delta\)-derivation \(\phi\) satisfies
\[
\phi(xy)=\delta\bigl(\phi(x)y+x\phi(y)\bigr),
\]
and a generalized \(\delta\)-derivation \(\chi\) associated with \(\phi\) satisfies
\[
\chi(xy)=\delta(\chi(x)y+x\phi(y))=\delta(\phi(x)y+x\chi(y)).
\]
That paper explicitly states that generalized \(\delta\)-derivations are a \(\delta\)-analogue of generalized derivations and notes that, when \(\delta=1\), the defining identity becomes the standard generalized derivation condition [1107.4420]. This suggests that “strongly generalized derivation” is best read as one member of a wider Leibniz-type hierarchy rather than as a universally standardized term.

Several later papers do not define “strongly generalized derivation” as a separate formal notion. Instead, they study generalized derivations subject to stronger-looking conditions, including range restrictions into radicals, \(k\)-centralizing and \(k\)-commuting constraints, square-closure conditions such as \((\delta^2,d^2)\) again being a generalized derivation, or additive decompositions into quasiderivation and quasicentroid parts [2201.06359] [1601.05345] [1406.1578]. In those settings, the strengthening is structural rather than terminological.

## 3. Automatic continuity on Banach algebras and \(C^*\)-algebras

The main analytic development of the subject concerns automatic continuity. In the order-one setting on algebras, one of the earliest systematic results shows that if \(A\) and \(B\) are complex algebras, \(A\) is simple and unital, and \(D:A\to B\) is a \((D,F,F,H)\)-derivation such that for some character \(\varphi\in\Phi\),
\[
\varphi(D(1))=\varphi(H(1))=0,\qquad \varphi(D(a_0))=0
\]
for some \(a_0\in A\setminus\{1\}\), then
\[
\dim B=1,\qquad \dim D(A),\,\dim F(A),\,\dim H(A)\le 1.
\]
If \(A\) and \(B\) are normed algebras, then \(D,F,H\) are continuous [2308.16367].

On operator algebras, a \( *\)-\((E,F,G,H)\)-derivation
\[
D(ab)=E(a)F(b)+G(a)H(b)
\]
from a unital \(C^*\)-subalgebra \(A\subseteq B(H)\) into \(B(H)\) is automatically continuous under explicit compatibility hypotheses. One version assumes \(E(I)=H(I)=I\),
\[
H(a)D(I)E(b)=-E(a)D(I)H(b)\qquad (a,b\in A),
\]
and that \(E+H\) is \(D\)-continuous; a second version assumes \(F(I)=G(I)=I\),
\[
G(a)D(I)F(b)=-F(a)D(I)G(b)\qquad (a,b\in A),
\]
and that \(F+G\) is \(D\)-continuous [2308.16367]. Closely related results show that if
\[
D(ab)=E(a)b+aH(b),
\]
then on a unital \(C^*\)-algebra \(D,E,H\) are generalized derivations and are automatically continuous in the \( *\)-case and in the commutative case; on a unital algebra, \(D\) and \(H\) are generalized derivations, and if every derivation on the algebra is continuous, then \(D,E,H\) are continuous [2308.16367].

Separating-space arguments extend these conclusions. If \(A\) is a Banach algebra, \(B\) a simple Banach algebra, and \(D:A\to B\) is both a continuous \((D,F,G,H)\)-derivation and a continuous \((H,F,G,D)\)-derivation, with \(H\) continuous and \(F,G\) surjective, then \(F\) and \(G\) are continuous, or else \(H\equiv 0\) and \(D\equiv 0\). If \(B\) is semiprime and either \(D\) is surjective with \(H\) continuous, or \(D\) is continuous with \(H\) surjective, then \(F\) and \(G\) are continuous [2308.16367].

A later \(C^*\)-algebra theorem shifts the codomain to a Banach \(\mathcal B\)-bimodule \(\mathcal M\). Let \(\mathcal A\) be a \(C^*\)-algebra, \(\mathcal B\) a Banach algebra, and \(D_1:\mathcal A\to\mathcal M\) an \((E,F,G,H)\)-derivation. If \(E,H:\mathcal A\to\mathcal M\) are linear, \(F,G:\mathcal A\to\mathcal B\) are continuous at \(0\), and one has either
\[
(G(ab)-G(a)G(b))\,H(c)=0
\]
or
\[
E(c)\,(F(ab)-F(a)F(b))=0
\]
for all \(a,b,c\in\mathcal A\), then \(D_1\) is automatically continuous [2509.06634]. Corollaries sharpen this in three directions: if at least one of \(F,G\) is a homomorphism, then every \((E,F,G,H)\)-derivation is continuous; if the separating-space annihilators
\[
\{m_0\in\mathcal M: G(\mathcal A)m_0=\{0\}\}=\{0\},\qquad
\{m_1\in\mathcal M: m_1F(\mathcal A)=\{0\}\}=\{0\},
\]
then \(E\) and \(H\) are also continuous; and if \(\mathcal B=\mathcal A\) with \(F,G\) surjective and continuous at \(0\), then \(D_1,E,H\) are continuous [2509.06634].

These theorems place strongly generalized derivations in the same automatic-continuity lineage as Sakai’s continuity of derivations on \(C^*\)-algebras, Ringrose’s continuity of bimodule derivations, and Johnson–Sinclair continuity on semisimple Banach algebras. The precise novelty is that continuity is forced for the asymmetric two-term identity
\[
D_1(ab)=E(a)F(b)+G(a)H(b)
\]
under mild algebraic side conditions rather than a classical Leibniz rule [2509.06634].

## 4. Decomposition phenomena in nonassociative, Hom-, and \(n\)-ary settings

A recurrent theme in nonassociative derivation theory is that generalized derivations admit a decomposition into a quasiderivation part and a centroid-like part. In \(3\)-Lie algebras, a generalized derivation is a linear map \(f\) for which there exist \(f_2,f_3,f'\) such that
\[
[f(x),y,z]+[x,f_2(y),z]+[x,y,f_3(z)] = f'([x,y,z]).
\]
The central structural theorem is
\[
GDer(A)=QDer(A)+Q\Gamma(A),
\]
where \(QDer(A)\) is the quasiderivation algebra and \(Q\Gamma(A)\) is the quasicentroid. The paper also proves that \([QDer(A),Q\Gamma(A)]\subseteq Q\Gamma(A)\) and that quasiderivations embed as derivations in the larger algebra
\[
A_t=A\otimes \bigl(tF[t]/(t^4)\bigr)
\]
via an explicit map \(l_u\) [1601.05345].

Exactly the same decomposition appears in multiplicative Hom-Jordan algebras:
\[
GDer(V)=QDer(V)+QC(V).
\]
There, generalized \(\alpha^k\)-derivations, quasiderivations, centroids, quasicentroids, and central derivations form a hierarchy
\[
ZDer(V)\subseteq Der(V)\subseteq QDer(V)\subseteq GDer(V),
\]
with
\[
ZDer(V)=C(V)\cap Der(V).
\]
The paper further proves that quasiderivations can be embedded as derivations in a larger Hom-Jordan algebra \(\dot V\) [1906.04551].

For multiplicative Hom-Lie superalgebras, the corresponding statement is again
\[
GDer(L)=QDer(L)+QC(L).
\]
The paper establishes closure properties such as \([QDer(L),QC(L)]\subseteq QC(L)\) and \([QC(L),QC(L)]\subseteq QDer(L)\), and it embeds \(QDer(L)\) into \(Der(\widetilde L)\) for a suitable larger Hom-Lie superalgebra \(\widetilde L\). If \(a\) is surjective and \(Z(L)=0\), then
\[
Der(\widetilde L)=\varphi(QDer(L))\oplus ZDer(\widetilde L)
\]
[1406.1578].

The same structural pattern extends to other nonassociative categories with variations in strength. For Lie triple systems,
\[
ZDer(T)\subseteq Der(T)\subseteq QDer(T)\subseteq GDer(T)\subseteq End(T),
\]
and quasiderivations embed into derivations of a larger Lie triple system \(\dot T\); if \(Z(T)=0\), then
\[
Der(\dot T)=\phi(QDer(T))\oplus ZDer(\dot T)
\]
[1412.7804]. For color \(n\)-ary \(\Omega\)-algebras, the hierarchy
\[
ZDer(T)\subseteq Der(T)\subseteq QDer(T)\subseteq GDer(T)\subseteq End(T)
\]
holds as well, and in the (anti)commutative case one has
\[
GDer(T)=QDer(T)+QC(T),
\qquad
\phi(QDer(T))\subseteq Der(\widehat T),
\]
with a direct-sum description
\[
Der(\widehat T)=\phi(QDer(T))\oplus ZDer(\widehat T)
\]
when the center vanishes [1506.00734].

These decomposition theorems do not usually employ the exact phrase “strongly generalized derivation.” A plausible implication is that, in nonassociative settings, the “strongly generalized” content is often encoded not by a new product formula but by a structural theorem identifying generalized derivations with sums of quasiderivation and quasicentroid-type components.

## 5. Banach-algebra generalized derivations under radical and centralizing constraints

A distinct but closely related direction studies generalized derivations on Banach algebras under strong spectral and radical hypotheses. Let \(A\) be a Banach algebra satisfying
\[
\operatorname{rad}(A)=\operatorname{rann}(A),
\qquad
A/\operatorname{rad}(A)\ \text{is commutative}.
\]
For a derivation \(d\), one then has
\[
d(A)\subseteq \operatorname{rad}(A).
\]
As a consequence,
\[
d^2=0 \quad \text{on }A,
\]
and every derivation is spectrally infinitesimal:
\[
r(d(a))=0 \qquad (a\in A).
\]
These are the starting points for the paper’s analysis of generalized derivations \((\delta,d)\) with
\[
\delta(ab)=a\delta(b)+d(a)b
\]
[2201.06359].

If \(A\) has a right identity, the paper proves the equivalence
\[
\delta(A)\subseteq \operatorname{rad}(A)
\iff
\delta \text{ is spectrally infinitesimal}
\iff
\delta=d.
\]
Hence, in this setting, the only generalized derivations whose range lies in the radical are the derivations themselves. Under the same assumption, every generalized derivation is spectrally bounded:
\[
r(\delta(a))\le C\,r(a).
\]
This is one of the clearest examples in which a “stronger” generalized-derivation property collapses the operator back to an ordinary derivation [2201.06359].

The paper also studies \(k\)-centralizing and \(k\)-commuting generalized derivations. For \(k\in\mathbb N\),
\[
[T(a),a]_k := [[T(a),a]_{k-1},a],\qquad [T(a),a]_1=[T(a),a].
\]
If \(A\) has a right identity and \((\delta,d)\) is a generalized derivation, then the following are equivalent:
\[
\delta \text{ is }k\text{-commuting},
\quad
\delta \text{ is }k\text{-centralizing},
\quad
\delta \text{ is a right multiplier},
\quad
\delta=R_b \text{ for some } b\in A.
\]
The associated set \(C(A)\) of centralizing generalized derivations is a Banach algebra, and
\[
C(A)\cong A/\operatorname{rad}(A).
\]
Thus centralizing generalized derivations are parametrized by the radical quotient [2201.06359].

A further strengthening asks when the square again defines a generalized derivation. If \((\delta,d)\) is a generalized derivation and \(A\) has a right identity, then the following are equivalent:
\[
[[\delta(a),a],\delta(a)]\in Z(A)\qquad (a\in A),
\]
\[
d\perp \delta,
\]
and
\[
(\delta^2,d^2)\ \text{is a generalized derivation of }A.
\]
The proof uses the computation
\[
\delta^2(ab)=a\delta^2(b)+2d(a)\delta(b)+d^2(a)b,
\]
so the mixed term \(d(a)\delta(b)\) must vanish [2201.06359]. This criterion is especially close in spirit to what some authors informally have in mind when speaking of stronger generalized-derivation conditions.

The same paper applies these results to group-related Banach algebras. For \(VN(G)\) when \(G\) is discrete amenable, for \(L^0(G)\) when \(G\) is abelian non-discrete, and for certain introverted subspaces \(X\subseteq VN(G)\), the standing hypotheses
\[
\operatorname{rad}(A)=\operatorname{rann}(A),
\qquad
A/\operatorname{rad}(A)\ \text{commutative}
\]
hold, and the generalized-derivation theorems above apply [2201.06359].

## 6. Higher-order Leibniz formulas, rigidity examples, and geometric extensions

One of the broadest conceptual messages in the literature is that generalized derivation identities admit higher-iterate Leibniz formulas. For an ordinary derivation,
\[
d^n(ab)=\sum_{k=0}^n \binom{n}{k}d^{\,n-k}(a)d^k(b).
\]
For an \((o,T)\)-derivation \(d\) with \([T,d]=[o,d]=0\),
\[
d^{n}(ab)=\sum_{k=0}^{n}\binom{n}{k}d^{\,n-k}(T^k(a))\,d^k(o^{\,n-k}(b)).
\]
For a \((g_1,h_1,g_2,h_2)\)-derivation under \([g_1,g_2]=[h_1,h_2]=0\),
\[
f^n(ab)=\sum_{k=0}^{n}\binom{n}{k}g_1^{\,n-k}\!\big(g_2^k(a)\big)\,h_2^{\,n-k}\!\big(h_1^k(b)\big).
\]
Without commutativity assumptions, the same paper derives a fully general formula using a non-commutative Newton expansion and the \({data}\)-product [2209.11848]. These results do not define strongly generalized derivations, but they show that generalized product decompositions remain stable under iteration in a binomial or noncommutative-binomial form.

Rigidity can be equally important. For generalized shifts on \(\ell^p(\tau)\), the restriction \(\sigma_\varphi|_{\ell^p(\tau)}\) is a \((\psi,\lambda)\)-derivation if and only if there exists
\[
\mathsf r=(r_\alpha)_{\alpha<\tau}\in\mathbb C^\tau
\]
such that
\[
\psi=\mathsf r\,\sigma_\varphi|_{\ell^p(\tau)},
\qquad
\lambda=((1)_\tau-\mathsf r)\,\sigma_\varphi|_{\ell^p(\tau)}.
\]
It is a \(\psi\)-derivation if and only if
\[
\psi=\tfrac12\,\sigma_\varphi|_{\ell^p(\tau)}.
\]
It is not a Jordan derivation and not a Jordan triple derivation, and
\[
\sigma_\varphi|_{\ell^p(\tau)}
\text{ is a generalized (Jordan, Jordan triple) derivation}
\iff
\varphi=\mathrm{id}_\tau.
\]
This is a particularly sharp example of how generalized-derivation behavior can force an operator to be trivial or identity-like [2104.02996].

Generalized derivations also support extensions beyond pure algebra. For the commutative algebra \(\mathcal A=C^\infty(M)\) on a Lorentzian manifold \(M\) of dimension \(N\),
\[
\mathrm{GDer}(\mathcal A)=\mathrm{span}_{\mathcal A}\bigl(\partial_0,\dots,\partial_{N-1},\mathrm{id}_{\mathcal A}\bigr).
\]
The extra basis element is denoted
\[
\partial_N:=\mathrm{id}_{\mathcal A},
\]
so the generalized derivation module has dimension \(N+1\). In the corresponding geometry, the special metric ansatz with
\[
g_{AB}=
\begin{pmatrix}
g_{\mu\nu} & 0\\
0 & \varepsilon\,\phi
\end{pmatrix}
\]
leads to a generalized Einstein–Hilbert action that reduces, for
\[
g_{\mu N}=0,\qquad g_{NN}=\varepsilon\,\phi^2,
\]
to the O’Hanlon action
\[
S_{\text{O'H}}=\frac{1}{2\kappa}\int\bigl(\phi R - V(\phi)\bigr)\sqrt{|g|}\,d^Nx,
\qquad
V(\phi)=\varepsilon\,N(N-1)\,\phi^{-1}.
\]
The paper also shows that the generalized Einstein equations with vanishing matter source are equivalent to a Kaluza–Klein theory satisfying the modified cylinder condition
\[
G_{AB}=e^{x^N} g_{AB}
\]
with a noncompact extra dimension [1301.0910]. Although this is a theory of generalized derivations rather than strongly generalized derivations in the modern order-\(n\) sense, it demonstrates the conceptual breadth of derivation enlargements based on generalized Leibniz rules.

Taken together, these results show that “strongly generalized derivation” names a concrete modern definition, but also sits inside a larger mathematical landscape. In associative operator theory it is an order-\(n\) finite-sum Leibniz scheme with strong automatic continuity properties; in Banach-algebra theory it is mirrored by restrictive spectral, centralizing, and square-closure conditions on generalized derivations; and in nonassociative settings it is approached through decomposition theorems of the form \(GDer=QDer+QC\) or \(GDer=QDer+Q\Gamma\).

Source: https://www.emergentmind.com/topics/strongly-generalized-derivation