- Research
- Open access
- Published:
Some results on noncommutative Hardy-Lorentz spaces
Journal of Inequalities and Applications volume 2015, Article number: 120 (2015)
Abstract
Let \(\mathcal{A}\) be a maximal subdiagonal algebra of a finite von Neumann algebra ℳ. For \(0< p<\infty\), we define the noncommutative Hardy-Lorentz spaces and establish the Riesz and Szegö factorizations on these spaces. We also present some results of Jordan morphism on these spaces.
1 Introduction
The concept of maximal subdiagonal algebras \(\mathcal{A}\), which appeared earlier in Arveson’s paper [1], unifies analytic function spaces and nonselfadjoint operator algebras. In fact, subdiagonal algebras are the noncommutative analogue of weak* Dirichlet algebras. In the case that ℳ has a finite trace, \(H^{p}(\mathcal{A})\) may be defined to be the closure of \(\mathcal{A}\) in the noncommutative \(L_{p}\) space \(L_{p}(\mathcal{M})\). Subsequently, Arveson’s pioneering work has been extended to different cases by several authors. For example, Marsalli and West [2] obtained a series of results including a Riesz factorization theorem, the dual relations between \(H^{p}(\mathcal{A})\) and \(H^{q}(\mathcal{A})\), and Labuschagne [3] proved the universal validity of Szegö’s theorem for finite subdiagonal algebras. Recently, the noncommutative \(H_{p}\) spaces have been developed by Blecher, Bekjan, Labuschagne, Xu and their coauthors in a series of papers.
The noncommutative Hardy spaces have received a lot of attention since Arveson’s pioneering work. Most results on the classical Hardy spaces on the torus have been established in this noncommutative setting. Here we mention only two of them directly related with the objective of this paper. After the fundamental work of Arveson, Labuschagne [3] proved the noncommutative Szegö type theorem for finite subdiagonal algebras, and Blecher and Labuschagne [4] gave several useful variants of this theorem for \(L^{p}(\mathcal{M})\). In [5], Bekjan and Xu presented the more general form of Szegö type factorization theorem: Let \(0< p, q\leq\infty\). Let \(x\in L^{p}(\mathcal{M})\) be an invertible operator such that \(x^{-1}\in L^{q}(\mathcal{M})\). Then there exist a unitary \(u\in\mathcal{M}\) and \(h\in H^{p}(\mathcal{A})\) such that \(x=uh\) and \(h^{-1}\in H^{q}(\mathcal{A})\). The second result we wish to mention concerns the following direct decomposition: Let \(1< p<\infty\). Then
This result is proved in [6] for \(p=2\), in [2] for the general case as above.
In this article we introduce the noncommutative Hardy-Lorentz spaces \(H^{p, \omega}(\mathcal{A})\). If \(\omega\equiv1\), the noncommutative Hardy-Lorentz spaces \(H^{p, \omega}(\mathcal{A})\) correspond to the noncommutative Hardy spaces \(H^{p}(\mathcal{A})\). By adapting the ideas and techniques in [2, 5, 7], we establish the Riesz and Szegö factorizations on these spaces. We also present some results of inner-outer type factorization and Jordan morphism according to noncommutative Hardy-Lorentz spaces.
The remainder of this article is organized as follows. After a short introduction to this article, Section 2 consists of some preliminaries and notations, including the noncommutative weighted Lorentz spaces and their elementary properties. Section 3 presents the Riesz and Szegö factorization of noncommutative Hardy-Lorentz spaces. Section 4 contains some results of outer operators according to noncommutative Hardy-Lorentz spaces. The last section is devoted to Jordan morphism on these spaces.
2 Preliminaries
Throughout this paper ℳ will be a finite von Neumann algebra with a faithful normal tracial state τ. We refer to [8, 9] for the theory of noncommutative integration. Let ℳ be a von Neumann algebra acting on a Hilbert space ℋ. We denote by \(P(\mathcal{M})\) the complete lattice of all (self-adjoint) projections in ℳ. The closed densely defined linear operator x in ℋ with domain \(D(x)\) is said to be affiliated with ℳ if and only if \(yx\subseteq xy\) for all \(y\in\mathcal{M}^{\prime}\). When x is affiliated with ℳ, x is said to be τ-measurable if for every \(\varepsilon>0\) there exists \(e\in P(\mathcal{M})\) such that \(e(H)\subseteq D(x)\) and \(\tau(e^{\perp})<\varepsilon\) (where for any projection e, we let \(e^{\perp}=1-e\)). The set of all τ-measurable operators will be denoted by \(L_{0}(\mathcal{M})\). The set \(L_{0}(\mathcal{M})\) is an ∗-algebra with sum and product being the respective closure of the algebraic sum and product. The topology of \(L_{0}(\mathcal{M})\) is determined by the convergence in measure. For \(0< p<{\infty}\), let
We define
Then \(( L^{p}(\mathcal{M}); \|\cdot\|_{p})\) is a Banach (or quasi-Banach for \(p < 1\)) space. As usual, we put \(L^{\infty}(\mathcal{M};\tau)=\mathcal{M}\) and denote by \(\|\cdot\|_{\infty}\) (\(=\|\cdot\|\)) the usual operator norm.
For \(x\in L_{0}(\mathcal{M})\), we define
where \(e_{(t,\infty)}(|x|)\) is the spectral projection of \(|x|\) associated with the interval \((t, \infty)\). The function \(t\rightarrow\lambda_{t}(x)\) is called the distribution function of x and \(t\rightarrow \mu_{t}(x)\) is the generalized singular number of x. We will denote simply by \(\lambda(x)\) and \(\mu(x)\) the functions \(t\rightarrow\lambda_{t}(x)\) and \(t\rightarrow \mu_{t}(x)\), respectively. It is easy to check that both are decreasing and continuous from the right on \((0, \infty)\) (cf. [10]).
It will be sometimes convenient to write \(L^{1}=L^{1}(I)\) for brevity, where \(I=[0, 1]\). When ω is a nonnegative, integrable function on \([0, 1]\) and not identically zero, we say that ω is a weight. For a given weight ω, we write \(W(t)=\int_{0}^{t} \omega (s)\,ds<\infty\), \(0\leq t\leq1\). We also agree that ‘decreasing’ or ‘increasing’ will mean ‘nonincreasing’ or ‘nondecreasing’, respectively.
Let \(L_{0}\) be the set of all Lebesgue measurable functions on I. For \(f\in L_{0}\), we define its nonincreasing rearrangement as
where m denotes the Lebesgue measure on I. The Lorentz space \(\Lambda_{\omega}^{p}\), \(0< p<\infty\), is a subspace of \(L_{0}\) such that
It is clear that
Let \(0< p<\infty\), we define
where \(f^{**}(t)=\frac{1}{t}\int_{0}^{t}f^{*}(s)\,ds\). Let ω be a weight in I, we write \(W(t)=\int_{0}^{t}\omega (s)\,ds\in\Delta_{2}\) if there exists some constant C such that \(W(2t)\leq C W(t)\), \(0< t<\frac{1}{2}\). For any \(0< p<\infty\), it is well known that \(\Lambda_{\omega}^{p}\) is a quasi-Banach space if and only if \(W(t)\in\Delta_{2}\) (cf. [11]). Since we deal further with the space \(\Lambda_{\omega}^{p}\) which is at least a quasi-Banach space, we will assume in what follows that \(W(t)\in \Delta_{2}\).
Let \(\Lambda_{\omega}^{p}\) be a quasi-Banach space. For \(0< s<\infty\), we define the dilation operator \(D_{s}\) on \(\Lambda _{\omega}^{p}\) by
Define the lower Boyd index \(\alpha_{\Lambda_{\omega}^{p}}\) of \(\Lambda _{\omega}^{p}\) by
and the upper Boyd index \(\beta_{\Lambda_{\omega}^{p}}\) of \(\Lambda _{\omega}^{p}\) by
Note that
and
It is clear that \(0\leq\alpha_{\Lambda_{\omega}^{p}}\leq\beta _{\Lambda_{\omega}^{p}}\leq\infty\) and \(\alpha_{\Lambda_{\omega}^{rp}} =r\alpha_{\Lambda_{\omega}^{p}}\), \(\beta_{\Lambda_{\omega}^{rp}} =r\beta_{\Lambda_{\omega}^{p}}\), \(r>0\). If \(\Lambda_{\omega}^{p}\) is a Banach function space, then \(1\leq\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}\leq \infty\). For further results about the Boyd index of quasi-Banach spaces, the reader is referred to [12, 13].
We consider a couple \((E_{0}, E_{1})\) of topological vector spaces \(E_{0}\) and \(E_{1}\), which are both continuously embedded in a topological vector space E. The morphisms \(T:(E_{0}, E_{1})\rightarrow(E_{0}, E_{1})\) in E are all bounded linear mappings from \(E_{0}+E_{1}\) to \(E_{0}+E_{1}\) such that \(T_{E_{0}}: E_{0}\rightarrow E_{0}\) and \(T_{E_{1}}: E_{1}\rightarrow E_{1}\). A (quasi-)Banach space A is said to be an intermediate space between \(E_{0}\) and \(E_{1}\) if E is continuously embedded between \(E_{0}\cap E_{1}\) and \(E_{0}+E_{1}\). The space E is called an interpolation space between \(E_{0}\) and \(E_{1}\) if, in addition, \(T:(E_{0}, E_{1})\rightarrow(E_{0}, E_{1})\) implies \(T:E\rightarrow E\). Further details may be found in [12, 14].
Let \(0< p<\infty\), we say that \(\Lambda_{\omega}^{p}\) has order continuous norm if for every net \((f_{i})\) in \(\Lambda_{\omega}^{p}\) such that \(f_{i} \downarrow0\) we have \(\|f_{i}\|_{p, \omega}\downarrow0\). It follows from Proposition 2.3.3 and Theorem 2.3.4 of [11] that the norm on \(\Lambda_{\omega}^{p}\) is order continuous. Then it follows from Theorem 3.2 of [12] that \(\Lambda_{\omega}^{r}\) is an interpolation space for the couple \((L^{p}, L^{q})\), where \(0< p<\alpha_{\Lambda_{\omega}^{r}}\leq\beta _{\Lambda_{\omega}^{r}}<q\leq\infty\). Let \(1< p<\infty\), it is well known (Theorem A, [15]) that \(\Lambda_{\omega}^{p}=\Gamma_{\omega}^{p}\) if and only if ω satisfies the condition \(B_{p}\) (\(\omega\in B_{p}\)), that is, there exists \(B>0\) such that \(\int_{t}^{1}s^{-p}\omega(t)\,dt\leq Bt^{-p}W(t)\), \(t\in I\). We refer to [11, 15] for these spaces.
A (quasi-)Banach function space E is called symmetric if for \(f\in L_{0}\) and \(g\in E\) with \(f^{*}\leq g^{*}\), we have \(f\in E\) and \(\|f\|_{E}\leq\|g\|_{E}\). It is called fully symmetric if, in addition, for \(f\in L_{0}\) and \(g\in E\) with \(\int_{0}^{t}f^{*}(s)\,ds\leq\int_{0}^{t}g^{*}(s)\,ds\), we have \(f\in E\) and \(\|f\|_{E}\leq\|g\|_{E}\). If \(1\leq p<\infty\) and ω is a non-increasing weighted function, it is clear that \(\Lambda_{\omega}^{p}\) is a fully symmetric Banach function space. Let \(0< p<\infty\), then \(\Gamma_{\omega}^{p}\) is a fully symmetric quasi-Banach function space.
Let \(x\in L_{0}(\mathcal{M})\) and \(0< p<\infty\). We define
By a simple computation we derive
The noncommutative Lorentz space \(\Lambda^{p}_{\omega}(\mathcal{M})\) is defined as the space of all τ-measurable operators affiliated with a finite von Neumann algebra ℳ such that \(\|x\|_{p, \omega}< \infty\). If \(\omega\equiv1\), then the noncommutative Lorentz space \(\Lambda^{p}_{\omega}(\mathcal{M})\) is the usual noncommutative \(L_{p}\) space \(L_{p}(\mathcal{M})\). If \(\omega=t^{\frac{p}{q}-1}\), \(p, q>0\), then \(\Lambda^{p}_{\omega}(\mathcal{M})=L^{q, p}(\mathcal{M})\). For further results about the noncommutative Lorentz spaces \(L^{q, p}(\mathcal{M})\), the reader is referred to [16].
If \(0< p<\infty\), we define
where \(x^{**}(t)=\frac{1}{t}\int_{0}^{t}\mu_{s}(x)\,ds\).
Since \(\Lambda_{\omega}^{p}\) and \(\Gamma_{\omega}^{p}\) are quasi-Banach spaces, then the following result follows from Theorem 4 of [17].
Proposition 2.1
Let \(0< p<\infty\), then \(\Lambda_{\omega}^{p}(\mathcal{M})\) and \(\Gamma_{\omega}^{p}(\mathcal{M})\) are quasi-Banach spaces.
We should introduce the Köthe dual spaces (associate spaces) generalizing the definition that can be found in [11] in the context of classical Lorentz space \(\Lambda_{\omega}^{p}\), \(0< p\leq\infty\). We define the Köthe dual space of \(\Lambda_{\omega}^{p}(\mathcal {M})\) by
If \(x\in\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}\), it is clear that
Lemma 2.1
Let ℳ have no minimal projection for every measurable function f with
then there exists \(x\in L_{0}(\mathcal{M})\) such that \(\mu_{t}(x)=f^{*}(t)\).
Proof
By Lemma 1.8 of [18], there exist \(Q_{k2^{-n}}\in\mathcal{M}_{\mathrm{proj}}\) such that \(\tau(Q_{k2^{-n}})=k2^{-n}\) and \(Q_{k_{1}2^{-n_{1}}}\leq Q_{k_{2}2^{-n_{2}}}\), \(k_{1}2^{-n_{1}}\leq k_{2}2^{-n_{2}}\), where \(n,k, n_{i}, k_{i}\in \mathbb{N}\), \(i=1,2\). Let \(Q_{\lambda}=\sup_{\frac{k}{2^{n}}\leq\lambda}Q_{k2^{-n}}\), \(\lambda \in[0, 1]\). Then \(\{Q_{\lambda}\}_{\lambda\in[0, 1]}\) is an increasing family of projections in ℳ, and it is clear that \(\tau(Q_{\lambda})=\lambda\), \(Q_{0}=0\) and \(Q_{1}=1\). Therefore, it is a spectral family of ℳ. Let \(f\in L_{0}\) with \(\lim_{t\rightarrow\infty}d_{f}(t)=0\). For \(x=\int_{0}^{\infty}f^{*}(\lambda)\,dQ_{\lambda}\), we have
which implies \(x\in L_{0}(\mathcal{M})\) and \(\mu_{t}(y)=f^{*}(t)\), \(t>0\). □
Lemma 2.2
Let \(1< p<\infty\). Then the injection \(\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\hookrightarrow L_{0}(\mathcal{M})\) is continuous.
Proof
For \(x\in\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\) with \(\|x\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}}\leq1\). Since \(\mu_{t}(x)\rightarrow0\), \(t\rightarrow\infty\), then \(|x|\) admits the Schmidt decomposition \(|x|=\int_{0}^{\infty}\mu_{t}(x)\,d\widetilde{e}_{t}\), where \(\widetilde{e}_{t}=e_{\mu_{t}(x)-0}\), \(t>0\), and \(e_{0-0}=1\) (cf. [19]). Given \(\delta>0\), it is clear that \(\mu_{t}(x)\geq\mu_{\delta}(x)\chi _{[\frac{\delta}{2}, \delta)}\). Thus \(|x|=\int_{0}^{\infty}\mu_{t}(x)\,d\widetilde{e}_{t}\geq\mu_{\delta}(x)q\), where \(q=\int_{0}^{\infty}\chi_{[\frac{\delta}{2}, \delta )}\,d\widetilde{e}_{t}\). Since \(\tau(q)<\infty\), we obtain \(q\in\Lambda_{\omega}^{p} (\mathcal {M})^{\prime}\). Therefore, \(\|x\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}} =\||x|\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}} \geq\mu_{\delta}(x)\|q\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}}\) and \(\|q\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}}^{-1}\geq\mu _{\delta}(x)\), which complete the proof. □
Proposition 2.2
Let \(1< p<\infty\). Then \(\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\) is a noncommutative Banach function space.
Proof
It is clear that \(\|\cdot\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}}\) is subadditive, homogenous and positive. If \(\|x\|_{\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}}=0\), then \(\|xy\|_{1}=0\) for every \(y\in\Lambda_{\omega}^{p} (\mathcal{M})\). Since \(e_{A}(|x|)\in\Lambda_{\omega}^{p} (\mathcal{M})\), we have \(xe_{A}(|x|)=0\), where A is a subset of \((0, \infty)\), which implies that \(x=0\). The proof of Theorem 8.11 of [20] shows that it is sufficient to prove the noncommutative form of the Riesz-Fischer theorem, i.e., we have an estimate
whenever the right-hand side is finite. Let \(\sum_{n=1}^{\infty}\|x_{n}\|_{\Lambda_{\omega}^{p} (\mathcal {M})^{\prime}}<\infty\), then \(\sum_{n=1}^{\infty}x_{n}\) converges to some x in \(L_{0}(\mathcal{M})\). Indeed, set \(z_{n}=\sum_{k=1}^{n} x_{k}\), it is clear that \(\{z_{n}\}_{n=1}^{\infty}\) is a Cauchy sequence in \(\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\). It follows from Lemma 2.2 that \(\{z_{n}\}_{n=1}^{\infty}\) converges to some x in \(L_{0}(\mathcal{M})\). Since \(\sum_{n=1}^{\infty}\|x_{n}y\|_{\Lambda_{\omega}^{p} (\mathcal {M})^{\prime}} \leq\sum_{n=1}^{\infty}\|x_{n}\|_{\Lambda_{\omega}^{p} (\mathcal {M})^{\prime}}<\infty\), then
holds for each \(y\in\Lambda_{\omega}^{p} (\mathcal{M})\) with \(\|y\|_{\Lambda_{\omega}^{p} (\mathcal{M})}\leq1\). Thus \(\sum_{n=1}^{\infty}x_{n}\in\Lambda_{\omega}^{p} (\mathcal {M})^{\prime}\) and
□
Proposition 2.3
Let ℳ have no minimal projection, then the associate space \(\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\) is a noncommutative Banach function space. For \(x\in\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\),
Proof
Let \(x \in\Lambda_{\omega}^{p} (\mathcal {M})^{\prime}\) and \(y\in\Lambda_{\omega}^{p} (\mathcal{M})\), then \(xy\in L^{1}(\mathcal{M})\). By Theorem 4.2 of [10], we have \(\tau(|xy|)\leq\int_{0}^{\infty}\mu_{t}(x)\mu_{t}(y)\,dt\). Thus
To prove the other inequality, let \(y\in\Lambda_{\omega}^{p} (\mathcal{M})\) with \(\|y\|_{\Lambda_{\omega}^{p} (\mathcal{M})}\leq1\). By Proposition 2.2 of [21], we obtain
Hence, \(\|x\|_{\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}}\geq \sup\{\int_{0}^{\infty}\mu_{t}(x)\mu_{t}(y)\,dt: \|y\|_{\Lambda_{\omega}^{p}(\mathcal{M})}\leq1\}\). □
Proposition 2.4
Let \(x\in\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\), then we have
Moreover, \((\Lambda^{p}_{\omega})^{\prime}(\mathcal{M})=\Lambda_{\omega}^{p} (\mathcal{M})^{\prime}\).
Proof
This result follows immediately from Lemma 2.1 and Proposition 2.3. First, let ℳ have no minimal projection. For every \(f\in\Lambda_{\omega}^{p}\), by Lemma 2.1, there exists \(y\in\Lambda_{\omega}^{p}(\mathcal{M})\) such that \(\mu(y)=f^{*}(t)\), \(t>0\). Thus, by Proposition 2.3, we obtain
If ℳ has minimal projections, we consider the von Neumann algebra tensor product \(\mathcal{M}\bar{\otimes}L^{\infty}([0, 1], m)\) denoted by \(\overline{\mathcal{M}}\), equipped with the tensor product trace \(\tau\otimes m\), then \(\overline{\mathcal{M}}\) has no minimal projection. By the trivial fact \(\mu(x)=\mu(x\otimes1)\) and argument of above, we have \(\|x\|_{\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}}= \|\mu_{t}(x)\|_{\Lambda_{\omega}^{p}(I)^{\prime}}\), which implies that
□
Definition 2.1
Let \(0< p\leq\infty\). If \(l\in\Lambda_{\omega}^{p} (\mathcal{M})^{*}\), then l is called normal if \(x_{\alpha}\downarrow0\) holds in \(\Lambda_{\omega}^{p} (\mathcal{M})\) implies \(l(x_{\alpha})\rightarrow0\).
If \(l\in\mathcal{M}^{*}\) then, by Theorem 5.11 of [22], l is normal if and only if l is ultra-weak topology on ℳ. Then a similar discussion of Lemma 5.10 of [23] leads to the following lemma.
Lemma 2.3
Let \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\), \(1< p<\infty\) and \(l\in\Lambda_{\omega}^{p}(\mathcal{M})^{*}\). If \(e\in \mathcal{M}_{\mathrm{proj}}\), we define \(l_{e}\) by setting \(l_{e}(x)=l(ex)\), \(x\in\mathcal{M}\), then \(l_{e}\) is a normal linear functional on ℳ.
Proof
Let \(e\in\mathcal{M}\) and \(x_{0}>x_{\alpha}\downarrow0\) hold in ℳ, then \(ex_{\alpha}e \downarrow0\) holds in \(L^{1}(\mathcal{M})\), and so \(\mu_{t}(ex_{\alpha}e)\downarrow0\) holds in \(L^{1}\). Thus,
By Theorem 2.3.4 of [11], we have \(\|e x_{\alpha}\|_{\Lambda _{\omega}^{p}(\mathcal{M})} =(\int_{0}^{1}\mu_{t}(ex_{\alpha})^{p}\omega(t))^{\frac{1}{p}}\downarrow0\). Therefore, \(l_{e}(x_{\alpha})\rightarrow0\). This tells us that \(l_{e}\) is a normal linear functional on ℳ. □
Proposition 2.5
Let \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\), \(1< p<\infty\). If \(l\in \Lambda_{\omega}^{p}(\mathcal{M})^{*}\), then there exists a unique operator \(y\in \Lambda_{\omega}^{p}(\mathcal{M})^{\prime}\) such that \(l(x)=\tau(yx)\) for every \(x\in\mathcal{M}\) and \(\|l\|\geq\|y\|_{\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}}\).
Proof
It follows from Theorem 3.2 of [12] that \(\Lambda_{\omega}^{p}\) is an interpolation space for the couple \((L^{1}, L^{\infty})\). Thus \(L^{\infty}\hookrightarrow\Lambda _{\omega}^{p}\hookrightarrow L^{1}\), and so \(\mathcal{M}\hookrightarrow\Lambda_{\omega}^{p}(\mathcal {M})\hookrightarrow L^{1}(\mathcal{M})\), where ‘↪’ denotes a continuous embedding. It follows that there exists some constant C such that
and \(W(t)\leq C\), \(t\in[0, 1]\). We apply Lemma 2.3 and some techniques from the proof of Theorem 5.11 in [23]. The result follows in the same way as in [23]. □
Proposition 2.6
Let \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\), \(1< p<\infty\). Then
-
(i)
\(\Lambda_{\omega}^{p}(\mathcal{M})^{*}=\Lambda_{\omega}^{p}(\mathcal {M})^{\prime}=(\Lambda_{\omega}^{p})^{\prime}(\mathcal{M}) =\Gamma_{ \widetilde{\omega} }^{q}(\mathcal{M})\), where \(\widetilde{\omega}(t)=t^{q}W(t)^{-q}\omega(t)\), \(t>0\).
-
(ii)
If \(e_{i}\) is an increasing sequence of projections in ℳ converging strongly to 1, we have \(\lim_{i\rightarrow\infty}\|xe_{i}-x\|_{p, \omega}=0\), \(\lim_{i\rightarrow\infty}\|e_{i}x-x\|_{p, \omega}=0\), \(\forall x\in\Lambda_{\omega}^{p}(\mathcal{M})\).
-
(iii)
\(\Lambda_{\omega}^{p}(\mathcal{M})^{*}\) separates points.
Proof
(i): Combining Proposition 2.4 with Remark 2.4.8 of [11], we obtain \(\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}=(\Lambda_{\omega}^{p})^{\prime}(\mathcal{M}) =\Gamma_{ \widetilde{\omega} }^{q}(\mathcal{M})\). For \(l\in\Lambda_{\omega}^{p}(\mathcal{M})^{*}\), by Proposition 2.5, there exists \(y\in\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}\) such that
Let \(l^{s}=l-l_{y}\), we have \(l^{s}\in\Lambda_{\omega}^{p}(\mathcal{M})^{*}\) and \(l^{s}(x)=0\), \(x\in\mathcal{M}\). For \(x\in\Lambda_{\omega}^{p}(\mathcal{M})\), it is well known that \(\tau(e_{(\frac{1}{n}, \infty)}(|x|))<\infty\), \(n=1, 2, 3, \ldots \) . Let \(x_{n}=xe_{[0, \frac{1}{n}]}(|x|)\), \(n=1, 2, \ldots \) , then \(x-x_{n}\in \mathcal{M}\). This implies that \(|l^{s}(x)|=|l^{s}(x_{n})|\leq\|l^{s}\|\|x_{n}\|_{p, \omega}\), \(n=1, 2, \ldots\) . Note that \(|x_{n}|\leq|x|\) and \(\mu(x_{n})\rightarrow0\), \(n\rightarrow \infty\). Then, by Theorem 2.3.4 of [11], we have \(\|x_{n}\|_{p, \omega}=\|\mu(x_{n})\|_{p, \omega}\rightarrow0\), \(n\rightarrow\infty\), and so \(|l^{s}(x)|=|l^{s}(x_{n})|=0\). Therefore, \(l^{s}=0\), that is, \(l=l_{y}\). (ii): Without loss of generality, we suppose \(0\leq x\in\Lambda^{p}_{\omega}(\mathcal{M})\). Let \(x=\int_{0}^{\infty}\lambda de_{\lambda}\) be the spectral decomposition of x. We write \(q_{i}=1-e_{i}\), then \(q_{i}\) converges strongly to 0 and \(q_{i}>q_{i+1}\), \(n=1,2,\ldots \) . Set \(x_{i}=(xq_{i}x)^{\frac{1}{2}}\), \(n=1,2,\ldots \) . Since \(\tau(1)<\infty\), then \(x_{i}^{2} \rightarrow0\) in the measure topology. Thus, by Theorem 2.3.4 of [11] and \(\mu_{t}( x_{n} ) \leq\mu _{t}(x)\), we have
It follows that \(\|xq_{n}\|_{p, \omega}=\|(xq_{n}x)^{\frac{1}{2}}\|_{p, \omega} =\|x_{n}\|_{p, \omega}\rightarrow0\), \(n\rightarrow\infty\). (iii): Suppose that there exists \(0\neq x_{0}\in\Lambda_{\omega}^{p}(\mathcal {M})\) such that \(l(x_{0})=0\) holds for all \(l\in\Lambda_{\omega}^{p}(\mathcal{M})^{*}\). Then, for every \(y\in\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}\), we have \(l_{y}(x_{0})=\tau(x_{0}y)=0\). By (i), we have \(\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}=\Gamma_{ \widetilde{\omega} }^{q}(\mathcal{M})\), then \(e_{(a, \infty)}(|x_{0}|)\in\Lambda_{\omega}^{p}(\mathcal {M})^{\prime}\), \(\forall a>0\). Thus \(\tau(x_{0}e_{(a, \infty)}(|x_{0}|))=l_{e_{(a, \infty )}(|x_{0}|)}(x_{0})=0\), \(\forall a>0\), which implies that \(x_{0}=0\), contradiction. Therefore, \(\Lambda_{\omega}^{p}(\mathcal{M})^{*}\) separates points. □
The identity in ℳ is denoted by 1, and we denote by \(\mathcal{D}\) a von Neumann subalgebra of ℳ; moreover, we let \(\mathcal{E}: \mathcal{M}\rightarrow\mathcal{D}\) be the unique normal faithful conditional expectation such that \(\tau\circ\mathcal{E}=\tau\). A finite subdiagonal algebra of ℳ with respect to ℰ (or \(\mathcal{D}\)) is a \(w^{*}\)-closed subalgebra \(\mathcal{A}\) of ℳ satisfying the following conditions:
-
(i)
\(\mathcal{A}+J\mathcal{A}\) is \(w^{*}\)-dense in ℳ;
-
(ii)
ℰ is multiplicative on \(\mathcal{A}\), i.e., \(\mathcal{E}(ab)=\mathcal{E}(a)\mathcal{E}(b)\) for all \(a, b \in \mathcal{A}\);
-
(iii)
\(\mathcal{A}\cap J\mathcal{A}=\mathcal{D}\).
\(\mathcal{D}\) is then called the diagonal of \(\mathcal{A}\), where \(J\mathcal{A}=\{x^{*}: x\in\mathcal{A}\}\). We say that \(\mathcal{A}\) is a maximal subdiagonal algebra in ℳ with respect to ℰ in case that \(\mathcal{A}\) is not properly contained in any other subalgebra of ℳ which is subdiagonal with respect to ℰ. It is proved by Exel (Theorem 7, [24]) that a finite subdiagonal algebra \(\mathcal{A}\) is automatically maximal. This maximality yields the following useful characterization of \(\mathcal{A}\):
where \(\mathcal{A}_{0}=\mathcal{A}\cap \operatorname{ker}\mathcal{E}\) (see [1]).
If K is a subset of \(\Lambda_{\omega}^{p}(\mathcal{M})\), \([K]_{ p, \omega}\) will denote the closure of K in \(\Lambda_{\omega}^{p}(\mathcal{M})\) (with respect to the \(w^{*}\)-topology in the case of \(p=\infty\)). Since ℳ is a finite von Neumann algebra, then \(\mathcal{M}=\Lambda_{\omega}^{\infty}(\mathcal{M})\subseteq\Lambda _{\omega}^{p}(\mathcal{M})\), \(0< p\leq\infty\).
Definition 2.2
Let ℳ be a finite von Neumann algebra, we define noncommutative weighted Hardy spaces by \(H^{p,\omega}(\mathcal{A})=[\mathcal{A}]_{ p, \omega}\) and \(H^{p,\omega}_{0}(\mathcal{A})=[\mathcal{A}_{0}]_{p, \omega}\).
In what follows, we will keep all previous notations throughout the paper, and C will always denote a constant, which may be different in different places. Unless otherwise stated, it will be assumed throughout that Lorentz spaces \(\Lambda_{\omega}^{p}\) satisfy the following property: for \(f\in L_{0}\) and \(g\in\Lambda_{\omega}^{p}\) with \(\int_{0}^{t}f^{*}(s)\,ds\leq\int_{0}^{t}g^{*}(s)\,ds\), we have
For two nonnegative (possibly infinite) quantities A and B, by \(A\lesssim B\) we mean that there exists a constant \(C>0\) such that \(A\leq CB\).
Remark 2.1
-
(i)
If \(\omega\equiv1\), then \(H^{p,\omega}(\mathcal {A})=H^{p}(\mathcal{A})\) and \(H^{p,\omega}_{0}(\mathcal{A})=H^{p}_{0}(\mathcal{A})\). In [25], Section 3 it is shown that for \(1\leq p\leq\infty\),
$$H^{p}(\mathcal{A})=[\mathcal{A}]_{p} = \bigl\{ x\in L^{p}(\mathcal{M}): \tau(xy)=0 \mbox{ for all } y\in \mathcal {A}_{0} \bigr\} $$and
$$H^{p}_{0}(\mathcal{A})=[\mathcal{A}_{0}]_{p} = \bigl\{ x\in L^{p}(\mathcal{M}): \tau(xy)=0 \mbox{ for all } y\in \mathcal {A} \bigr\} . $$Subsequently, Bekjan and Xu (Proposition 3.3, [5]) showed that \(H^{q}(\mathcal{A})=H^{p}(\mathcal{A})\cap L^{q}(\mathcal{M})\) and \(H^{q}_{0}(\mathcal{A})=H^{p}_{0}(\mathcal{A})\cap L^{q}(\mathcal{M})\), where \(0< p<q\leq\infty\).
-
(ii)
Let \(0< p\leq q\leq\infty\). Since ℳ is a finite von Neumann algebra, then \(\Lambda_{\omega}^{q}(\mathcal{M})\subseteq\Lambda_{\omega}^{p}(\mathcal{M})\).
-
(iii)
Let \(0< r_{1}<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda _{\omega}^{p}}<r_{2}\leq\infty\). It follows from Theorem 3.2 of [12] that \(\Lambda_{\omega}^{p}\) is an interpolation space for the couple \((L^{r_{1}}, L^{r_{2}})\). Thus \(L^{r_{2}}\hookrightarrow\Lambda_{\omega}^{p}\hookrightarrow L^{r_{1}}\), and so \(L^{r_{2}}(\mathcal{M})\hookrightarrow\Lambda_{\omega}^{p}(\mathcal {M})\hookrightarrow L^{r_{1}}(\mathcal{M})\), where ‘↪’ denotes a continuous embedding. Moreover, this implies that \(H^{r_{2}}(\mathcal{A})\hookrightarrow H^{p, \omega}(\mathcal {A})\hookrightarrow H^{r_{1}}(\mathcal{A})\).
-
(iv)
Let \(1< p<\infty\) and \(\omega\in B_{p}\). Then Theorem A of [15] implies that \(\Lambda_{\omega}^{p}\) satisfies property (2.4).
-
(v)
If \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\) and \(\Lambda_{\omega}^{p}\) satisfies property (2.4), it follows from Proposition 1 of [26] that ℰ is bounded on \(\Lambda_{\omega}^{p}(\mathcal{M})\).
3 Riesz and Szegö factorization of noncommutative weighted Hardy spaces
Proposition 3.1
Let \(1< p<\infty\) and \(1< \alpha_{\Lambda_{\omega}^{p}}\leq\beta _{\Lambda_{\omega}^{p}}<\infty\). Then
Moreover, if \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda _{\omega}^{p}}<\infty\), \(1< p<\infty\) and \(r<\alpha_{\Lambda_{\omega}^{p}}\), then
Proof
Since \(1< \alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\) and \(\mathcal{A}_{0}\) is an ideal of \(\mathcal{A}\), then
For \(x\in H^{p, \omega}(\mathcal{A})\), there exist \(x_{n}\in\mathcal{A}\) such that \(\|x-x_{n}\|_{p, \omega}\rightarrow0\), \(n\rightarrow\infty\) and \(\tau(x_{n}y)=0\), \(\forall y\in\mathcal{A}_{0}\), \(n=1, 2, \ldots \) . Hence, \(x_{n}\rightarrow x\) in the measure topology. This means that \(x_{n}y\rightarrow xy\) in the measure topology holds for every \(y\in\mathcal{A}_{0}\). By Remark 2.1(iii), we obtain \(\Lambda_{\omega}^{p}(\mathcal{M})\subseteq L^{1}(\mathcal{M})\). Using Theorem 3.5 of [10], we obtain
which implies that \(\tau(xy)=0\), and so
Conversely, let \(x\in\{x\in\Lambda_{\omega}^{p}(\mathcal{M}): \tau(xy)=0, \forall y\in\mathcal{A}_{0}\}\) and \(x\notin H^{p, \omega}(\mathcal{A})\). By (i) and (iii) of Proposition 2.6, there exists \(y\in\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}\) such that \(\tau(xy)\neq0\) and \(\tau(ya)=0\), \(\forall a\in H^{p, \omega}(\mathcal{A})\). Since \((f+g)^{**}\leq f^{**}+g^{**}\), \(\Gamma_{ \widetilde{\omega} }^{q}\) is Banach function spaces over \([0, 1]\), where \(\frac{1}{q}+\frac{1}{p}=1\). By Corollary II. 6.7 of [27], we have \(\Gamma_{ \widetilde{\omega} }^{q}\subseteq L^{1}\), which implies that \(\Gamma_{ \widetilde{\omega} }^{q}(\mathcal{M})\subseteq L^{1}(\mathcal{M})\). Combining this with Proposition 2.6, we obtain \(y\in\Lambda_{\omega}^{p}(\mathcal{M})^{\prime}= \Gamma_{ \widetilde{\omega} }^{q}(\mathcal{M})\subseteq L^{1}(\mathcal{M})\). Note that \(\tau(ya)=0\), \(\forall a\in\mathcal{A}\subseteq H^{p, \omega}(\mathcal{A})\). Then by Proposition 3.9 of [2] and Remark 2.1(i), we have \(y\in H^{1}_{0}(\mathcal{A})\) and \(\mathcal{E}(y)=0\). Set \(1\leq s<\alpha_{\Lambda_{\omega}^{p}}\), then \(x\in\{z\in L_{s}(\mathcal{M}): \tau(za)=0, \forall a\in \mathcal{A}_{0}\}=H^{s}(\mathcal{A})\). By Corollary 2.2 of [5], we deduce \(\tau(xy)=\tau(\mathcal{E}(xy))=\tau(\mathcal{E}(x)\mathcal {E}(y))=0\). This is a contradiction, and so the first equality holds.
A similar argument to the proof of above shows that
On the other hand, let \(x\in\{x\in\Lambda_{\omega}^{p}(\mathcal{M}): \tau(xy)=0, \forall y\in\mathcal{A}\}\). Since \(\mathcal{D}\subseteq\mathcal{A}\), then \(\tau(xy)=\tau (\mathcal{E}(x)y)=0\), \(\forall y\in\mathcal{D}\). It follows that \(\mathcal{E}(x)=0\). Since \(x\in\Lambda_{\omega}^{p}(\mathcal{M}) \subseteq L^{1}(\mathcal {A})\), it follows from Corollary 2.2 of [5] that \(\tau(xy)=\tau(\mathcal{E}(xy))=\tau(\mathcal{E}(x)\mathcal{E}(y))=0\) holds for all \(y\in\mathcal{A}_{0}\). This implies that \(x\in H^{p, \omega}(\mathcal{A})\), and so there exist \(x_{n}\in\mathcal{A}\), \(n=1, 2, \ldots \) , such that \(\|x-x_{n}\|_{p, \omega}\rightarrow0\), \(n\rightarrow\infty\). On the other hand, Remark 2.1(v) shows that ℰ is bounded on \(\Lambda_{\omega}^{p}(\mathcal{M})\). It follows from the fact \(\mathcal{E}(x)=0\) that
Since \(x_{n}-\mathcal{E}(x_{n})\in\mathcal{A}_{0}\), then \(x\in H^{p, \omega}_{0}(\mathcal{A})\), which implies the second equality holds.
For the third equality, it is clear that \(H^{r}(\mathcal{A})\cap\Lambda^{p}_{\omega}(\mathcal{M}) \supseteq H^{p, \omega}(\mathcal{A})\). Conversely, if \(r\geq1\) and \(x\in H^{r}(\mathcal{A})\cap\Lambda ^{p}_{\omega}(\mathcal{M})\), then \(x\in\{z\in\Lambda_{\omega}^{p}(\mathcal{M}): \tau(za)=0, \forall a\in\mathcal{A}_{0}\}\) since \(H^{r}(\mathcal{A})=\{z\in L_{r}(\mathcal{A}): \tau(za)=0, \forall a\in\mathcal{A}_{0}\}\) and \(L_{r}(\mathcal{A})\supseteq\Lambda_{\omega}^{p}(\mathcal{M})\). By the first equality, we have \(x\in H^{p, \omega}(\mathcal{A})\). On the other hand, if \(r<1\) and \(x\in H^{r}(\mathcal{A})\cap\Lambda^{p}_{\omega}(\mathcal {M})\), there exists \(1\leq s<\alpha_{\Lambda_{\omega}^{p}}\) such that \(\Lambda_{\omega}^{s}(\mathcal{M})\subseteq L_{s}(\mathcal{M})\), and so, by Proposition 3.3 of [5], we have \(H^{r}(\mathcal{A})\cap\Lambda_{\omega}^{p}(\mathcal{M})= H^{r}(\mathcal{A})\cap L_{s}(\mathcal{M})\cap\Lambda_{\omega}^{p}(\mathcal{M})= H^{s}(\mathcal{A})\cap\Lambda_{\omega}^{p}(\mathcal{M})= H^{p, \omega}(\mathcal{A})\). Similarly, the fourth equality holds. □
Theorem 3.1
Let \(0< p, q<\infty\) and \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\). If \(x\in\Lambda_{\omega}^{p}(\mathcal{M})\) is an invertible operator such that \(x^{-1}\in\Lambda_{\omega}^{q}(\mathcal{M})\), then there exist a unitary \(u\in\mathcal{M}\) and \(h\in H^{p, \omega }(\mathcal{A})\) such that \(x=uh\), \(h^{-1}\in H^{q, \omega}(\mathcal{A})\).
Proof
By the fact \(\alpha_{\Lambda_{\omega}^{rp}}=r\alpha_{\Lambda_{\omega}^{p}}\), \(\beta_{\Lambda_{\omega}^{rp}}=r\beta_{\Lambda_{\omega}^{p}}\), \(r>0\) and \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\), we have \(0<\alpha_{\Lambda_{\omega}^{q}}\leq\beta_{\Lambda_{\omega}^{q}}<\infty\). First, let \(1< p, q\leq\infty\), \(\alpha_{\Lambda_{\omega}^{p}}>1\), \(\alpha _{\Lambda_{\omega}^{q}}>1\), and let \(x\in\Lambda_{\omega}^{p}(\mathcal{M})\) with \(x^{-1}\in\Lambda_{\omega}^{q}(\mathcal{M})\). Take \(r_{1}\), \(r_{2}\) with \(1\leq r_{1}<\alpha_{\Lambda_{\omega}^{p}}\), \(1\leq r_{2}<\alpha_{\Lambda_{\omega}^{q}}\). By Theorem 3.1 of [5], there exist a unitary \(u\in\mathcal{M}\) and \(h\in H^{r_{1}}(\mathcal{A})\) such \(h=u^{*}x\) and \(h^{-1}\in H^{r_{2}}(\mathcal{A})\). It is clear that \(h=u^{*}x\in\Lambda_{\omega}^{p}(\mathcal{M})\). It follows from Proposition 3.1 that \(h\in H^{p, \omega}(\mathcal{A})\). Similarly, \(h^{-1}\in H^{q, \omega}(\mathcal{A})\).
On the other hand, if \(\min(p, q)\leq1\) or \(\min(\alpha_{\Lambda_{\omega}^{p}}, \alpha_{\Lambda_{\omega}^{p}})\leq1\), there exists an integer n such that \(\min(np, nq)>1\) and \(\min(n\alpha_{\Lambda_{\omega}^{p}}, n\alpha_{\Lambda_{\omega}^{q}})= \min(\alpha_{\Lambda_{\omega}^{np}}, \alpha_{\Lambda_{\omega}^{nq}})>1\). Let \(x=v|x|\) be the polar decomposition of x. We write \(x_{1}=v|x|^{\frac{1}{n}}\), \(x_{k}=|x|^{\frac{1}{n}}\), \(2\leq k\leq n\), then \(x=x_{1}x_{2}\cdots x_{n}\). Thus, \(x_{k}\in\Lambda_{\omega}^{np}(\mathcal{M})\) and \(x_{k}^{-1} \in\Lambda_{\omega}^{nq}(\mathcal{M})\). Therefore, there exist \(u_{n}\in\mathcal{M}\) and \(h_{n}\in H^{np, \omega}(\mathcal{A})\) such that \(x_{n}=u_{n}h_{n}\) and \(x_{n}^{-1}\in H^{nq, \omega}(\mathcal{A})\). Repeating this argument, we again get the same factorization for \(x_{n-1}u_{n}\): \(x_{n-1}u_{n}=u_{n-1}h_{n-1}\), and then for \(x_{n-2}u_{n-1}\), and so on. In this way, we obtain the factorization \(x=uh_{1}\cdots h_{n}\), \(u\in\mathcal{M}\), where \(u\in\mathcal{M}\) is a unitary and \(h_{k}\in H^{np, \omega}(\mathcal{A})\) such that \(h_{k}^{-1}\in H^{nq, \omega}(\mathcal{A})\). We write \(h=h_{1}\cdots h_{n}\), then \(x=uh\) is the desired factorization. □
Remark 3.1
-
(i)
Let \(x\in\Lambda_{\omega}^{\infty}(\mathcal{M})=\mathcal{M}\) be an invertible operator such that \(x^{-1}\in\Lambda_{\omega}^{\infty}(\mathcal{M})\). Theorem 3.1 of [5] implies that there exist a unitary \(u\in\mathcal{M}\) and \(h\in \mathcal{A}\) such that \(x=uh\), \(h^{-1}\in\mathcal{A}\).
-
(ii)
Let \(0< p<\infty\) and \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta _{\Lambda_{\omega}^{p}}<\infty\). If \(x\in\Lambda_{\omega}^{p}(\mathcal{M})\) is an invertible operator such that \(x^{-1}\in\Lambda_{\omega}^{\infty}(\mathcal{M})=\mathcal{M}\), then there exist a unitary \(u\in\mathcal{M}\) and \(h\in H^{p, \omega }(\mathcal{A})\) such that \(x=uh\), \(h^{-1}\in\mathcal{A}\).
Indeed, if \(1< p<\infty\), \(1<\alpha_{\Lambda_{\omega}^{p}}\), take \(r>0\) with \(1\leq r<\alpha_{\Lambda_{\omega}^{p}}\). By Theorem 3.1 of [5], there exist a unitary \(u\in\mathcal{M}\) and \(h\in H^{r}(\mathcal{A})\) such that \(h=u^{*}x\) and \(h^{-1}\in \mathcal{A}\). Since \(h=u^{*}x\in\Lambda_{\omega}^{p}(\mathcal{M})\), it follows from Proposition 3.1 that \(h\in H^{p, \omega}(\mathcal{A})\). Therefore, by adapting the proof of Theorem 3.1, we complete the proof.
Proposition 3.2
Let \(0< p<q<\infty\) and \(0<\alpha_{\Lambda_{\omega}^{q}}\leq\beta_{\Lambda_{\omega}^{q}}<\infty\). Then
Proof
Since \(0< p<q<\infty\), then \(\Lambda_{\omega}^{q}(\mathcal{M})\subseteq\Lambda_{\omega}^{p}(\mathcal{M})\), which implies that \(H^{p, \omega}(\mathcal{A})\cap\Lambda_{\omega}^{q}(\mathcal{M})\supseteq H^{q, \omega}(\mathcal{A})\). Conversely, let \(x\in H^{p, \omega}(\mathcal{A})\cap\Lambda_{\omega}^{q}(\mathcal{M})\). We write \(a=(x^{*}x+1)^{\frac{1}{2}}\), then \(a\in\Lambda_{\omega}^{q}(\mathcal{M})\) and \(a^{-1}\in\mathcal{M}\). Applying Remark 3.1(ii) to a, there exist \(u\in\mathcal{M}\) and \(h\in H^{q, \omega}(\mathcal{A})\) such that \(a=uh\) and \(h^{-1}\in\mathcal{A}\). Since \(h^{*}h=x^{*}x+1\), then there exists \(v\in\mathcal{M}\) with \(\|v\|\leq1\) such that \(x=vh\). Since \(\alpha_{\Lambda_{\omega}^{rp}}=r\alpha_{\Lambda_{\omega}^{p}}\), \(r>0\) and \(0<\alpha_{\Lambda_{\omega}^{q}}\), then there exists \(0<\alpha_{\Lambda_{\omega}^{p}}\). Let \(0< r<\alpha _{\Lambda_{\omega}^{p}}\). By Remark 2.1(iii), we deduce \(x\in H^{p, \omega}(\mathcal{A})\subseteq H^{r}(\mathcal{A})\). Thus \(v=xh^{-1}\in H^{p, \omega}(\mathcal{A})\subseteq H^{r}(\mathcal{A})\). It follows from Proposition 3.3 of [5] that \(v\in\mathcal{A}\). Therefore, \(x\in\mathcal{A}\cdot H^{q, \omega}(\mathcal{A})=H^{q, \omega}(\mathcal{A})\). Similarly, the second equality holds. □
Theorem 3.2
Let \(0< p<\infty\), \(0< q\leq\infty\) and \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\). Then, for every \(x\in H^{p, \omega}(\mathcal{A})\) and \(\varepsilon>0\), there exist \(y\in H^{q, \omega}(\mathcal{A})\) and \(z\in H^{r, \omega}(\mathcal{A})\) such that \(x=yz\), where \(\frac{1}{p}=\frac{1}{q}+\frac{1}{r}\).
Proof
The case where \(\max\{q, r\}=\infty\) is trivial. Thus we assume \(\max\{q, r\}<\infty\). Let \(x\in H^{p, \omega}(\mathcal{A})\) and \(\varepsilon>0\). We write \(a=(x^{*}x+\varepsilon)^{\frac{1}{2}}\), then \(x\in\Lambda _{\omega}^{p}(\mathcal{M})\) and \(a^{-1}\in\mathcal{M}\). Let \(v\in\mathcal{M}\) be a contraction operator such that \(x=va\). Now applying Remark 3.1(ii) to \(a^{\frac{p}{r}}\), we have \(a^{\frac{p}{r}}=uz\), \(z^{-1}\in\mathcal{M}\), where u is a unitary in ℳ and \(z\in H^{r, \omega }(\mathcal{A})\). Let \(y=va^{\frac{p}{q}}u\in\Lambda_{\omega}^{q}(\mathcal{M})\), then \(x=yz\) and so \(y=xz^{-1}\). Since \(x\in H^{p, \omega}(\mathcal{A})\) and \(z^{-1}\in \mathcal{A}\), we have \(y\in H^{p, \omega}(\mathcal{M})\). By Proposition 3.2 and the fact \(p< q\), we obtain \(y\in H^{q, \omega}(\mathcal{A})\). □
Remark 3.2
Let \(0< p, q\leq\infty\), \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty \) and \(\frac{1}{p}=\frac{1}{q}+\frac{1}{r}\). For every \(x\in H^{p, \omega}(\mathcal{A})\) and \(\varepsilon>0\), let y, z be as in Theorem 3.2, then \(\|y\|_{q, \omega}\|z\|_{r, \omega}\leq C(1+\varepsilon)\|x\|_{p, \omega}\) for some constant C (independent of x). Indeed, the case \(p=\infty\) follows from Theorem 3.4 of [5] since \(\Lambda_{\omega}^{\infty}(\mathcal {M})=\mathcal{M}\). The other case follows from Theorem 3.2.
4 Outer operators according to \(H^{p,\omega}(\mathcal{A})\)
Let x be a τ-measurable operator with \(\lim_{t\rightarrow \infty}\mu_{t}(x)=0\). The Fuglede-Kadison determinant \(\Delta(x)\) is defined by
where \(d\nu_{|x|}\) denotes the probability measure on \(\mathbb{R}^{+}\) which is obtained by composing the spectral measure of \(|x|\) with the trace τ. We refer the reader to [1, 8] for more information on determinant.
Proposition 4.1
Let \(0< p<q\leq\infty\), \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\) and \(h\in H^{q}_{\omega}(\mathcal{A})\), then
-
(i)
\([h\mathcal{A}]_{p, \omega}= H^{p,\omega}(\mathcal{A})\) if and only if \([h\mathcal{A}]_{q, \omega}= H^{q,\omega}(\mathcal{A})\).
-
(ii)
\([\mathcal{A}h]_{p, \omega}= H^{p,\omega}(\mathcal{A})\) if and only if \([\mathcal{A}h]_{q, \omega}= H^{q,\omega}(\mathcal{A})\).
-
(iii)
\([\mathcal{A}h\mathcal{A}]_{p, \omega}= H^{p,\omega}(\mathcal{A})\) if and only if \([\mathcal{A}h\mathcal{A}]_{q, \omega}= H^{q,\omega}(\mathcal{A})\).
Proof
We shall prove only the third equivalence. The other cases are similar. Since \(p< q\), then \(H^{q,\omega}(\mathcal{A})\) is dense in \(H^{p,\omega}(\mathcal{A})\). It follows from \([\mathcal{A}h\mathcal{A}]_{\Lambda^{q}_{\omega}(\mathcal{M})}= H^{q,\omega}(\mathcal{A})\) that \([\mathcal{A}h\mathcal{A}]_{\Lambda^{p}_{\omega}(\mathcal{M})}= H^{p,\omega}(\mathcal{A})\). Conversely, since \(\alpha_{\Lambda_{\omega}^{p}}>0\), there is \(r>0\) such that \(\Lambda_{\omega}^{p}(\mathcal{M})\subseteq L_{r}(\mathcal{M})\), and so \([\mathcal{A}h\mathcal{A}]_{r}= H^{r}(\mathcal{A})\). Using Proposition 4.1 of [5], we have \([\mathcal{A}h\mathcal{A}]_{\infty}= H^{\infty}(\mathcal{A})\), which means that \([\mathcal{A}h\mathcal{A}]_{q, \omega}= H^{q,\omega}(\mathcal{A})\). □
Definition 4.1
Let \(0< p<\infty\) and \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta _{\Lambda_{\omega}^{p}}<\infty\). An operator \(h\in H^{p,\omega}(\mathcal{A})\) is called left outer, right outer or bilaterally outer according to \([h\mathcal{A}]_{p, \omega}= H^{p,\omega}(\mathcal{A})\), \([\mathcal{A}h]_{p, \omega}= H^{p,\omega}(\mathcal{A})\) or \([\mathcal{A}h\mathcal{A}]_{p, \omega}= H^{p,\omega}(\mathcal{A})\).
Remark 4.1
-
(i)
Proposition 4.1 justifies the relative independence of the index p in Definition 4.1.
-
(ii)
Since \(\Lambda_{\omega}^{\infty}(\mathcal{M})=\mathcal{M}\), Proposition 4.1 of [5] and Proposition 4.1 imply that Definition 4.1 coincides with the definition in the sense of [5].
Proposition 4.2
Let \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\) and \(h\in H^{p,\omega}(\mathcal{A})\).
-
(i)
If h is left or right outer, then \(\Delta(h)= \Delta(\mathcal {E}(h))\). Conversely, if \(\Delta(h)= \Delta(\mathcal{E}(h))\) and \(\Delta(h)> 0\), then h is left and right outer (so bilaterally outer too).
-
(ii)
If \(\mathcal{A}\) is antisymmetric and h is bilaterally outer, then \(\Delta(h)= \Delta(\mathcal{E}(h))\).
Proof
Since \(\alpha_{\Lambda_{\omega}^{p}}>0\), then there is \(r>0\) such that \(H^{p,\omega}(\mathcal{A})\subseteq H^{r}(\mathcal{A})\). Applying Theorem 4.4 of [5] to \(h\in H^{p,\omega}(\mathcal {A})\subseteq H^{r}(\mathcal{A})\), we obtain that (i) and (ii) hold. □
In the classical function algebra setting, one assumes that \(\mathcal{D}=\mathcal{A}\cap J\mathcal{A}\) is one-dimensional, which forces \(\mathcal{E}=\tau(\cdot)1\). If in our setting this is the case, then we say that \(\mathcal{A}\) is antisymmetric. It is worth remarking that the antisymmetric maximal subdiagonal subalgebras of commutative von Neumann algebras are precisely the weak* Dirichlet algebras. It is clear that \(\mathcal{A}\) is antisymmetric if and only if \(\dim\mathcal{D}= 1\) (equivalently, \(\mathcal {D}=\mathbb{C}1\)).
The following corollary is a consequence of Proposition 4.2.
Corollary 4.1
Let \(h\in H^{p,\omega}(\mathcal{A})\), \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\).
-
(i)
If \(\Delta(h)> 0\), then h is left outer if and only if h is right outer.
-
(ii)
Assume that \(\mathcal{A}\) is antisymmetric, then the following properties are equivalent:
-
(a)
h is left outer;
-
(b)
h is right outer;
-
(c)
h is bilaterally outer;
-
(d)
\(\Delta(\mathcal{E}(h))= \Delta(h)> 0\).
-
(a)
We will say that h is outer if it is at the same time left and right outer. If \(h\in H^{p}_{\omega}(\mathcal{A})\) with \(\Delta(h)> 0\), then h is outer if and only if \(\Delta(h)= \Delta(\mathcal{E}(h))\). Also in the case that \(\mathcal{A}\) is antisymmetric, h with \(\Delta (h)> 0\) is outer if and only if it is left, right or bilaterally outer.
The following corollary is an immediate consequence of Corollary 4.7 of [5] and the proof of Proposition 4.1.
Corollary 4.2
Let \(0< p, q\leq\infty\) and \(0<\alpha _{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\). If \(h\in H^{p,\omega}(\mathcal{A})\) with \(h^{-1}\in H^{q,\omega }(\mathcal{A})\), then h is outer.
Theorem 4.1
Let \(0< p<\infty\) and \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta _{\Lambda_{\omega}^{p}}<\infty\). If \(x\in\Lambda^{p}_{\omega}(\mathcal{M})\) with \(\Delta(x)> 0\), then there exist a unitary \(u\in\mathcal{M}\) and an outer \(h\in H^{p,\omega}(\mathcal{A})\) such that \(x=uh\).
Proof
First, let \(\alpha_{\Lambda_{\omega}^{p}}>1\) and \(p>1\), there is \(\alpha _{\Lambda_{\omega}^{p}}>r>1\) such that \(\mathcal{M}\subseteq\Lambda^{p}_{\omega}(\mathcal{M})\subseteq L_{r}(\mathcal{M})\) and \(\mathcal{A}\subseteq H^{p, \omega}(\mathcal{A})\subseteq H^{r}(\mathcal{A})\). Applying Theorem 4.8 of [5] to \(x\in\Lambda^{p}_{\omega}(\mathcal{M})\subseteq L_{r}(\mathcal{M})\), there exist a unitary \(u\in\mathcal{M}\) and an outer \(h\in H^{r}(\mathcal{A})\) such that \(x=uh\). It follows from Proposition 3.1 that \(h=u^{*}x\in\Lambda^{p}_{\omega}(\mathcal{M})\cap H^{r}(\mathcal {A})=H^{p, \omega}(\mathcal{M})\). By adapting the proof of Theorem 3.1, we complete the proof. □
Corollary 4.3
Let \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\) and \(x\in\Lambda^{p}_{\omega}(\mathcal{M})\) with \(0< p<\infty\) such that \(\Delta(x)> 0\), then there exist a unitary \(u\in\mathcal{A}\) (inner) and an outer \(h\in H^{p,\omega}(\mathcal{A})\) such that \(x=uh\).
Proof
Let \(\alpha_{\Lambda_{\omega}^{p}}>1\) and \(p>1\), there exists \(\alpha _{\Lambda_{\omega}^{p}}>r>1\) such that \(\mathcal{M}\subseteq\Lambda^{p}_{\omega}(\mathcal{M})\subseteq L_{r}(\mathcal{M})\) and \(\mathcal{A}\subseteq H^{p, \omega}(\mathcal{A})\subseteq H^{r}(\mathcal{A})\). Applying Corollary 4.9 of [5] to \(x\in\Lambda^{p}_{\omega}(\mathcal{M})\subseteq L_{r}(\mathcal{M})\), there exist a unitary \(u\in\mathcal{A}\) and an outer \(h\in H^{r}(\mathcal{A})\) such that \(x=uh\). Thus, Proposition 3.1 implies that \(h=u^{*}x\in\Lambda^{p}_{\omega}(\mathcal{M})\cap H^{r}(\mathcal {A})=H^{p, \omega}(\mathcal{M})\). By adapting the end of the proof of Theorem 3.1, we get the desired factorization of x. For simplicity we consider only the case \(\alpha_{\Lambda_{\omega}^{p}}>\frac{1}{2}\) and \(p>\frac{1}{2}\). Let \(x=v|x|\) be the polar decomposition of x. Using a similar discussion of above to \(|x|^{\frac{1}{2}}\), there exist a unitary \(u_{1}\in\mathcal{A}\) and an outer \(h_{1}\in H^{2p,\omega}(\mathcal{A})\) such that \(|x|^{\frac{1}{2}}=u_{1}h_{1}\) since \(\Delta(|x|^{\frac{1}{2}})>0\). Similarly, we have \(v|x|^{\frac{1}{2}}u_{1}=u_{2}h_{2}\), where \(u_{2}\in\mathcal{A}\) and an outer \(h_{2}\in H^{2p,\omega }(\mathcal{A})\). Therefore, \(u=u_{2}\) and \(h=h_{2}h_{1}\) yield the desired factorization of x. □
Corollary 4.4
Let \(0< p<\infty\), \(0<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda _{\omega}^{p}}<\infty\) and \(h\in H^{p,\omega}(\mathcal{A})\) with \(\Delta(h)> 0\), then h is outer if and only if for any \(x\in H^{p,\omega}(\mathcal {A})\) with \(|x|= |h|\), we have \(\Delta(\mathcal{E}(x))\leq\Delta(\mathcal{E}(h))\).
Proof
Let h be outer and \(x\in H^{p,\omega}(\mathcal{A})\) with \(|x|= |h|\). Taking \(0< r<\alpha_{\Lambda_{\omega}^{p}}\), we obtain that \(x\in H^{r}(\mathcal{A})\). From Corollary 4.11 of [5], we get \(\Delta(\mathcal{E}(x))\leq\Delta(\mathcal{E}(h))\). Conversely, since \(0<\alpha_{\Lambda_{\omega}^{p}}\), there is \(0< r<\alpha_{\Lambda_{\omega}^{p}}\) such that \(H^{p, \omega}(\mathcal{A})\subseteq H^{r}(\mathcal{A})\). From Corollary 4.11 of [5] and Remark 4.1(i), we obtain that h is outer. □
5 Jordan morphism
Let x be an operator, we write \(\operatorname{Re}x=\frac{x+x^{*}}{2}\) and \(\operatorname{Im}x=\frac{x-x^{*}}{2i}\). If \(u\in \operatorname{Re}\mathcal{A}\), then \(u=\operatorname{Re}x\) for some \(x\in\mathcal{A}\). Let \(a=x-\frac{1}{2}\mathcal{E}(x-x^{*})\), it is clear that \(a\in\mathcal{A}\), \(u=\operatorname{Re}a\) and \(\mathcal{E}(\operatorname{Im}a)=0\). Therefore, there exists \(\widetilde{u}=\operatorname{Im}a\in \operatorname{Re}\mathcal{A} \) such that \(a=u+i\widetilde{u}\in\mathcal{A}\) and \(\mathcal{E}(\widetilde{u})=\mathcal{E}(\operatorname{Im}a)=0\). By a similar discussion of [2], we have \(\widetilde{u}\in \operatorname{Re}\mathcal{A}\) is unique. Thus, we can define \(\widetilde{u}=\operatorname{Im}a\), where \(a\in\mathcal{M}\) is the unique element of ℳ with \(u=\operatorname{Re}a\) and \(\mathcal{E}(\operatorname{Im}a)=0\). It is obvious that \(\sim: u\mapsto\widetilde{u}\) is real linear. We called \(\widetilde{u}\) the conjugate of u. We define the Herglotz transform \(H: \Lambda_{\omega}^{p}(\mathcal{M}) \rightarrow\Lambda_{\omega}^{p}(\mathcal{M})\) by \(H(u)=u+i\widetilde{u}\). It is clear that H is real linear.
Theorem 5.1
Let \(\Lambda_{\omega}^{p}\) be a fully symmetric quasi-Banach function space and \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\). The real linear maps
extend to real linear maps
If \(x\in\Lambda_{\omega}^{p}(\mathcal{M})^{sa}\), then \(H(x)=x+i\widetilde{x}\in H^{p, \omega}(\mathcal{A})\) and \(\mathcal{E}(\widetilde{x})=0\). Both ∼ and H are bounded.
Proof
Let \(r_{1}, r_{2}>0\) with \(1< r_{1}<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<r_{2}< \infty\). By Theorem 5.4 of [2], we have that ∼ and H extend to bounded real linear maps
and \(\mathcal{E}(\widetilde{x})=0\), \(H(x)=x+i\widetilde{x}\in H^{r_{i}}(\mathcal{A})\), \(i=1, 2\). Now consider the standard complexification \(\overline{\thicksim}: L^{r_{i}}(\mathcal{M}) \rightarrow L^{r_{i}}(\mathcal{M})\), \(i=1, 2\), that is, \(\overline{\widetilde{x}}=\widetilde{\operatorname{Re}x} +i\widetilde{\operatorname{Im}x}\). From the method in [2], we know that \(x, y\prec x+iy\), \(x, y\in L^{r_{i}}(\mathcal{M})^{sa}\), \(i=1, 2\). Thus,
and so \(\overline{\thicksim}\) is bounded on \(L^{r_{i}}(\mathcal{M})\), \(i=1, 2\). Since \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty\), it follows from Theorem 3.7 of [12] that \(\Lambda_{\omega}^{p}(\mathcal{M})\) is an interpolation space for the couple \((L^{r_{1}}(\mathcal{M}), L^{r_{2}}(\mathcal{M}))\), which implies that \(\overline{\thicksim}\) is bounded on \(\Lambda_{\omega}^{p}(\mathcal{M})\). Thus, both ∼ and H are bounded. By the discussion of above we know that \(H(x)\in H^{r_{1}}(\mathcal{M})\), \(\forall x\in L^{r_{1}}(\mathcal{M})^{sa}\). Since \(\Lambda_{\omega}^{p}(\mathcal{M})\subseteq L^{r_{1}}(\mathcal{M})\) and H is bounded, we have \(H(x)\in H^{r_{1}}(\mathcal{M})\), \(\mathcal{E}(\widetilde{x})=0\) and \(H(x)\in\Lambda_{\omega}^{p}(\mathcal{M})\) hold for all \(x\in\Lambda_{\omega}^{p}(\mathcal{M})^{sa}\). Combining this with Proposition 3.1, we obtain \(H(x)\in H^{p, \omega}(\mathcal{A})\). □
Using the same method as that in Theorem 6.2 of [2], we obtain the following result.
Theorem 5.2
Let \(\Lambda_{\omega}^{p}\) be a fully symmetric quasi-Banach function space and \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda_{\omega}^{p}}<\infty \), then
The relevant projections are \(x\mapsto\frac{1}{2}[x+i\overline{\widetilde{x}}-\mathcal{E}(x)]\); \(x\mapsto\mathcal{E}(x)\); \(x\mapsto\frac{1}{2}[x-i\overline{\widetilde{x}}-\mathcal{E}(x)]\), where \(\overline{\thicksim}:\Lambda_{\omega}^{p}(\mathcal{M}) \rightarrow\Lambda_{\omega}^{p}(\mathcal{M})\) is the standard complexification of ∼, that is, \(\overline{\widetilde{x}}= \widetilde{\operatorname{Re}x}+i\widetilde{\operatorname{Im}x}\).
Let \(B_{1}\), \(B_{2}\) be two Banach algebras, a linear map \(\varphi: B_{1}\rightarrow B_{2}\) is a homomorphism if \(\varphi(ab)=\varphi(a)\varphi(b)\), \(a, b \in B_{1}\). We say that a linear map \(\varphi: B_{1}\rightarrow B_{2}\) is an anti-morphism if \(\varphi(ab)=\varphi(b)\varphi(a)\), \(a, b \in B_{1}\). Given a unital Banach algebra B with identity 1, under the term irreducible representation of B, we understand a continuous homomorphism \(\pi: B\rightarrow\mathcal{B}(X)\), where \(\mathcal{B}(X)\) is the set of all bounded linear operators on some Banach space X such that \(\pi(B)\) is an irreducible subalgebra of \(\mathcal{B}(X)\) in the sense of admitting only trivial invariant subspaces. A Jordan morphism between two von Neumann algebras \(\mathcal{M}_{1}\) and \(\mathcal{M}_{2}\) is a linear mapping \(\varphi: \mathcal{M}_{1}\rightarrow\mathcal {M}_{2}\) which preserves the Jordan product, i.e., \(\varphi(ab+ba)=\varphi(a)\varphi (b)+\varphi(b)\varphi(a)\) for all \(a, b\in\mathcal{M}_{1}\). For further information, we refer the reader to [7].
Lemma 5.1
Let \(1\leq p<\infty\) and let \(\varphi:\mathcal{A}\rightarrow\mathcal{A}\) be a Jordan morphism such that \(\pi\circ\varphi\) is either a morphism or an anti-morphism for each irreducible representation π of ℳ. If φ extends to a bounded map \(\varphi: H^{p, \omega}(\mathcal{A})\rightarrow H^{p, \omega }(\mathcal{A})\) with norm \(\|\varphi\|_{p, \omega}\), then for any integer \(0< k\leq p\), φ extends to a map \(\varphi: H^{\frac{p}{k}, \omega}(\mathcal{A})\rightarrow H^{\frac {p}{k}, \omega}(\mathcal{A})\) with norm not exceeding \(2C^{2k-1}(\|\varphi\|_{p, \omega})^{k}\), where C is a constant.
Proof
Let \(\varepsilon>0\) and \(x\in\mathcal{A}\) be given. Using Remark 3.2, we obtain \(h_{1}, \ldots, h_{k}\in\mathcal{A}\) so that \(x=h_{1}h_{2}\cdots h_{k}\) with
By Lemma 4.5 of [7], we have
Thus, by the fact that \(\Lambda_{\omega}^{p}\) is a fully symmetric quasi-Banach function space, we deduce
This implies that the required estimate holds. □
Lemma 5.2
Let \(\varphi:\mathcal{A}\rightarrow\mathcal{A}\) be a continuous identity-preserving Jordan morphism extending continuously to a map \(\varphi: H^{p, \omega}(\mathcal{A})\rightarrow H^{p, \omega }(\mathcal{A})\), \(1\leq p<\infty\) with norm \(\|\varphi\|_{p, \omega}\). For any \(k\in \mathbb{N} \) and any \(x\in\mathcal{A}\) such that \(\varphi(x)\) is normal, we have \(\|\varphi(x)\|_{pk, \omega}\leq(\|\varphi\|_{p, \omega})^{\frac {1}{k}}\|x\|_{pk, \omega}\). Let \(1< p<\infty\) and \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta _{\Lambda_{\omega}^{p}}<\infty\). If, in addition, φ is contractive on \(\mathcal{A}\) and \(\Lambda_{\omega}^{p}\) is a fully symmetric quasi-Banach function space, it extends to a unique bounded hermitian map \(\widetilde{\varphi}\) on \(\Lambda_{\omega}^{p}(\mathcal{M})\) which is positive on \(\mathcal{A}+J\mathcal{A}\subseteq\Lambda _{\omega}^{p}(\mathcal{M})\). If \(\widetilde{\varphi}\) is positive on all of \(\Lambda_{\omega}^{p}(\mathcal{M})\), we get
where C is a constant. If, moreover, \(\widetilde{\varphi}\) satisfies the condition
we obtain the estimate
where C is a constant.
Let B be a linear complement of \(\mathcal{D}\) in \(\mathcal{A}\). It is well known that \(J\mathcal{A}=\mathcal{D}\oplus JB \) and hence that \(\mathcal{A}+J\mathcal{A} =B\oplus\mathcal{D}\oplus JB\). Let \(\varphi:\mathcal{A}\rightarrow\mathcal{M}\) be a contractive identity-preserving Jordan morphism, we may extend φ to a map \(\widetilde{\varphi}\) on \(\mathcal {A}+J\mathcal{A}\) by setting \(\varphi=\widetilde{\varphi}\) on \(\mathcal{A}\) and defining the action on JB by \(\widetilde{\varphi}(x)=\varphi(x^{*})^{*}\), \(x\in JB\). Since φ is order-preserving on \(\mathcal{D}\), it follows that \(\widetilde{\varphi}\) preserves adjoints. Indeed, \(\widetilde{\phi}\) is order-preserving, positive and contractive on \(\mathcal {A}+J\mathcal{A}\). For further information, we refer the reader to [7].
Proof
Our proof follows Labuschagne’s argument of (Lemma 4.8, [7]). Given some \(x\in\mathcal{A}\) such that \(\varphi(x)\) is normal, then
Let \(1< p<\infty\), \(1<\alpha_{\Lambda_{\omega}^{p}}\leq\beta_{\Lambda _{\omega}^{p}}<\infty\), and let φ be an identity-preserving contractive Jordan morphism on \(\mathcal{A}\). From the preceding discussion (above the proof of this lemma), φ uniquely extends to a positive map on \(\mathcal{A}+J\mathcal{A}\). Moreover, φ then acts positively on the von Neumann algebra \(\mathcal{D}\). By Theorem 5.2, we have \(H^{p, \omega}(\mathcal{A})=H^{p, \omega}_{0}(\mathcal{A})\oplus \Lambda_{\omega}^{p}(\mathcal{D})\). Since \(\mathcal{D}\) is dense in \(\Lambda_{\omega}^{p}(\mathcal{D})\), then the continuous action of φ on \(H^{p, \omega}(\mathcal {A})\) (and hence also on \(L^{p}(\mathcal{M})\)) ensures that the unique extension of φ to \(H^{p, \omega }(\mathcal{A})\) acts positively on \(\Lambda_{\omega}^{p}(\mathcal{D})\). The extension of φ to a hermitian map \(\widetilde{\varphi}\) on \(\Lambda_{\omega}^{p}(\mathcal{M})\) can now be defined by
Since this construction canonically contains that construction in the discussion above this lemma, \(\widetilde{\varphi}\) constructed as (5.4) will restrict to \(\mathcal{A}+J\mathcal{A}\), which yields precisely the map in the preceding discussion (above the proof of this lemma). From Theorem 5.2, the \(\|\cdot\|_{p, \omega}\) boundedness of \(\widetilde{\varphi}\) follows from the \(\|\cdot\|_{p, \omega}\) boundedness of φ and of the conjugate linear map \(x\rightarrow x^{*}\). Let \(1\leq q<\infty\), we write \(|x|_{q}=(\frac{1}{2}(|x|^{q}+|x^{*}|^{q}))^{\frac{1}{q}}\). It is well known that \(|x|^{q}\leq2|x|_{q}^{q}\), and so \(|x|\leq2^{\frac{1}{q}}|x|_{q}\). If \(\widetilde{\varphi}\) acts positively on all of \(\Lambda_{\omega}^{p}(\mathcal{M})\), it will surely map \(\|\cdot\|_{\infty}\) boundedly ℳ into ℳ. By Lemma 7.3 of [28], we get
Since \(\widetilde{\varphi}(y)^{2}\leq\widetilde{\varphi}(y^{2})\) for each \(y\in\mathcal{M}^{+}\), we have
Now note that
which implies that
Combining this with (5.5), we get
Consequently, given \(x\in\mathcal{A}\), the positivity of \(\widetilde {\varphi}\) and the fact that \(|\varphi(x)|\leq2^{\frac{1}{2}}|\varphi(x)|_{2}\) imply that
The final inequality is a consequence of Lemma 2.1 of [29]. Observe that \(|x|\leq2^{\frac{1}{2^{k}}}|x|_{2^{k}}\), a similar proof using \(|x|_{2^{k}}\) instead of \(|x|_{2}\) will suffice in the case that \(\widetilde{\varphi}\) satisfies condition (5.2). □
Lemma 5.3
Let \(1< p\leq\infty\) be given and let \((q_{n})\) be a sequence of reals in \([1, p)\) increasing to p. Then \(x\in\Lambda_{\omega}^{p}(\mathcal{M})\) if and only if \(x\in\bigcap_{n=1}^{\infty}\Lambda_{\omega}^{q_{n}}\subseteq\Lambda _{\omega}^{p}(\mathcal{M})\), and \(\sup_{n}\|x\|_{q_{n}, \omega}<\infty\). In this case \(\|x\|_{p, \omega}=\lim_{n\rightarrow\infty}\|x\| _{q_{n}, \omega}\).
Proof
Let \(x\in\Lambda_{\omega}^{p}(\mathcal{M})\subseteq\Lambda_{\omega}^{q_{n}}(\mathcal{M})\), then
Thus, \(\limsup_{n\rightarrow\infty}\|x\|_{q_{n}, \omega}\leq\|x\|_{p, \omega}\). By the Levi lemma and the dominated convergence theorem, we have
□
Lemma 5.4
Let \(1\leq p<\infty\) be given and let \((q_{n})\) be a sequence of reals in \((p, \infty)\) decreasing to p. Given \(x\in\bigcup_{n=1}^{\infty}\Lambda_{\omega}^{q_{n}}\subseteq\Lambda _{\omega}^{p}(\mathcal{M})\), we have \(\|x\|_{p, \omega}=\lim_{n\rightarrow\infty}\|x\|_{q_{n}, \omega}\).
Proof
Let \(x\in \Lambda_{\omega}^{q_{n}}\subseteq\Lambda_{\omega}^{q_{n+1}}\subseteq\Lambda_{\omega}^{p}(\mathcal{M})\). Let \(E_{1}=\{t\in[0, 1]: \mu_{t}(x)\geq1\}\) and \(E_{2}=\{t\in[0, 1]: \mu _{t}(x)<1\}\), then \(\mu_{t}(x)^{p}\leq\mu_{t}(x)^{q_{n}}\), \(t\in E_{1} \) and \(\mu_{t}(x)^{p}\geq\mu_{t}(x)^{q_{n}}\), \(t\in E_{2}\). This implies that
and
which yield the desired result. □
Using the same method as that in Theorem 4.12 of [7], we obtain the following result.
Theorem 5.3
Let \(\Lambda_{\omega}^{p}\) be a fully symmetric quasi-Banach function space, and let \(\varphi:\mathcal{A}\rightarrow\mathcal{A}\) be a contractive identity-preserving Jordan morphism such that:
-
(a)
\(\pi\circ\varphi\) is either a homomorphism or an anti-morphism for each irreducible representation of \(\mathcal{M}\supseteq\mathcal{A}\); and
-
(b)
for some \(1< p<\infty\), φ extends to a bounded map \(\varphi:H^{p, \omega}(\mathcal{A})\rightarrow H^{p, \omega }(\mathcal{A})\).
If the canonical hermitian extension \(\widetilde{\varphi}\) to all of \(\Lambda_{\omega}^{p}(\mathcal{M})\) (see Lemma 5.2) is even positive, then, for any other \(1\leq q <\infty\), if \(1<\alpha_{\Lambda_{\omega}^{q}}\leq\beta_{\Lambda_{\omega}^{q}}<\infty\), φ will induce a unique closed operator \(\overline{\varphi}:D(\overline{\varphi}) \subseteq H^{q, \omega}(\mathcal{A})\rightarrow H^{q, \omega }(\mathcal{A})\) with the following properties:
-
(i)
\(D(\overline{\varphi})\supseteq\bigcup_{r>q}H^{r, \omega }(\mathcal{A})\);
-
(ii)
for any \(r>q\), \(\overline{\varphi}\) restricts to a bounded map \(\overline{\varphi}:H^{r, \omega}(\mathcal{A})\rightarrow H^{q, \omega}(\mathcal{A})\).
-
(iii)
for any \(q>s\geq1\), \(\overline{\varphi}\) extends uniquely to a bounded map \(\overline{\varphi}:H^{q, \omega}(\mathcal{A})\rightarrow H^{s, \omega}(\mathcal{A})\).
If \(\widetilde{\varphi}:\Lambda_{\omega}^{p}(\mathcal{M})\rightarrow \Lambda_{\omega}^{p}(\mathcal{M})\) satisfies the slightly stronger condition (5.2) in Lemma 5.2, then \(\overline{\varphi}:H^{q, \omega}(\mathcal{A})\rightarrow H^{q, \omega}(\mathcal{A})\) itself is a bounded everywhere defined operator.
References
Arveson, WB: Analyticity in operator algebras. Am. J. Math. 89, 578-642 (1967)
Marsalli, M, West, G: Noncommutative \(H^{p}\) spaces. J. Oper. Theory 40, 339-355 (1998)
Labuschagne, LE: A noncommutative Szegö theorem for subdiagonal subalgebras of von Neumann algebras. Proc. Am. Math. Soc. 133, 3643-3646 (2005)
Blecher, DP, Labuschagne, LE: Applications of the Fuglede-Kadison determinant: Szegö theorem and outers for noncommutative \(H^{p}\). Trans. Am. Math. Soc. 360, 6131-6147 (2008)
Bekjan, TN, Xu, Q: Riesz and Szegö type factorizations for noncommutative Hardy spaces. J. Oper. Theory 62, 215-231 (2009)
McAsey, M, Muhly, P, Saito, KS: Nonselfadjoint crossed products (invariant subspaces and maximality). Trans. Am. Math. Soc. 248, 381-410 (1979)
Labuschagne, LE: Analogues of composition operators on non-commutative \(H^{p}\)-spaces. J. Oper. Theory 49, 115-141 (2003)
Pisier, G, Xu, Q: Noncommutative \(L^{p}\)-spaces. In: Handbook of the Geometry of Banach Spaces, vol. 2, pp. 1459-1517 (2003)
Terp, M: \(L^{p}\) Spaces associated with von Neumann algebras. Notes, Math. Institute, Copenhagen University (1981)
Fack, T, Kosaki, H: Generalized s-numbers of τ-measurable operators. Pac. J. Math. 123, 269-300 (1986)
Carro, MJ, Raposo, JA, Soria, J: Recent developments in the theory of Lorentz spaces and weighted inequalities. Mem. Am. Math. Soc. 187, 877 (2007)
Dirksen, S, de Pagter, B, Potapov, D, Sukochev, F: Resenthal inequalities in noncommutative symmetric spaces. J. Funct. Anal. 261, 2890-2925 (2011)
Lindenstrauss, J, Tzafriri, L: Classical Banach Spaces. II. Springer, Berlin (1979)
Bergh, J, Löfström, J: Interpolation Spaces. Springer, New York (1976)
Kaminska, A, Maligranda, L: Order convexity and concavity of Lorentz spaces \(\Lambda_{p, \omega}\), \(0< p<\infty\). Stud. Math. 160, 267-286 (2004)
Jao, Y: Burkholder’s inequalities on noncommutative Lorentz martingale spaces. Proc. Am. Math. Soc. 138, 2431-2441 (2010)
Sukochev, F: Completeness of quasi-normed symmetric operator spaces. Indag. Math. 25, 376-388 (2014)
Guido, D, Isola, T: Singular traces on semifinite von Neumann algebras. J. Funct. Anal. 134, 451-485 (1995)
Dodds, PG, Dodds, TK, de Pagter, B: Fully symmetric operator spaces. Integral Equ. Oper. Theory 15, 942-972 (1992)
Kalton, NJ, Sukochev, FA: Symmetric norms and spaces of operators. J. Reine Angew. Math. 621, 81-121 (2008)
Watanabe, K: Some results on non-commutative Banach function spaces. II. Tohoku Math. J. 22, 349-364 (1993)
Strtil, S, Zsido, L: Lectures on von Neumann Algebras. Abacus Press, Tunbridge Wells (1979)
Dodds, PG, Dodds, TK, de Pagter, B: Noncommutative Köthe duality. Trans. Am. Math. Soc. 339, 717-750 (1993)
Exel, R: Maximal subdiagonal algebras. Am. J. Math. 110, 775-782 (1988)
Saito, KS: A note on invariant subspaces for finite maximal subdiagonal algebras. Proc. Am. Math. Soc. 77, 348-352 (1979)
Harada, T, Kosaki, H: Trace Jensen inequality and related weak majorization in semi-finite von Neumann algebras. J. Oper. Theory 63, 129-150 (2010)
Bennett, C, Sharpley, R: Interpolation of Operators. Pure and Applied Mathematics, vol. 129. Academic Press, New York (1988)
Størmer, E: Positive linear maps of operator algebras. Acta Math. 110, 233-278 (1963)
Dodds, PG, Dodds, TK, de Pagter, B: On p-convexity and q-concavity in non-commutative symmetric spaces. Integral Equ. Oper. Theory 78, 91-114 (2014)
Acknowledgements
The authors are grateful to the referee for useful suggestions. This work was partially supported by NSFC Grant No. 11371304 and NSFC Grant No. 11401507.
Author information
Authors and Affiliations
Corresponding author
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
The main idea of this paper was proposed by the corresponding author YZH. YZH and JJS prepared the manuscript initially and performed all the steps of the proofs in this research. All authors read and approved the final manuscript.
Rights and permissions
Open Access This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.
About this article
Cite this article
Han, Y., Shao, J. Some results on noncommutative Hardy-Lorentz spaces. J Inequal Appl 2015, 120 (2015). https://doi.org/10.1186/s13660-015-0642-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s13660-015-0642-3