---
title: 'Multiplier Algebra: Unital Envelope for Non-Unital Algebras'
url: https://www.emergentmind.com/topics/multiplier-algebra
type: topic
---

# Multiplier Algebra: Unital Envelope for Non-Unital Algebras

Searching arXiv for the cited paper and closely related multiplier algebra references.
Multiplier algebra is the unital envelope attached to a non-unital algebra or ring with non-degenerate multiplication. In the algebraic and \(C^\ast\)-algebraic settings, it is characterized as the largest algebra containing \(A\) as an essential ideal, and it is modeled by double centralizers; in the strict topology, it is the completion of \(A\). The same construction is the basic tool for multiplier Hopf algebras, where comultiplications naturally land in \(M(A\otimes A)\), and it has analytic incarnations in topological \(^\ast\)-algebras and function spaces, including the noncommutative Schwartz space and multiplier algebras of reproducing-kernel Hilbert spaces [2507.08769] [2103.05352] [2403.06863] [1703.09677].

## 1. Algebraic definition and universal property

Let \(A\) be an algebra over a field, not assumed unital. The basic hypothesis is non-degeneracy of the product: \(A\) is left non-degenerate if \(ab=0\) for all \(b\in A\) implies \(a=0\), right non-degenerate if \(ba=0\) for all \(b\in A\) implies \(a=0\), and non-degenerate if both conditions hold. Under this hypothesis, the multiplier algebra is defined by double centralizers:
\[
M(A)=\{(L,R)\in \operatorname{End}(A)\times \operatorname{End}(A)\mid L(a)b=aR(b)\ \text{for all}\ a,b\in A\}.
\]
For \(x=(L,R)\in M(A)\), one writes
\[
x a:=L(a),\qquad a x:=R(a),
\]
so the defining identity becomes \((ax)b=a(xb)\). The multiplication and identity are
\[
(L,R)\cdot(L',R')=(L\circ L',\,R'\circ R),\qquad 1_{M(A)}=(\operatorname{id}_A,\operatorname{id}_A),
\]
and \(M(A)\) is a unital associative ring or algebra [2507.08769].

The canonical embedding of \(A\) into \(M(A)\) is
\[
A\hookrightarrow M(A),\qquad a\mapsto (L_a,R_a),\qquad L_a(b)=ab,\ R_a(b)=ba.
\]
When \(A\) is non-degenerate, this map is injective, and \(A\) is identified with a two-sided ideal of \(M(A)\). The essential-ideal property is decisive: \(A\triangleleft B\) is essential if \(xB=0\Rightarrow x=0\), equivalently \(Ax=0\Rightarrow x=0\). In this sense, \(M(A)\) is the largest unital ring containing \(A\) as an essential two-sided ideal. More precisely, if \(B\) is any unital ring containing \(A\) as a two-sided ideal, there exists a unique unital homomorphism
\[
j:B\to M(A)
\]
whose restriction to \(A\) is the given embedding, determined by
\[
a\cdot j(b)=ab,\qquad j(b)\cdot a=ba,\qquad a\in A.
\]
Moreover, \(j\) is injective if and only if \(A\) is essential in \(B\) [2507.08769].

One-sided multiplier algebras are also fundamental. They are
\[
L(A)=\{y:A\to A\ \text{additive}\mid y(ab)=(ya)b\},
\]
\[
R(A)=\{z:A\to A\ \text{additive}\mid (ab)z=a(bz)\}.
\]
The multiplier algebra can be characterized as the pullback of \(L(A)\) and \(R(A)\) inside the space of \(A\)-bilinear maps \(A\otimes A\to A\). This formulation is particularly useful for functoriality: a non-degenerate homomorphism \(\gamma:A\to M(B)\) extends uniquely to a unital homomorphism
\[
\gamma_1:M(A)\to M(B),
\]
and the assignment \(A\mapsto M(A)\) yields a functor from non-degenerate idempotent rings and non-degenerate homomorphisms to unital rings [2507.08769].

## 2. Strict topology and local units

The strict topology is the natural topology on multiplier algebras. If \(A\) is non-degenerate and \(T\) is a ring containing \(A\), then the strict topology on \(T\) has neighborhood basis
\[
\mathcal{O}(x;a_1,\dots,a_n)=\{y\in T\mid a_i y=a_i x\ \text{and}\ y a_i=x a_i,\ \forall i\}.
\]
A net \((x_\alpha)\) converges strictly to \(x\) if and only if, for every \(a\in A\), one has \(a x_\alpha=a x\) and \(x_\alpha a=x a\) eventually. In this topology, \(M(A)\) is the completion of \(A\): every strict Cauchy net in \(A\) determines an element of \(M(A)\), and every multiplier arises this way [2507.08769].

Local units control strict density. An algebra \(A\) has local units if for every finite subset \(F\subset A\) there exists \(e\in A\) such that
\[
ea=a,\qquad ae=a,\qquad a\in F.
\]
The central equivalence is
\[
A\ \text{is strictly dense in}\ M(A)\ \Longleftrightarrow\ A\ \text{has local units}.
\]
If \(A\) has local units, then \(A\) is non-degenerate, idempotent, and firm; nets of local units converge strictly to the identity of \(M(A)\). In many commutative examples, local units can be chosen idempotent, while in \(^\ast\)-rings with local units they can be refined to self-adjoint, and even positive, local units [2507.08769].

Standard examples make the construction concrete. For an infinite set \(X\), the algebra \(K(X)\) of finitely supported functions with pointwise product has local units given by characteristic functions of finite subsets, and
\[
M(K(X))\cong C(X)
\]
with pointwise multiplication. By contrast, \(c_0\) is non-degenerate but has no local units, and
\[
M(c_0)=\ell^\infty;
\]
accordingly, \(c_0\) is not strictly dense in \(\ell^\infty\). For coalgebraic examples, if \(C\) is co-Frobenius, then the rational dual \(\operatorname{Rat}({}_{C^\ast}C^\ast)\) can have two-sided local units and multiplier algebra \(C^\ast\). In the Hopf-theoretic direction, the underlying algebras of multiplier Hopf algebras, algebraic quantum hypergroups, weak multiplier Hopf algebras, and algebraic quantum groupoids have local units [2507.08769].

In the \(C^\ast\)-algebraic case, the strict topology is generated by the seminorms \(p_a(x)=\|xa\|\) and \(q_a(x)=\|ax\|\), \(a\in A\). With this topology, \(A\) embeds into \(M(A)\) as a dense essential ideal, and multiplication by multipliers is continuous in the strict topology [2103.05352].

## 3. \(C^\ast\)-algebraic realizations and ultrapowers

For a \(C^\ast\)-algebra \(A\), the multiplier algebra has several equivalent descriptions. It is the algebra of double centralizers, and it is also the idealizer of \(A\) in the bidual:
\[
M(A)=\{m\in A^{\ast\ast}:mA\subseteq A,\ Am\subseteq A\}.
\]
If \(A\subset B(H)\) is represented non-degenerately, one may write
\[
M(A)=\{b\in B(H):bA\subset A\ \text{and}\ Ab\subset A\}.
\]
In all these formulations, \(A\) is an essential ideal in \(M(A)\), and \(M(A)\) is the maximal unital \(C^\ast\)-algebra containing \(A\) as such an ideal [1110.6858] [1903.07249].

Two canonical examples are pervasive. For a locally compact Hausdorff space \(X\),
\[
M(C_0(X))=C_b(X),
\]
and for the compact operators on a Hilbert space,
\[
M(K(\ell_2))=B(\ell_2).
\]
The corona algebra
\[
Q(A):=M(A)/A
\]
records the failure of \(A\) to be unital and is the natural codomain of Busby invariants of extensions [1110.6858] [1903.07249].

An ultrapower construction provides a concrete model for \(M(A)\). Let \(A\subset B(H)\) be faithful and non-degenerate, \((e_i)\) a bounded approximate identity indexed by a directed set \(I\), and \(\mathcal U\) a cofinal ultrafilter on \(I\). The ultrapower is
\[
A^{\mathcal U}=\prod_I A\,/\,c_{\mathcal U}(A),
\qquad
c_{\mathcal U}(A)=\{(a_i):\lim_{\mathcal U}\|a_i\|=0\}.
\]
Inside \(A^{\mathcal U}\), one defines the subalgebra \(A^{s\mathcal U}\) of classes represented by nets that are \(\mathcal U\)-strict convergent, and the ideal \(J\subset A^{s\mathcal U}\) of those converging \(\mathcal U\)-strictly to \(0\). The main identification is
\[
A^{s\mathcal U}/J\cong M(A),
\]
and the class of the approximate unit \([(e_i)]\) becomes the unit of \(M(A)\). Equivalently, \(M(A)\) can be recovered as the idealizer of the diagonal copy \(\Delta(A)\subset A^{\mathcal U}\) modulo the annihilator of \(\Delta(A)\) [1903.07249].

This ultrapower model is compatible with the classical strict topology. In it, strict multipliers are precisely the \(\mathcal U\)-strict limits of bounded nets from \(A\), and the identifications
\[
M(A)=\{x\in A^{\ast\ast}:xA\subset A,\ Ax\subset A\}
\]
and
\[
M(A)\cong A^{s\mathcal U}/J
\]
become two realizations of the same object [1903.07249].

## 4. Local multiplier algebras and quasi-multipliers

The local multiplier algebra organizes multipliers over essential ideals. If \(\mathcal E(A)\) denotes the directed set of closed, two-sided essential ideals of a \(C^\ast\)-algebra \(A\), then
\[
M_{\mathrm{loc}}(A)\cong \varinjlim_{I\in\mathcal E(A)} M(I).
\]
Iterating this construction produces the tower
\[
A\subset M_{\mathrm{loc}}(A)\subset M_{\mathrm{loc}}^{2}(A)\subset\cdots\subset I(A),
\]
where \(I(A)\) is the injective envelope. In the commutative case, \(M_{\mathrm{loc}}(A)\) is a commutative \(AW^\ast\)-algebra and hence injective, so the iteration stabilizes immediately; for simple \(A\), one has \(M_{\mathrm{loc}}(A)=M(A)\), and again the iteration stabilizes [1110.6858].

For separable \(C^\ast\)-algebras, the behavior of \(M_{\mathrm{loc}}^{2}(A)\) is governed by ideal-structure and topology of \(\operatorname{Prim}(A)\). There are separable examples with
\[
M_{\mathrm{loc}}^{2}(A)\supsetneq M_{\mathrm{loc}}(A),
\]
including tensor-product constructions \(A=C(X)\otimes B\) with \(X\) perfect and \(B\) simple non-unital. On the positive side, if \(A\) is quasicentral, separable, and \(\operatorname{Prim}(A)\) contains a dense \(G_\delta\) subset of closed points, then
\[
M_{\mathrm{loc}}^{2}(A)=M_{\mathrm{loc}}(A),
\]
and every derivation of \(M_{\mathrm{loc}}(A)\) is inner [1110.6858].

A related enlargement is the quasi-multiplier algebra
\[
QM(A)=\{x\in A^{\ast\ast}:AxA\subset A\},
\]
together with the one-sided multiplier algebras
\[
LM(A)=\{x\in A^{\ast\ast}:xA\subset A\},\qquad RM(A)=\{x\in A^{\ast\ast}:Ax\subset A\}.
\]
One always has
\[
M(A)=LM(A)\cap RM(A)\subset QM(A).
\]
For \(\sigma\)-unital \(A\), the coincidence \(QM(A)=M(A)\) has a sharp structural characterization: it holds if and only if \(A\) is the direct sum of a dual \(C^\ast\)-algebra and a locally unital \(C^\ast\)-algebra. Here “locally unital” means that there is a family of ideals \(\{I_j\}\) with \((\sum I_j)^{-}=A\) and, for each \(j\), an element \(u_j\in A\) such that
\[
(1-u_j)I_j=I_j(1-u_j)=\{0\}.
\]
In the simple \(\sigma\)-unital case, this reduces to:
\[
QM(A)=M(A)\quad\Longleftrightarrow\quad A\ \text{is unital or elementary}.
\]
The same paper proves \(LM(A)=M(A)\) if and only if \(QM(A)=M(A)\) [1812.11086].

## 5. Multiplier-valued comultiplication and Hopf-type structures

Multiplier algebras are indispensable when Hopf-type structures are imposed on non-unital algebras. Let \(A\) be a non-degenerate algebra. A multiplier Hopf algebra is a pair \((A,\Delta)\) with
\[
\Delta:A\to M(A\otimes A)
\]
an algebra homomorphism that is coassociative in \(M(A\otimes A\otimes A)\). The canonical maps are
\[
T_1(a\otimes b)=\Delta(a)(1\otimes b),\qquad
T_2(c\otimes a)=(c\otimes 1)\Delta(a),
\]
\[
T_3(a\otimes b)=(1\otimes b)\Delta(a),\qquad
T_4(c\otimes a)=\Delta(a)(c\otimes 1).
\]
In Van Daele’s definition, \((A,\Delta)\) is a multiplier Hopf algebra when \(T_1\) and \(T_2\) have range in \(A\otimes A\) and are bijections; it is regular when all four canonical maps have range in \(A\otimes A\) and are bijections [2403.06863].

A notable rigidity theorem states that this regularity can be recovered from one-sided data. A left multiplier Hopf algebra assumes regularity and bijectivity only for \(T_1\) and \(T_4\), together with the corresponding covered form of coassociativity; a right multiplier Hopf algebra makes the analogous assumption on \(T_2\) and \(T_3\). In either case one can construct a unique counit \(\varepsilon:A\to\mathbb C\) and a unique antipode \(S:A\to A\), prove that \(S\) is a bijective anti-automorphism, and then deduce that all four canonical maps are bijections with range in \(A\otimes A\). Thus every single sided multiplier Hopf algebra is automatically a regular multiplier Hopf algebra [2403.06863].

Weak multiplier bialgebras extend weak bialgebras to the non-unital setting by adjoining a canonical idempotent
\[
E\in M(A\otimes A)
\]
that replaces \(\Delta(1)\). In this framework, source and target maps take values in \(M(A)\), and their images \(B_L,B_R\subset M(A)\) are the base algebras. Under regularity and fullness assumptions, these base algebras carry coseparable co-Frobenius coalgebra structures; the multiplication on \(B_R\) is non-degenerate, \(B_R\) has local units, and the module category over \(A\) becomes monoidal via tensor product over the base algebra [1306.1466].

The categorical generalization is the theory of multiplier Hopf monoids in a braided monoidal category. A multiplier bimonoid is specified by fusion morphisms \(t_1,t_2:A\otimes A\to A\otimes A\) and a counit \(e:A\to I\); it is a multiplier Hopf monoid when \(t_1\) and \(t_2\) are isomorphisms. In \(\mathrm{Vect}_{\mathbb C}\), this recovers Van Daele’s definition. For a multiplier Hopf monoid, one proves existence and uniqueness of an antipode in a multiplier-valued sense; for a regular multiplier Hopf monoid, duals lift to module and comodule categories, and a Fundamental Theorem of Hopf modules holds [1511.03806].

## 6. Analytic multiplier algebras in function and operator spaces

In several analytic settings, “multiplier algebra” denotes the algebra of pointwise multipliers of a function space. For a Hilbert function space \(H\) on the unit ball \(B_d\),
\[
\operatorname{Mult}(H)=\{\varphi:B_d\to\mathbb C:\varphi f\in H\ \text{for all}\ f\in H\},
\]
with norm \(\|\varphi\|_{\operatorname{Mult}(H)}=\|M_\varphi\|_{B(H)}\). For regular unitarily invariant spaces, \(\operatorname{Mult}(H)\) is a weak-\(^\ast\) closed subalgebra of \(B(H)\). This algebra supports a multivariable functional calculus: if a commuting tuple \(T\) admits an \(A(H)\)-functional calculus and is completely non-unitary, then \(T\) is \(\operatorname{Mult}(H)\)-absolutely continuous, so the polynomial calculus extends to a weak-\(^\ast\) continuous completely contractive homomorphism
\[
\Phi_T:\operatorname{Mult}(H)\to B(\mathcal K).
\]
For tuples with a spherical unitary summand, absolute continuity is characterized by a \(\operatorname{Mult}(H)\)-Henkin spectral measure [1703.09677].

The same terminology appears in concrete spaces of holomorphic functions. For the Dirichlet space \(\mathcal D\), the multiplier algebra \(\mathcal M(\mathcal D)\) satisfies a Wolff-type ideal theorem: if \(H\in\mathcal M(\mathcal D)\) and \(F=(f_1,f_2,\dots)\subset \mathcal M(\mathcal D)\) obey \(\|M_C\|\le 1\) and
\[
|H(z)|\le \sum_j |f_j(z)|^2,\qquad z\in\mathbb D,
\]
then
\[
H^3\in I(f_1,f_2,\dots).
\]
The corresponding radical criterion is:
\[
H\in \operatorname{Rad} I(f_1,\dots,f_N)
\]
if and only if there exist \(C_0<\infty\) and \(m\in\mathbb N\) such that
\[
|H(z)|^m\le C_0\sum_{j=1}^N |f_j(z)|^2,\qquad z\in\mathbb D.
\]
For the Drury–Arveson space \(H^2_n\) and the Besov–Dirichlet type spaces \(H_{m,s}\), if \(f\) is a multiplier and \(|f(z)|\ge c>0\) on the ball, then \(f^t\) is a multiplier for every \(t\in\mathbb R\); moreover, for non-vanishing \(f\), \(\log f\) is a multiplier if and only if \(\log f\) is bounded [1302.5732] [2405.14401].

A substantially different analytic example is the noncommutative Schwartz space \(S_{nc}=(s',s)\). Its multiplier algebra \(ML(s',s)\) is identified simultaneously as the double centralizer algebra of \(S_{nc}\), as the maximal \(O^\ast\)-algebra on the Schwartz domain \(s\), and as a Köthe-type matrix \(PLB\)-space \(\Lambda(A)\). The algebra \(S_{nc}\) is an essential ideal in \(ML(s',s)\). Topologically, \(ML(s',s)\) is a nuclear, ultrabornological \(PLS\)-space; algebraically, it is neither a \(Q\)-algebra nor \(m\)-convex. Nonetheless, the closed graph theorem, open mapping theorem, and uniform boundedness principle remain valid [2103.05352].

Further analytic uses are highly structured. For generalized Toeplitz kernels, multipliers between kernels are characterized by an intrinsic embedding condition and a Smirnov-class factor condition \(h g^{-1}w\in \overline N\) or \(\overline{\mathcal N^+}\); in the upper half-plane, the existence of such multipliers is linked to Beurling–Malliavin densities, Pólya sequences, and the spectral theory of entire functions [2507.03452]. For the Herz algebra \(A_p(G)\), the pointwise multiplier algebra
\[
B_p(G)=\{m\in C_b(G):mu\in A_p(G)\ \text{for all}\ u\in A_p(G)\}
\]
is a commutative Banach algebra, and weak compactness of the multiplier operator \(M_m\) forces \(G\) to be discrete; when \(G\) is discrete and amenable, the weakly compact multipliers are exactly \(A_p(G)\) [1601.04282].

Across these algebraic, coalgebraic, and analytic contexts, the common function of a multiplier algebra is to supply the correct ambient unital object: it is the largest unital extension of a non-unital algebra as an essential ideal, the strict completion of the original algebra, and the natural codomain for operations—comultiplication, functional calculus, or pointwise multiplication—that do not remain internal to the underlying non-unital structure [2507.08769] [2403.06863].

Source: https://www.emergentmind.com/topics/multiplier-algebra