betaFeedback

WorksMathematics

Erratum to Power boundedness and related properties for weighted composition operators on S(Rd)\mathscr{S}(\mathbb{R}^{d}), Journal of Functional Analysis 288 (2025), 110745

Abstract

As was kindly pointed out to us by S. Li, the proofs of Theorem 2.11 and Lemma 3.1 in our original paper contain errors that require modifications of the corresponding statements. We also correct a gap in the proof of Lemma 2.1, whose statement remains valid. Furthermore, we explain the consequences for several subsequent results and provide the necessary corrections to their statements and proofs.

Journal of Functional Analysis 288 (2025), 110745

Introduction

In [1], we studied power boundedness and related properties of weighted composition operators Cψ,φC_{\psi,\varphi} on the Schwartz space S(Rd)\mathcal{S}(\mathbb{R}^{d}), where Cψ,φ(f)=ψ⋅(f∘φ)C_{\psi,\varphi}(f)=\psi\cdot(f\circ\varphi) for f∈S(Rd)f\in\mathcal{S}(\mathbb{R}^{d}), with ψ∈C∞(Rd)\psi\in C^{\infty}(\mathbb{R}^{d}) and a smooth mapping φ:Rd→Rd\varphi:\mathbb{R}^{d}\to\mathbb{R}^{d}, the symbol of Cψ,φC_{\psi,\varphi}.

As kindly pointed out to us by S. Li, the proofs of Theorem 2.11 and Lemma 3.1 in [1] contain errors. We are grateful to S. Li for bringing these issues to our attention and for further investigating their consequences [2].

In the present erratum, we correct Theorem 2.11 by imposing an additional boundedness assumption on the mapping φ\varphi, and Lemma 3.1 by strengthening the original condition (3.3). We also provide a new proof of Lemma 2.1, whose statement remains valid despite a gap in the original proof.

These corrections affect several subsequent results. In particular, we amend Theorems 4.5 and 4.12, whose corrected statements correspond to Remarks 4.6 and 4.13, respectively, of [1]. Furthermore, we remove an incorrect assertion from Proposition 5.4, withdraw Proposition 5.5, and modify Theorems 5.8 and 5.10 accordingly. We also provide new proofs where required.

We are grateful to S. Li for bringing the errors in the proofs of Theorem 2.11 and Lemma 3.1 to our attention, and for sharing his further analysis with us [2]. In this erratum, we restrict ourselves to correcting the affected statements and proofs, without reproducing his additional results.

Corrected proof of Lemma 2.1

The proof of Lemma 2.1 in [1] incorrectly assumes that the points at which the relevant coefficient becomes large can be chosen to tend to infinity. This does not follow from hypothesis [1 (2.3)]. The lemma itself remains valid, as the following proof shows.

Replacement proof of Lemma 2.1. Fix α,λ∈N0\alpha,\lambda\in\mathbb{N}_{0} with ∣λ∣≤∣α∣|\lambda| \leq|\alpha| and p>0p > 0 satisfying [1 (2.3)]. Suppose, by contradiction, that for every f∈S(Rd)f \in\mathcal{S}(\mathbb{R}^{d}),

sup⁡i∈Isup⁡x∈Rd(1+∣x∣)p∣(ψi⋅(f∘φi))(α)(x)∣<∞.\sup_{i\in I}\sup_{x\in\mathbb{R}^{d}}(1+|x|)^{p}\left|(\psi_i\cdot(f\circ\varphi_i))^{(\alpha)}(x)\right|<\infty.

For i∈Ii\in I and x∈Rdx\in\mathbb{R}^{d} define the continuous linear functional

Λi,x(f)=(1+∣x∣)p(ψi⋅(f∘φi))(α)(x),f∈S(Rd).\Lambda_{i,x}(f)=(1+|x|)^{p}(\psi_i\cdot(f\circ\varphi_i))^{(\alpha)}(x),\qquad f\in\mathcal{S}(\mathbb{R}^{d}).

By assumption, the family (Λi,x)i,x(\Lambda_{i,x})_{i,x} is pointwise bounded. Since S(Rd)\mathcal{S}(\mathbb{R}^{d}) is a Fréchet space, the uniform boundedness principle gives C>0C>0 and an integer m≥∣α∣m\geq|\alpha| such that

∣Λi,x(f)∣≤C∥f∥m(i∈I, x∈Rd, f∈S(Rd)).(1)|\Lambda_{i,x}(f)|\leq C\|f\|_{m}\qquad(i\in I,\ x\in\mathbb{R}^{d},\ f\in\mathcal{S}(\mathbb{R}^{d})). \tag*{(1)}

where

∥f∥m=max⁡∣β∣≤msup⁡y∈Rd(1+∣y∣)m∣f(β)(y)∣.\|f\|_{m}=\max_{|\beta|\leq m}\sup_{y\in\mathbb{R}^{d}}(1+|y|)^{m}|f^{(\beta)}(y)|.

Choose g∈D(B(0,1))g\in\mathcal{D}(B(0,1)) such that

g(λ)(0)=1,g(β)(0)=0(∣β∣≤∣α∣, β≠λ).g^{(\lambda)}(0)=1,\qquad g^{(\beta)}(0)=0\qquad(|\beta|\leq|\alpha|,\ \beta\neq\lambda).

For each i∈I,x∈Rdi\in I,x\in\mathbb{R}^{d} put fi,x(y)=g(y−φi(x))f_{i,x}(y)=g(y-\varphi_i(x)). Translation yields a constant CmC_m independent of i,xi,x for which

∥fi,x∥m≤Cm(1+∣φi(x)∣)m.(2)\|f_{i,x}\|_{m}\leq C_m(1+|\varphi_i(x)|)^{m}. \tag*{(2)}

By formula [1 (2.2)] and the prescribed jet of gg,

(ψi⋅(fi,x∘φi))(α)(x)=Fα,λφi,ψi(x).(\psi_i\cdot(f_{i,x}\circ\varphi_i))^{(\alpha)}(x)=F_{\alpha,\lambda}^{\varphi_i,\psi_i}(x).

Applying (1) to fi,xf_{i,x} and then (2), we obtain

sup⁡i∈Isup⁡x∈Rd(1+∣x∣)p(1+∣φi(x)∣)m∣Fα,λφi,ψi(x)∣≤CCm<∞,\sup_{i\in I}\sup_{x\in\mathbb{R}^{d}}\frac{(1+|x|)^{p}}{(1+|\varphi_i(x)|)^{m}}\left|F_{\alpha,\lambda}^{\varphi_i,\psi_i}(x)\right|\leq CC_m<\infty,

which contradicts [1 (2.3)]. This proves the lemma. □

Amendment to Theorem 2.11

Theorem 2.11 is false as stated. Nevertheless, under the additional hypothesis that φ\varphi is bounded, the proof is correct and, under this additional assumption, is simplified. Accordingly, the statement is replaced here by its bounded-symbol version:

Let φ:Rd→Rd\varphi:\mathbb{R}^{d}\to\mathbb{R}^{d} be smooth and bounded. Then Cψ,φC_{\psi,\varphi} acts on S(Rd)\mathcal{S}(\mathbb{R}^{d}) for every ψ∈S(Rd)\psi\in\mathcal{S}(\mathbb{R}^{d}) if and only if φj∈OM(Rd)\varphi_j\in O_M(\mathbb{R}^{d}) for all 1≤j≤d1\leq j\leq d.

The restriction to bounded symbols is not intended to be optimal. For extensions and related statements we refer to [2].

Amendment to Lemma 3.1

The proof of Lemma 3.1 contains the same unjustified assumption as the original proof of Lemma 2.1, namely that the points witnessing the unboundedness in condition (3.3) of [1] can be chosen arbitrarily far from the origin. Unlike Lemma 2.1, however, Lemma 3.1 is false as stated. Its conclusion nevertheless remains valid if condition (3.3) is replaced by the following stronger condition: there is α∈N0d\alpha\in\mathbb{N}_{0}^{d} such that

lim sup⁡R→∞sup⁡i∈Isup⁡∣x∣>R∣φi(α)(x)∣(1+∣φi(x)∣)q=∞(3)\limsup_{R\to\infty}\sup_{i\in I}\sup_{|x|>R}\frac{|\varphi_i^{(\alpha)}(x)|}{(1+|\varphi_i(x)|)^q}=\infty \tag*{(3)}

for all q>0q>0.

Indeed, following the original proof up to equation (3.4) in [1], we choose α∈N0d\alpha\in\mathbb{N}_{0}^{d} with minimal ∣α∣|\alpha| such that (3.3′) holds for every q>0q>0. Again, ∣α∣≥1|\alpha|\geq1, and for every β∈N0d\beta\in\mathbb{N}_{0}^{d} with β≤α\beta\leq\alpha, β≠0\beta\ne0, there is q′(β)>0q^{\prime(\beta)}>0 such that

lim sup⁡R→∞sup⁡i∈Isup⁡∣x∣>R∣φi(α−β)(x)∣(1+∣φi(x)∣)q′(β)<∞.(4)\limsup_{R\to\infty}\sup_{i\in I}\sup_{|x|>R}\frac{|\varphi_i^{(\alpha-\beta)}(x)|}{(1+|\varphi_i(x)|)^{q^{\prime(\beta)}}}<\infty. \tag*{(4)}

The remainder of the proof proceeds as in [1] with the points xjx_j chosen sufficiently far from the origin so that the bounds provided by (3.5′) apply. Thus, the global estimates (3.5) in [1] are replaced by the corresponding estimates valid outside a sufficiently large ball.

We emphasize that conditions (3.1) and (3.3) of [1 Lemma 3.1] do not, in general, imply condition (3.3′). Consequently, the amendment does not restore the original statement or its subsequent applications without further verification.

However, we explicitly point out that, if I={1}I=\{1\}, then condition (3.3) of [1] and condition (3.3′) are equivalent. In particular, Theorem 3.3 of [1] remains valid.

Amendments to Sections 4 and 5

The amendment to Lemma 3.1 affects several results in Sections 4 and 5 of [1]. In the following, we describe the necessary modifications, provide corrected proofs where needed, and withdraw those assertions that cannot be maintained.

  1. Theorem 4.5. Due to the amendment to Lemma 3.1 and the fact that ψn,φ\psi^{n,\varphi} need not be of small decay with respect to φn\varphi_n, even when ψ\psi is of small decay with respect to φ\varphi, condition (ii)(b) of Theorem 4.5 does not, in general, follow from condition (i). Theorem 4.5 must therefore be replaced by the following statement, which is precisely the assertion of Remark 4.6 of [1]. We include a proof for the reader’s convenience.

Let φ:Rd→Rd\varphi:\mathbb{R}^{d}\to\mathbb{R}^{d} be a smooth mapping and let ψ∈C∞(Rd)\psi\in C^{\infty}(\mathbb{R}^{d}). Then, for the following conditions, (ii) implies (i), and (i) implies (ii)(a).

(i) Cψ,φC_{\psi,\varphi} acts on S(Rd)\mathcal{S}(\mathbb{R}^{d}) and is power bounded.

(ii) We have the following two conditions.

(a) For all p>0p>0 and α∈N0d\alpha\in\mathbb{N}_{0}^{d} there exists q>0q>0 such that

sup⁡n∈Nsup⁡x∈Rd(1+∣x∣)p(1+∣φn(x)∣)q∣(ψn,φ)(α)(x)∣<∞.\sup_{n\in\mathbb{N}}\sup_{x\in\mathbb{R}^{d}}\frac{(1+|x|)^p}{(1+|\varphi_n(x)|)^q}\left|(\psi^{n,\varphi})^{(\alpha)}(x)\right|<\infty.

(b) For all α∈N0d\alpha\in\mathbb{N}_{0}^{d} there exists q>0q>0 such that

sup⁡n∈Nsup⁡x∈Rd1(1+∣φn(x)∣)q∣φn(α)(x)∣<∞.\sup_{n\in\mathbb{N}}\sup_{x\in\mathbb{R}^{d}}\frac{1}{(1+|\varphi_n(x)|)^q}\left|\varphi_n^{(\alpha)}(x)\right|<\infty.

Proof. Suppose that (ii)(a) and (ii)(b) hold. By Corollary 2.4 of [1], Cψ,φC_{\psi,\varphi} acts on S(Rd)\mathcal{S}(\mathbb{R}^{d}). Moreover, (2.1) and (2.2) of [1] imply that {Cψ,φnf:n∈N}\{C_{\psi,\varphi}^{n}f:n\in\mathbb{N}\} is bounded in S(Rd)\mathcal{S}(\mathbb{R}^{d}) for every f∈S(Rd)f\in\mathcal{S}(\mathbb{R}^{d}). Thus, the Uniform Boundedness Principle yields the power boundedness of Cψ,φC_{\psi,\varphi}.

Conversely, suppose that (i) holds. Since Fα,0φn,ψn,φ=(ψn,φ)(α)F_{\alpha,0}^{\varphi^{n},\psi^{n,\varphi}}=(\psi^{n,\varphi})^{(\alpha)} by equation (2.1) of [1], Lemma 2.1 applied to the family (ψn,φ)n∈N(\psi^{n,\varphi})_{n\in\mathbb{N}} and to the smooth symbols (φn)n∈N(\varphi_n)_{n\in\mathbb{N}} yields (ii)(a). □

  1. Theorem 4.12. For the same reasons as in Theorem 4.5, condition (ii)(b) of Theorem 4.12 does not, in general, follow from condition (i). Thus, Theorem 4.12 must be replaced by the following statement, which is precisely the assertion of Remark 4.13 of [1]. Again, we include a proof for the reader’s convenience.

Let φ:Rd→Rd\varphi:\mathbb{R}^{d}\to\mathbb{R}^{d} be a smooth mapping and let ψ∈C∞(Rd)\psi\in C^{\infty}(\mathbb{R}^{d}). Then, for the following conditions, (ii) implies (i) and (i) implies (ii)(a).

(i) Cψ,φC_{\psi,\varphi} acts on S(Rd)\mathcal{S}(\mathbb{R}^{d}) and is topologizable (m-topologizable).

(ii) We have the following two conditions.

(a) For all p>0p>0 and α∈N0d\alpha\in\mathbb{N}_{0}^{d} there exists q>0q>0 such that, for every n∈Nn\in\mathbb{N},

sup⁡x∈Rd(1+∣x∣)p(1+∣φn(x)∣)q∣(ψn,φ)(α)(x)∣<∞(≤Mn for some M>0).\sup_{x\in\mathbb{R}^{d}}\frac{(1+|x|)^{p}}{(1+|\varphi_{n}(x)|)^{q}}\left|(\psi^{n,\varphi})^{(\alpha)}(x)\right|<\infty\quad(\leq M^{n}\text{ for some }M>0).

(b) For all α∈N0d\alpha\in\mathbb{N}_{0}^{d} there exists q>0q>0 such that, for every n∈Nn\in\mathbb{N},

sup⁡x∈Rd∣φn(α)(x)∣(1+∣φn(x)∣)q<∞(≤Mn for some M>0).\sup_{x\in\mathbb{R}^{d}}\frac{|\varphi_{n}^{(\alpha)}(x)|}{(1+|\varphi_{n}(x)|)^{q}}<\infty\quad(\leq M^{n}\text{ for some }M>0).

Proof. Suppose that (ii)(a) and (ii)(b) hold. By Corollary 2.4 of [1], Cψ,φC_{\psi,\varphi} acts on S(Rd)\mathcal{S}(\mathbb{R}^{d}). Moreover, equations (2.1) and (2.2) of [1] imply that, for every p>0p>0 and α∈N0d\alpha\in\mathbb{N}_{0}^{d}, there exists q>0q>0 such that

∀n∈N ∃Cn>0 ∀f∈S(Rd):∥Cψ,φnf∥p,α≤Cnmax⁡∣λ∣≤∣α∣∥f∥q,λ,\forall n\in\mathbb{N}\ \exists C_{n}>0\ \forall f\in\mathcal{S}(\mathbb{R}^{d}):\quad\|C_{\psi,\varphi}^{n}f\|_{p,\alpha}\leq C_{n}\max_{|\lambda|\leq|\alpha|}\|f\|_{q,\lambda},

where the constants CnC_{n} can be chosen of the form MnM^{n} for some M>0M>0 in the m-topologizable case. Hence, Cψ,φC_{\psi,\varphi} is topologizable (m-topologizable).

Conversely, suppose that (i) holds. By Proposition 4.10 of [1], for every p>0p>0 and α∈N0d\alpha\in\mathbb{N}_{0}^{d} there exists q>0q>0 such that

sup⁡x∈Rd(1+∣x∣)p(1+∣φn(x)∣)q∣Fα,0φn,ψn,φ(x)∣<∞(≤Mn for some M>0)\sup_{x\in\mathbb{R}^{d}}\frac{(1+|x|)^{p}}{(1+|\varphi_{n}(x)|)^{q}}\left|F_{\alpha,0}^{\varphi^{n},\psi^{n,\varphi}}(x)\right|<\infty\quad(\leq M^{n}\text{ for some }M>0)

for every n∈Nn\in\mathbb{N}. Since Fα,0φn,ψn,φ=(ψn,φ)(α)F_{\alpha,0}^{\varphi^{n},\psi^{n,\varphi}}=(\psi^{n,\varphi})^{(\alpha)} by equation (2.1) of [1], condition (ii)(a) follows. □

  1. Proposition 5.4. Since condition (ii)(b) of Theorem 4.5 is not necessary for the power boundedness of Cψ,φC_{\psi,\varphi} on S(Rd)\mathcal{S}(\mathbb{R}^{d}), the proof of Proposition 5.4 in [1] is invalid for ∣a∣>1|a|>1. We therefore restrict Proposition 5.4 to the following statement, whose proof is already contained in [1].

Let φ:R→R\varphi:\mathbb{R}\to\mathbb{R}, x↦ax+bx\mapsto ax+b, with b∈Rb\in\mathbb{R} and 0<∣a∣<10<|a|<1, and let ψ\psi be a nonzero polynomial. Then Cψ,φC_{\psi,\varphi} is not power bounded.

We note that the conclusion also holds for ∣a∣>1|a| > 1 whenever ψ(b1−a)≠0\psi\left(\frac{b}{1-a}\right) \ne0, as observed in [2].

  1. Proposition 5.5. Proposition 5.5 is false as stated and must be withdrawn. We refer to [2] for related results.

  1. Theorem 5.8. In condition (ii), remove the alternative concerning smooth weights ψ\psi of small decay with respect to φ\varphi. Thus, condition (ii) must be replaced by the requirement that Cψ,φC_{\psi,\varphi} is power bounded on S(Rd)\mathcal{S}(\mathbb{R}^{d}) for every ψ∈OM(Rd)\psi\in\mathcal{O}_{M}(\mathbb{R}^{d}).

  1. Theorem 5.10. In conditions (f), (g), and (h), remove the alternative concerning smooth weights ψ\psi of small decay with respect to φ\varphi. Thus, these conditions must be replaced by the following:

(f) Cψ,φC_{\psi,\varphi} is power bounded on S(R)\mathcal{S}(\mathbb{R}) for every ψ∈OM(R)\psi\in\mathcal{O}_{M}(\mathbb{R}).

(g) (Cψ,φn)n∈N(C_{\psi,\varphi}^{n})_{n\in\mathbb{N}} converges to 00 in Lb(S(R))\mathcal{L}_{b}(\mathcal{S}(\mathbb{R})) for every ψ∈OM(R)\psi\in\mathcal{O}_{M}(\mathbb{R}).

(h) Cψ,φC_{\psi,\varphi} is (uniformly) mean ergodic on S(R)\mathcal{S}(\mathbb{R}) for every ψ∈OM(R)\psi\in\mathcal{O}_{M}(\mathbb{R}).

Moreover, the original proof of the implication (e)⇒\Rightarrow(b) relies on Proposition 5.5 and is therefore invalid. To complete the proof of the corrected theorem, we establish the implication (e)⇒\Rightarrow(a) directly, using Cesàro means, and recall that the equivalence of conditions (a)–(d) has been established in the literature; see [1] for the references. We write

Anf=1n∑m=1nCφmf.A_n f = \frac{1}{n}\sum_{m=1}^{n} C_{\varphi}^{m}f.

Assume (e), i.e. {An:n∈N}\{A_n:n\in\mathbb{N}\} is bounded in Lb(S(R))\mathcal{L}_{b}(\mathcal{S}(\mathbb{R})), and suppose, to derive a contradiction, that (a) does not hold, i.e. that φ\varphi has a fixed point. Let x0x_0 be its largest fixed point. Since φ(x)−x\varphi(x)-x has no zeros on (x0,∞)(x_0,\infty), it has constant sign there.

Case 1: φ(x)>x\varphi(x)>x for every x>x0x>x_0. For every such xx, the sequence φn(x)\varphi^n(x) is strictly increasing. It cannot be bounded, since its limit would be a fixed point greater than x0x_0. Hence φn(x)→+∞\varphi_n(x)\to+\infty. For every f∈S(R)f\in\mathcal{S}(\mathbb{R}) it follows that

Anf(x)→0(x>x0),Anf(x0)=f(x0).(5)A_n f(x)\to0\quad(x>x_0),\qquad A_n f(x_0)=f(x_0). \tag*{(5)}

By (e), the sequence (Anf)n∈N(A_n f)_{n\in\mathbb{N}} is bounded in S(R)\mathcal{S}(\mathbb{R}). Since S(R)\mathcal{S}(\mathbb{R}) is a Montel space, (Anf)n∈N(A_n f)_{n\in\mathbb{N}} has a convergent subsequence. Its limit is a continuous function, contradicting the pointwise limits in (5) for a choice of ff with f(x0)≠0f(x_0)\ne0.

Case 2: φ(x)<x\varphi(x)<x for every x>x0x>x_0. Then φ\varphi has negative leading coefficient. Set θ=φ2\theta=\varphi^2. We distinguish the parity of deg⁡φ\deg\varphi.

If deg⁡φ\deg\varphi is odd, then θ\theta has odd degree and positive leading coefficient. Let x1x_1 be its largest fixed point, which exists since x0x_0 is a fixed point of θ\theta. We have θ(x)>x\theta(x)>x for x>x1x>x_1. Thus θn(x)→+∞\theta_n(x)\to+\infty for all x>x1x>x_1, by the monotonicity argument in Case 1. Since φ(t)→−∞\varphi(t)\to-\infty as t→+∞t\to+\infty, the odd iterates φ2n+1(x)\varphi_{2n+1}(x) also tend to infinity in modulus. Consequently Anf(x)→0A_n f(x)\to0 for x>x1x>x_1 and every f∈S(R)f\in\mathcal{S}(\mathbb{R}).

If deg⁡φ\deg\varphi is even, then θ\theta has even degree and negative leading coefficient. Let x1x_1 now be the smallest fixed point of θ\theta. Since θ(x)−x→−∞\theta(x)-x\to-\infty as x→−∞x\to-\infty and there are no fixed points of θ\theta below x1x_1, we have θ(x)<x\theta(x)<x for x<x1x<x_1. The sequence θn(x)\theta_n(x) is strictly decreasing and must tend to −∞-\infty for every x<x1x<x_1; otherwise its finite limit would be a fixed point below x1x_1. Again the odd iterates φ2n+1(x)=φ(θn(x))\varphi_{2n+1}(x)=\varphi(\theta_n(x)) tend to infinity in modulus. Hence Anf(x)→0A_nf(x)\to0 for x<x1x<x_1.

In either subcase, θ(x1)=x1\theta(x_1)=x_1, so the orbit of x1x_1 under φ\varphi has period at most two and

Anf(x1)→f(x1)+f(φ(x1))2.A_nf(x_1)\to\frac{f(x_1)+f(\varphi(x_1))}{2}.

Choose f∈S(R)f\in\mathcal{S}(\mathbb{R}) such that f(x1)=f(φ(x1))=1f(x_1)=f(\varphi(x_1))=1. The pointwise limit of (Anf)n∈N(A_nf)_{n\in\mathbb{N}} is therefore discontinuous at x1x_1, since it vanishes on an adjacent half-line. On the other hand, by (e), (Anf)n∈N(A_nf)_{n\in\mathbb{N}} is bounded in the Montel space S(R)\mathcal{S}(\mathbb{R}) and hence has a convergent subsequence whose limit is continuous. This contradiction proves (e)⇒\Rightarrow(a).

References

References

  1. [1]V. Asensio, Enrique E. Jordá, and T. Kalmes. Power boundedness and related properties for weighted composition operators on ℒ(ℝ^d). J. Funct. Anal., 288(3):Paper No. 110745, 33, 2025.DOI
  2. [2]S. Li. Personal communication, 2026. Email/Discussion on October 12, 2026.

Paper details

vlnc:W_vp3r237b
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents