betaFeedback

WorksMathematics

Similarity cancellation and primitive operator roots

Abstract

Kaplansky's second test problem for similarity asks whether A⊕A∼B⊕BA\oplus A\sim B\oplus B implies A∼BA\sim B for bounded operators on a separable infinite-dimensional Hilbert space. We answer this question in the negative, even for compact operators whose singular values satisfy any prescribed positive upper bounds tending to zero. We also show that cancellation can fail under the stronger hypothesis J2(T)∼J2(S)J_2(T)\sim J_2(S) considered by Marcoux, Radjavi, Troscheit, and Zhang \cite{MRTZ}. For these operators, finite amplifications and primitive roots are similar exactly at even multiplicities and even orders, respectively. In contrast, cancellation in fixed matrix size admits a uniform bound on the condition number. This extends the finite type~II results of Jiang, Ma, Shi, and Zhang \cite{JMSZ}. The bound also holds for simultaneous similarity of arbitrary families of matrices and can be chosen to tend to one as the amplified bound tends to one. We conclude that, in a finite type~II von Neumann algebra, cancellation of doubles and cancellation of primitive square roots are each equivalent to boundedness of the orders of its nonzero homogeneous summands.

Introduction

Let HH be a complex separable infinite-dimensional Hilbert space. We denote by B(H)\mathcal{B}(H) the algebra of bounded operators on HH. Two operators AA and BB are similar, written A∼BA \sim B, if YA=BYY A = B Y for some bounded invertible operator YY. We use the same notation for operators on different Hilbert spaces, with YY acting between those spaces. All Hilbert space direct sums are orthogonal. For a positive integer qq, write

A(q)=A⊕⋯⊕A⏟q copiesA^{(q)} = \underbrace{A \oplus\cdots\oplus A}_{q\ \text{copies}}

In his study of abelian groups, Kaplansky asked whether G⊕G≅F⊕FG \oplus G \cong F \oplus F forces G≅FG \cong F [16]. This is the second of his three test problems

2020 Mathematics Subject Classification. Primary 47A45, 47A65; Secondary 46L10, 47B10.

Key words and phrases. Similarity, finite amplification, primitive operator roots, compact operators, von Neumann algebras. for direct sums. Kadison and Singer [15] brought these questions to Hilbert space operators, and Azoff [1] studied them in several categories of operator algebras. Related versions of Kaplansky’s test problems for quasisimilarity were studied by Bercovici [2] and, for bitriangular operators, by Davidson and Herrero [5]. The answer depends on the equivalence relation being considered. For unitary equivalence, Kadison and Singer proved that

A⊕A≃uB⊕B⟹A≃uB,A \oplus A \simeq_{u} B \oplus B \Longrightarrow A \simeq_{u} B,

where ≃u\simeq_{u} denotes unitary equivalence. There is also a multiplicity theory for approximate unitary equivalence. Hadwin’s rank criterion for unital representations on a separable Hilbert space [8 Theorem 2.5] gives a formulation of Voiculescu’s theorem suited to this question; see also [4 §II.5]. In particular, finite amplifications cancel for approximate unitary equivalence in B(H)\mathcal{B}(H), as shown in [19 Proposition 4.4]. Related questions concerning distances between unitary orbits of matrix amplifications were studied by Marcoux, Sarkowicz, and Zhang [20]. Their results concern a different quantity from the condition number considered here. For similarity, the question is

A⊕A∼B⊕B⟹A∼B ?(1)A \oplus A \sim B \oplus B \Longrightarrow A \sim B\ ? \tag*{(1)}

Jiang [11 Theorem 4.6] proved cancellation of finite amplifications for strongly irreducible Cowen–Douglas operators. Jiang, Guo, and Ji [12] developed the similarity classification of Cowen–Douglas operators using the ordered KK-group of the commutant algebra. For weighted Hardy spaces of polynomial growth, Hou and Jiang [10 Theorem B] proved cancellation of doubles for the Hermitian holomorphic vector bundles induced by functions analytic in a neighbourhood of the closed unit disc.

Marcoux, Radjavi, Troscheit, and Zhang record this question in [19 §1.3], including the case in which AA and BB are compact.

The same authors consider a stronger hypothesis. Following their notation, for n≥2n \ge2 let

Jn(A)(h1,…,hn)=(h2,…,hn,Ah1).(2)J_{n}(A)(h_{1},\ldots,h_{n})=(h_{2},\ldots,h_{n},Ah_{1}). \tag*{(2)}

They call Jn(A)J_{n}(A) the primitive nnth root of A(n)A^{(n)}. In particular,

J2(A)=(0IA0),Jn(A)n=A(n).J_{2}(A)=\begin{pmatrix} 0 & I \\ A & 0 \end{pmatrix}, \qquad J_{n}(A)^{n}=A^{(n)}.

Thus, similarity of J2(A)J_{2}(A) and J2(B)J_{2}(B) implies similarity of the doubles. The question in [19 Remark 5.3] is whether

J2(A)∼J2(B)⟹A∼B.(3)J_{2}(A) \sim J_{2}(B) \Longrightarrow A \sim B. \tag*{(3)}

Their Theorem 4.10 proves this implication when one of the operators is compact. A counterexample to cancellation in a unital ring, due to J. Bell, is recorded in [19 Example 5.4]. That algebraic example does not settle either question in B(H)\mathcal{B}(H).

We answer both questions in the negative. Our counterexample to (3) is noncompact, as required by the theorem just cited. The same pair distinguishes even from odd multiplicities, both for direct sums and for primitive roots.

Theorem 1.1. For every ε>0\varepsilon> 0, there are noncompact operators T,S∈B(H)T, S \in\mathcal{B}(H) with ∥T∥\lVert T\rVert, ∥S∥<ε\lVert S\rVert< \varepsilon such that

T(q)∼S(q)⟺q is even(q≥1),(4)T^{(q)} \sim S^{(q)} \qquad\Longleftrightarrow\qquad q\ \text{is even} \qquad(q \ge1), \tag*{(4)}
Jn(T)∼Jn(S)⟺n is even(n≥2).(5)J_n(T) \sim J_n(S) \qquad\Longleftrightarrow\qquad n\ \text{is even} \qquad(n \ge2). \tag*{(5)}

The operators have the same point spectrum. Every eigenspace is one-dimensional, and the eigenvectors of each operator have dense linear span.

The answer to (1) is negative even when both operators are compact. In fact, their singular values can lie below any prescribed positive sequence tending to zero. For a compact operator RR, let sj(R)s_j(R) denote its singular values in nonincreasing order, counted with multiplicity.

Theorem 1.2. Let (aj)(a_j) be a nonincreasing sequence of positive numbers tending to zero. There are compact operators K,L∈B(H)K, L \in\mathcal{B}(H) such that

sj(K),sj(L)≤aj(j≥1)(6)s_j(K), s_j(L) \le a_j \qquad(j \ge1) \tag*{(6)}

and

K(q)∼L(q)⟺q is even(q≥1).(7)K^{(q)} \sim L^{(q)} \qquad\Longleftrightarrow\qquad q\ \text{is even} \qquad(q \ge1). \tag*{(7)}

Both operators are injective and have dense range. Their common nonzero spectrum consists of positive algebraically simple eigenvalues, and their eigenvectors have dense linear span.

For an arbitrary positive null sequence (aj)(a_j), apply Theorem 1.2 to the nonincreasing sequence bj=min⁡1≤i≤jaib_j = \min_{1 \le i \le j} a_i. Since 0<bj≤aj0 < b_j \le a_j and bj→0b_j \to0, the same upper bounds follow without a monotonicity assumption. In particular, KK and LL may belong to every Schatten class. Their positive eigenvalues should not be confused with positivity of the operators.

The counterexamples are orthogonal sums of diagonalizable matrices. In finite dimension, the Jordan form gives cancellation of finite amplifications; see [19 §1.3]. For an infinite orthogonal sum, however, the similarities chosen on the summands must have uniformly bounded norms and inverse norms. Pointwise similarity without such bounds is not enough, as the examples of Deckard and Pearcy already show [7 §2]. We therefore also consider the condition number of a similarity. For matrices A,B∈Mn(C)A,B \in M_n(\mathbb{C}), set

s(A,B)=inf⁡{∥Y∥∥Y−1∥:Y∈GLn(C),  YA=BY},(8)s(A,B)=\inf\{\lVert Y\rVert\lVert Y^{-1}\rVert:Y\in GL_n(\mathbb{C}),\;YA=BY\}, \tag*{(8)}

with s(A,B)=∞s(A,B)=\infty when AA and BB are not similar. For n,q≥1n,q\ge1 and t≥1t\ge1, put

θn,q(t)=sup⁡{s(A,B):A,B∈Mn(C),  s(A(q),B(q))≤t}.(9)\theta_{n,q}(t)=\sup\{s(A,B):A,B\in M_n(\mathbb{C}),\;s(A^{(q)},B^{(q)})\le t\}. \tag*{(9)}

For q=2q=2, this is the function introduced by Jiang, Ma, Shi, and Zhang [13 Definition 4.1].

Theorem 1.3. For every n,q≥1n,q\ge1 and t≥1t\ge1,

θn,q(t)<∞.\theta_{n,q}(t)<\infty.

Jiang, Ma, Shi, and Zhang relate this bound to similarity cancellation in homogeneous type InI_n von Neumann algebras [13 Theorem 4.3]. They prove cancellation under property (J)(J) [13 Theorem 3.27] and, without this assumption, for n≤3n\le3 [13 Proposition 4.7]. Theorem 1.3, together with the continuous selection theorem of Deckard and Pearcy [7], removes the extra assumption for every finite nn and every multiplicity qq.

Two further conclusions follow from the proof. The same uniform bound, with a possibly different constant, holds when one operator is required to intertwine an arbitrary family of matrices simultaneously. The constant does not depend on the number of matrices in the family. For a single pair, we also prove

lim⁡t↓1θn,q(t)=1(n,q fixed).(10)\lim_{t\downarrow1}\theta_{n,q}(t)=1\qquad(n,q\text{ fixed}). \tag*{(10)}

Thus, a similarity with condition number close to one remains so after cancellation, provided the matrix size is fixed. Proposition 2.2 proves the same assertion for a common intertwiner of arbitrary families, uniformly in the index set.

The dependence on matrix size cannot be omitted. The finite-dimensional operators used in Theorems 1.1 and 1.2 have uniformly controlled similarities after doubling, while every similarity at an odd multiplicity has arbitrarily large condition number. Consequently, there is a fixed t0t_0 such that

θn,2(t0)<∞for every n,sup⁡n≥1θn,2(t0)=∞.(11)\theta_{n,2}(t_0)<\infty\quad\text{for every }n,\qquad\sup_{n\ge1}\theta_{n,2}(t_0)=\infty. \tag*{(11)}

These pairs also show that s(A(2),B(2))<s(A,B)s(A^{(2)},B^{(2)})<s(A,B) can occur, contrary to the equality suggested in [13 Remark 4.2]; see Corollary 5.1. The relation between a dimension-independent bound and cancellation in finite type I algebras was established in [13 Corollary 4.6]. We decide cancellation for each bounded product ∏jMdj(C)\prod_j M_{d_j}(\mathbb{C}): both the double and primitive-square-root questions have affirmative answers for all pairs in the product if and only if sup⁡jdj<∞\sup_j d_j < \infty. More generally, for a finite type I von Neumann algebra, the same two properties are equivalent to a common finite bound on the orders of its nonzero homogeneous summands. The centres of those summands may be diffuse.

The proofs of the positive and negative results both concern uniform bounds for finite-dimensional intertwiners. To prove Theorem 1.3, we take a limit of the commutants of Ak⊕BkA_k \oplus B_k inside a fixed matrix space. The limit is a finite-dimensional algebra, so Krull–Schmidt cancellation applies. The resulting intertwiner can be approximated by exact intertwiners for the original pairs.

For the counterexamples, the starting point is the even and odd decomposition of a Sobolev space on the sphere. Restriction to a finite set gives two Hilbert norms on the same vector space of functions. An intertwiner between multiplication operators with distinct eigenvalues must act pointwise. Applying it to constant functions and lifting to the odd Sobolev space produces a continuous odd matrix-valued function. At an odd multiplicity its determinant must vanish, by Borsuk–Ulam. Compactness of the inclusion into C(S2)C(S^2) then forces the condition number to be large. At multiplicity two, a fixed invertible block operator induces similarities on all the quotients. Its entries are chosen to satisfy the additional identity for primitive square roots. Taking orthogonal sums with disjoint finite spectra gives the two pairs of operators.

The paper is organised as follows. § 2 proves the fixed-size bound, its simultaneous version, and (10), and then treats homogeneous type In_n algebras. § 3 constructs the Hilbert space quotients and their intertwiners. § 4 proves Theorems 1.1 and 1.2 by taking orthogonal sums. § 5 compares bounded and unbounded matrix sizes and characterises cancellation in finite type I von Neumann algebras.

Cancellation in fixed matrix size

We first prove the uniform estimate for matrices. The argument uses cancellation inside a finite-dimensional algebra and compactness of the set of orthogonal projections on a fixed matrix space.

By Krull–Schmidt cancellation, P⊕q≅Q⊕qP^{\oplus q} \cong Q^{\oplus q} implies P≅QP \cong Q for finite-dimensional modules over a finite-dimensional algebra; see [17 Theorems 4.2 and 5.5].

We apply this result to right ideals in finite-dimensional operator algebras, without assuming self-adjointness. Let CC be a finite-dimensional unital algebra and let e,f∈Ce, f \in C be idempotents. Suppose

U∈Mq(fCe),V∈Mq(eCf),VU=e(q),UV=f(q).(12)U \in M_q(fCe), \qquad V \in M_q(eCf), \qquad VU = e^{(q)}, \qquad UV = f^{(q)}. \tag*{(12)}

Left multiplication by UU identifies (eC)⊕q(eC)^{\oplus q} with (fC)⊕q(fC)^{\oplus q}. Cancellation gives eC≅fCeC \cong fC. An isomorphism between these right modules is left multiplication by an element u∈fCeu \in fCe, and its inverse is left multiplication by some v∈eCfv \in eCf. Thus,

vu=e,uv=f.(13)vu = e, \qquad uv = f. \tag*{(13)}

We are now ready to prove Theorem 1.3.

Proof of Theorem 1.3. Fix nn, qq and tt, and suppose that θn,q(t)=∞\theta_{n,q}(t) = \infty. For each kk, choose Ak,Bk∈Mn(C)A_k, B_k \in M_n(\mathbb{C}) with

s(Ak(q),Bk(q))≤t,s(Ak,Bk)>k.(14)s(A_k^{(q)}, B_k^{(q)}) \le t, \qquad s(A_k, B_k) > k. \tag*{(14)}

By the definition of the infimum, there is an invertible Xk∈Mnq(C)X_k \in M_{nq}(\mathbb{C}) such that

XkAk(q)=Bk(q)Xk,∥Xk∥∥Xk−1∥<t+1.(15)X_k A_k^{(q)} = B_k^{(q)} X_k, \qquad\lVert X_k\rVert\lVert X_k^{-1}\rVert< t+1. \tag*{(15)}

Multiplying XkX_k by a positive scalar, we may assume

∥Xk∥=∥Xk−1∥<t+1.(16)\lVert X_k\rVert= \lVert X_k^{-1}\rVert< \sqrt{t+1}. \tag*{(16)}

Consider the commutant

Ck={Ak⊕Bk}′⊂M2n(C)C_k = \{A_k \oplus B_k\}' \subset M_{2n}(\mathbb{C})

and the two fixed projections

e=(In000),f=(000In).e = \begin{pmatrix} I_n & 0 \\ 0 & 0 \end{pmatrix}, \qquad f = \begin{pmatrix} 0 & 0 \\ 0 & I_n \end{pmatrix}.

Both projections belong to every CkC_k. Equip M2n(C)M_{2n}(\mathbb{C}) with the Hilbert–Schmidt inner product, and let PkP_k be the orthogonal projection onto the linear subspace CkC_k. The maps PkP_k act on a fixed finite-dimensional Hilbert space and have norm one. After passing to a subsequence, we may suppose that

Pk⟶P,Xk⟶XP_k \longrightarrow P, \qquad X_k \longrightarrow X

in norm. The limit PP is an orthogonal projection. For every h∈Cnqh \in\mathbb{C}^{nq}, (2.5) gives ∥Xkh∥≥(t+1)−1/2∥h∥\lVert X_k h\rVert\ge(t+1)^{-1/2}\lVert h\rVert. The same lower bound holds for XX. Since XX is square and finite-dimensional, it is invertible, and

Xk−1⟶X−1.X_k^{-1} \longrightarrow X^{-1}.

Put C=ran⁡P\mathcal{C}=\operatorname{ran} P. We first check that C\mathcal{C} is an algebra. If R,S∈CR,S\in\mathcal{C}, then Rk=PkR→RR_k=P_kR\to R and Sk=PkS→SS_k=P_kS\to S. Since RkSk∈CkR_kS_k\in\mathcal{C}_k,

Pk(RkSk)=RkSk.P_k(R_kS_k)=R_kS_k.

Taking limits gives P(RS)=RSP(RS)=RS. Hence, RS∈CRS\in\mathcal{C}. Moreover, I,e,f∈CI,e,f\in\mathcal{C}, since each of these matrices is fixed by every PkP_k. Thus, C\mathcal{C} is a finite-dimensional unital algebra.

Write Xk=[Xij(k)]X_k=[X_{ij}^{(k)}] and Xk−1=[Zij(k)]X_k^{-1}=[Z_{ij}^{(k)}] in q×qq\times q blocks of size nn. Equation (15) and its inverse give

(00Xij(k)0)∈fCke,(0Zij(k)00)∈eCkf.\begin{pmatrix} 0 & 0\\ X_{ij}^{(k)} & 0 \end{pmatrix}\in f\mathcal{C}_ke,\qquad \begin{pmatrix} 0 & Z_{ij}^{(k)}\\ 0 & 0 \end{pmatrix}\in e\mathcal{C}_kf.

If Rk∈CkR_k\in\mathcal{C}_k and Rk→RR_k\to R, then PkRk=RkP_kR_k=R_k implies PR=RPR=R. Therefore the corresponding corner matrices formed from the entries of XX and X−1X^{-1} belong to fCef\mathcal{C}e and eCfe\mathcal{C}f. Denote the resulting q×qq\times q matrices over C\mathcal{C} by UU and VV. The equations X−1X=XX−1=InqX^{-1}X=XX^{-1}=I_{nq} give (12).

By (13), there are u∈fCeu\in f\mathcal{C}e and v∈eCfv\in e\mathcal{C}f with vu=evu=e and uv=fuv=f. Write

u=(00Y0),v=(0Z00).u=\begin{pmatrix} 0 & 0\\ Y & 0 \end{pmatrix},\qquad v=\begin{pmatrix} 0 & Z\\ 0 & 0 \end{pmatrix}.

Then, ZY=YZ=InZY=YZ=I_n, so YY is invertible.

For each kk, the matrix uk=fPk(u)eu_k=fP_k(u)e belongs to fCkef\mathcal{C}_ke and tends to uu. Write its lower-left block as YkY_k. We have

YkAk=BkYk,Yk⟶Y.Y_kA_k=B_kY_k,\qquad Y_k\longrightarrow Y.

For all sufficiently large kk, YkY_k is invertible and Yk−1→Y−1Y_k^{-1}\to Y^{-1}. Thus, s(Ak,Bk)≤∥Yk∥∥Yk−1∥s(A_k,B_k)\le\lVert Y_k\rVert\lVert Y_k^{-1}\rVert stays bounded, contradicting (14). □

The proof does not require the sequences (Ak)(A_k) and (Bk)(B_k) to converge, or even to be bounded. It is the dimension of the ambient commutant space that is fixed. Also, the limiting algebra C\mathcal{C} need not itself be a commutant.

Simultaneous similarity

The proof just given also applies to a joint commutant. This allows the same operator to intertwine any number of matrices, without a commutativity assumption on the family.

Theorem 2.1. Fix nn, q≥1q\ge1 and t≥1t\ge1. There is a constant C=C(n,q,t)C=C(n,q,t) with the following property. Let (Aα)α∈I(A_\alpha)_{\alpha\in I} and (Bα)α∈I(B_\alpha)_{\alpha\in I} be families in Mn(C)M_n(\mathbb{C}), indexed by the same nonempty set. Suppose that one invertible matrix XX satisfies

XAα(q)=Bα(q)X(α∈I),∥X∥∥X−1∥≤t.XA_\alpha^{(q)}=B_\alpha^{(q)}X\quad(\alpha\in I),\qquad\lVert X\rVert\lVert X^{-1}\rVert\le t.

Then, there is an invertible Y∈Mn(C)Y \in M_n(\mathbb{C}) such that

YAα=BαY(α∈I),∥Y∥∥Y−1∥≤C.Y A_\alpha= B_\alpha Y \qquad(\alpha\in I), \qquad\lVert Y\rVert\lVert Y^{-1}\rVert\le C.

The constant is independent of II and of the two families.

We apply the preceding commutant argument to all the equations at once.

Proof. Suppose that no such constant exists. For each kk, choose two families (Ak,α)α∈Ik(A_{k,\alpha})_{\alpha\in I_k} and (Bk,α)α∈Ik(B_{k,\alpha})_{\alpha\in I_k} with a common amplified intertwiner XkX_k of condition number at most tt, but no common intertwiner of condition number at most kk. After balancing XkX_k, we have ∥Xk∥=∥Xk−1∥≤t\lVert X_k\rVert= \lVert X_k^{-1}\rVert\le\sqrt{t}.

Apply the proof of Theorem 1.3 with the joint commutants

Ck={Ak,α⊕Bk,α:α∈Ik}′⊂M2n(C).\mathcal{C}_k=\{A_{k,\alpha}\oplus B_{k,\alpha}:\alpha\in I_k\}'\subset M_{2n}(\mathbb{C}).

Each block of XkX_k satisfies every intertwining equation at the kkth stage, and the blocks of Xk−1X_k^{-1} satisfy the reverse equations. Thus the same limiting algebra and corner argument give an element u∈fCeu\in f\mathcal{C}e with invertible lower-left block. The lower-left blocks YkY_k of fPk(u)efP_k(u)e converge to that block and satisfy

YkAk,α=Bk,αYk(α∈Ik).Y_kA_{k,\alpha}=B_{k,\alpha}Y_k \qquad(\alpha\in I_k).

For all sufficiently large kk, the matrices YkY_k are invertible and their condition numbers are bounded, contrary to the choice of the families.

The index set may be infinite. The proof takes place in M2n(C)M_{2n}(\mathbb{C}) and uses only the common intertwining equations, not a bound on the number of those equations.

Condition numbers close to one

For an invertible matrix, condition number one means that a scalar multiple is unitary. The next proposition shows that this remains a uniform limiting statement after cancellation in a fixed size.

Proposition 2.2. For every fixed nn, q≥1q\ge1,

θn,q(1)=1,lim⁡t↓1θn,q(t)=1.(17)\theta_{n,q}(1)=1,\qquad\lim_{t\downarrow1}\theta_{n,q}(t)=1. \tag*{(17)}

More generally, for every ε>0\varepsilon>0 there is δ>0\delta>0, depending only on nn, qq, ε\varepsilon, with the following property. Let (Aα)α∈I(A_\alpha)_{\alpha\in I} and (Bα)α∈I(B_\alpha)_{\alpha\in I} be families in Mn(C)M_n(\mathbb{C}) indexed by the same nonempty set. If an invertible X∈Mnq(C)X\in M_{nq}(\mathbb{C}) satisfies

XAα(q)=Bα(q)X(α∈I),∥X∥∥X−1∥≤1+δ,X A_\alpha^{(q)}=B_\alpha^{(q)}X \qquad(\alpha\in I), \qquad\lVert X\rVert\lVert X^{-1}\rVert\le1+\delta,

then there is an invertible Y∈Mn(C)Y \in M_n(\mathbb{C}) such that

YAα=BαY(α∈I),∥Y∥∥Y−1∥≤1+ε.Y A_\alpha= B_\alpha Y \quad(\alpha\in I), \qquad\|Y\|\|Y^{-1}\| \le1+\varepsilon.

In this case the limiting algebra contains the adjoints of the corner entries. This allows us to choose a partial isometry before returning to the original commutants.

Proof. We prove the assertion for families first. If it fails, there are ε>0\varepsilon> 0 and, for each kk, families (Ak,α)α∈Ik(A_{k,\alpha})_{\alpha\in I_k} and (Bk,α)α∈Ik(B_{k,\alpha})_{\alpha\in I_k} with a common amplified intertwiner XkX_k of condition number at most 1+k−11+k^{-1}, but no common intertwiner of condition number at most 1+ε1+\varepsilon. After balancing,

∥Xk∥=∥Xk−1∥≤1+k−1.\|X_k\|=\|X_k^{-1}\|\le\sqrt{1+k^{-1}}.

Use the joint commutants from the proof of Theorem 2.1 and the projections and subsequences from the proof of Theorem 1.3. We obtain Pk→PP_k\to P, Xk→XX_k\to X, and the unital algebra C=ran⁡P\mathcal{C}=\operatorname{ran}P. Here XX and X−1X^{-1} have norm at most one, so XX is unitary.

Let A=C∩C∗\mathcal{A}=\mathcal{C}\cap\mathcal{C}^*. This is a finite-dimensional unital C∗\mathcal{C}^*-algebra containing ee and ff. The lower-left corner matrices formed from the blocks of XX belong to C\mathcal{C}. Their adjoints also belong to C\mathcal{C}, since X−1=X∗X^{-1}=X^* and the upper-right corner matrices formed from the blocks of XkX_k have limits in C\mathcal{C}. Hence, the matrices U,VU,V of (2.1) belong to the corresponding corners over A\mathcal{A}, and V=U∗V=U^*.

Apply (2.2) in A\mathcal{A}. We obtain u∈fAeu\in f\mathcal{A}e and v∈eAfv\in e\mathcal{A}f with vu=evu=e and uv=fuv=f. Thus, uu is an invertible map from eC2ne\mathbb{C}^{2n} onto fC2nf\mathbb{C}^{2n}. By continuous functional calculus in the corner eAee\mathcal{A}e [4 §I.4], its polar part

w=u(u∗u)−1/2w=u(u^*u)^{-1/2}

belongs to fAef\mathcal{A}e and satisfies w∗w=ew^*w=e and ww∗=fww^*=f. The inverse square root here is taken in eAee\mathcal{A}e, whose identity is ee. Write

w=(00W0);w=\begin{pmatrix} 0 & 0 \\ W & 0 \end{pmatrix};

then WW is unitary.

The lower-left blocks WkW_k of fPk(w)efP_k(w)e satisfy

WkAk,α=Bk,αWk(α∈Ik)W_kA_{k,\alpha}=B_{k,\alpha}W_k \quad(\alpha\in I_k)

and converge to WW. For all sufficiently large kk, they are invertible and

∥Wk∥∥Wk−1∥⟶1,\|W_k\|\|W_k^{-1}\|\longrightarrow1,

contrary to the choice of the families.

Now fix ε>0\varepsilon>0 and choose δ>0\delta>0 as in the assertion for families. For a single pair with s(A(q),B(q))≤1+δ/2s(A^{(q)},B^{(q)})\le1+\delta/2, the definition of the infimum gives an amplified intertwiner with condition number less than 1+δ1+\delta. The assertion just proved yields s(A,B)≤1+ϵs(A,B) \le1+\epsilon. Hence θn,q(1+δ/2)≤1+ϵ\theta_{n,q}(1+\delta/2) \le1+\epsilon, which proves the limit in (17). Finally, 1≤θn,q(1)≤θn,q(t)1 \le\theta_{n,q}(1) \le\theta_{n,q}(t) for t>1t>1, so θn,q(1)=1\theta_{n,q}(1)=1.

The matrix size is fixed throughout (17). Section 5 shows that a bound valid in all sizes is a different question.

Homogeneous type I-n algebras

We next apply the fixed-size estimate to continuous matrix-valued functions. Deckard and Pearcy studied these algebras in [6, 7]; the uniform norm requirement in their similarity theorem is the point needed here.

A homogeneous type InI_n von Neumann algebra is isomorphic to Mn(A)M_n(\mathcal{A}) for an abelian von Neumann algebra A\mathcal{A}. We may write A=C(Ω)\mathcal{A}=C(\Omega), where Ω\Omega is compact and hyperstonean, hence Stonian; see [14 §6.6]. We use the following continuous selection result of Deckard and Pearcy [7 Theorem 1, p. 858]: if A,B∈Mn(C(Ω))A,B\in M_n(C(\Omega)) are pointwise similar on a dense subset of Ω\Omega, with intertwiners whose norms and inverse norms are all less than one fixed number MM, then there is an invertible Y∈Mn(C(Ω))Y\in M_n(C(\Omega)) with YA=BYYA=BY and

∥Y∥, ∥Y−1∥≤M.\lVert Y\rVert,\ \lVert Y^{-1}\rVert\le M.

Theorem 2.3. Let M\mathcal{M} be a homogeneous type InI_n von Neumann algebra, with n<∞n<\infty, and let A,B∈MA,B\in\mathcal{M}. If A(q)∼B(q)A^{(q)}\sim B^{(q)} in Mq(M)M_q(\mathcal{M}) for some positive integer qq, then A∼BA\sim B in M\mathcal{M}. In particular, Jq(A)∼Jq(B)J_q(A)\sim J_q(B) in Mq(M)M_q(\mathcal{M}) implies A∼BA\sim B for every q≥2q\ge2.

We can now apply the selection theorem to the pointwise similarities.

Proof. Write M=Mn(C(Ω))\mathcal{M}=M_n(C(\Omega)) as above. Suppose that XA(q)=B(q)XXA^{(q)}=B^{(q)}X for an invertible X∈Mq(M)X\in M_q(\mathcal{M}), and set t=∥X∥∥X−1∥t=\lVert X\rVert\lVert X^{-1}\rVert. Evaluation at ω∈Ω\omega\in\Omega gives

s(A(ω)(q),B(ω)(q))≤t.s\left(A(\omega)^{(q)},B(\omega)^{(q)}\right)\le t.

By Theorem 1.3,

s(A(ω),B(ω))≤θn,q(t)<∞(ω∈Ω).s(A(\omega),B(\omega))\le\theta_{n,q}(t)<\infty\qquad(\omega\in\Omega).

Choose M>θn,q(t)M>\sqrt{\theta_{n,q}(t)}. For each ω\omega, the definition of the infimum gives a pointwise intertwiner with condition number less than M2M^2. A scalar multiple makes both its norm and its inverse norm less than MM. The theorem of Deckard and Pearcy applies on all of Ω\Omega and gives a similarity in M\mathcal{M}. For primitive roots, take qqth powers and apply the assertion just proved. □\square

Hilbert quotients and finite-dimensional intertwiners

We now construct finite-dimensional pairs with uniformly bounded similarities of their doubles, but without such a bound before amplification. We work with two fixed Hilbert spaces and take their quotients by the functions that vanish on a finite set.

Even and odd quotient spaces

We choose the Hilbert spaces so that evaluation is continuous and smooth multiplication is bounded. Let H=H2(S2)\mathcal{H}=H^2(\mathbb{S}^2) be the complex Sobolev space on the unit sphere in R3\mathbb{R}^3, equipped with an orthogonally invariant Sobolev norm, normalised so that ∥1∥H=1\|1\|_{\mathcal{H}}=1. Multiplication by every smooth function is bounded on H\mathcal{H}, and the natural inclusion

H⟶C(S2)(18)\mathcal{H}\longrightarrow C(\mathbb{S}^2) \tag*{(18)}

is compact. These are consequences of the smooth-multiplier, Sobolev embedding, and Rellich theorems; see [23 Chapter 4, §§2–3]. In particular, compactness follows by factoring through Ht(S2)H^t(\mathbb{S}^2) for 1<t<21<t<2, using Propositions 3.3–3.4 of that chapter. We use the continuous representative of each element of H\mathcal{H}.

The operator

(Γf)(x)=f(−x)(\Gamma f)(x)=f(-x)

is a unitary involution. Its orthogonal eigenspaces

H+=ker⁡(Γ−I),H−=ker⁡(Γ+I)\mathcal{H}_{+}=\ker(\Gamma-I),\qquad\mathcal{H}_{-}=\ker(\Gamma+I)

consist of the even and odd functions. A smooth even multiplier preserves both spaces, while a smooth odd multiplier maps each into the other.

Let ex+e_x^+ and ex−e_x^- denote evaluation at xx on H+\mathcal{H}_{+} and H−\mathcal{H}_{-}. Orthogonal invariance gives constants a,b>0a,b>0 such that

∥ex+∥=a,∥ex−∥=b(x∈S2).(19)\|e_x^+\|=a,\qquad\|e_x^-\|=b\qquad(x\in\mathbb{S}^2). \tag*{(19)}

The first functional is nonzero because it evaluates the constant function. The second is nonzero because a smooth odd function can be chosen not to vanish at any prescribed point. By (3.1) and the Arzelà–Ascoli theorem [3 Theorem VI.3.8], the unit ball of H−\mathcal{H}_{-} is uniformly equicontinuous. Hence,

sup⁡∣x−y∣<η∥ex−−ey−∥⟶0(η↓0).(20)\sup_{|x-y|<\eta}\|e_x^- - e_y^-\|\longrightarrow0\qquad(\eta\downarrow0). \tag*{(20)}

The norm in this formula is the dual norm on H−\mathcal{H}_{-}.

Let EE be a finite nonempty subset of S2S^2 containing no antipodal pair. Let HEH_E and KEK_E be the restrictions to EE of H+\mathcal{H}_+ and H−\mathcal{H}_-, with the quotient norms

∥u∥HE=inf⁡{∥f∥H:f∈H+, f∣E=u},∥u∥KE=inf⁡{∥f∥H:f∈H−, f∣E=u}.(21)\begin{aligned} \lVert u\rVert_{H_E} &= \inf\{\lVert f\rVert_{\mathcal{H}} : f \in\mathcal{H}_+, \ f|_E = u\},\\ \lVert u\rVert_{K_E} &= \inf\{\lVert f\rVert_{\mathcal{H}} : f \in\mathcal{H}_-, \ f|_E = u\}. \tag*{(21)} \end{aligned}

Both spaces contain every function on EE. Indeed, prescribe the values u(p)u(p) and ±u(p)\pm u(p) at pp and −p-p, respectively. Smooth interpolation on E∪(−E)E \cup(-E), followed by the corresponding symmetrisation, gives an even or odd function with restriction uu; see the bump-function construction in [18 Chapter 2]. Thus, HEH_E and KEK_E are the same vector space of functions, equipped with different Hilbert norms.

The restriction maps have closed kernels. By the Hilbert space projection theorem [3 Theorems I.2.5–I.2.7], each quotient has a linear isometric lift obtained by choosing the representative of least norm. In particular, there is an isometry

LE:KE⟶H−,(LEu)∣E=u.L_E: K_E \longrightarrow\mathcal{H}_-, \qquad(L_Eu)|_E = u.

The constant function on EE has HEH_E-norm at most one.

For p∈Ep \in E, evaluation on either quotient has the same norm as on the corresponding full space, by the isometric identification of the dual of a quotient with the annihilator of its kernel [3 Theorem III.10.2]. On the qq-fold sums, write

epH:HE(q)⟶Cq,epK:KE(q)⟶Cqe_p^H : H_E^{(q)} \longrightarrow\mathbb{C}^q, \qquad e_p^K : K_E^{(q)} \longrightarrow\mathbb{C}^q

for the componentwise evaluations. By (19) and the Riesz representation theorem,

(epH(epH)∗)=a2Iq,epK(epK)∗=b2Iq.(22)(e_p^H(e_p^H)^*) = a^2 I_q, \qquad e_p^K(e_p^K)^* = b^2 I_q. \tag*{(22)}

For any scalar function λ:E→C\lambda: E \to\mathbb{C}, let AE(λ)A_E(\lambda) and BE(λ)B_E(\lambda) be multiplication by λ\lambda on HEH_E and KEK_E:

AE(λ)u=λu,BE(λ)u=λu.(23)A_E(\lambda)u = \lambda u, \qquad B_E(\lambda)u = \lambda u. \tag*{(23)}

If λ\lambda is injective, both operators are diagonalizable with simple eigenvalues λ(p)\lambda(p). The eigenspace at λ(p)\lambda(p) consists of the functions supported at pp. Pointwise bounds on λ\lambda alone need not bound these operator norms uniformly as EE varies.

Odd multiplicities

We show that the norms of a similarity and its inverse cannot both stay bounded as the finite sets and their antipodes fill the sphere.

We call E∪(−E)E \cup(-E) an η\eta-net if each point of the sphere is at distance less than η\eta from this set. We shall use the fact that every continuous odd function S2→CS^2 \to\mathbb{C} has a zero. This is the Borsuk–Ulam theorem applied to the real and imaginary parts; see [9 Theorem 1.10, p. 32].

Proposition 3.1. For every M≥1M \ge1, there is η>0\eta> 0 with the following property. Let EE contain no antipodal pair, suppose that E∪(−E)E \cup(-E) is an η\eta-net, and let λ:E→C\lambda: E \to\mathbb{C} be injective. For every odd positive integer qq, each invertible operator

Y:HE(q)⟶KE(q),YAE(λ)(q)=BE(λ)(q)Y,Y : H_E^{(q)} \longrightarrow K_E^{(q)}, \qquad Y A_E(\lambda)^{(q)} = B_E(\lambda)^{(q)}Y,

satisfies

∥Y∥∥Y−1∥>M.(24)\lVert Y\rVert\lVert Y^{-1}\rVert> M. \tag*{(24)}

The choice of η\eta is independent of qq, of EE, and of the values of λ\lambda.

The constant function and the evaluation operators give the required estimate.

Proof. Put γ=b/a\gamma= b/a. By (20), choose η>0\eta> 0 such that

∣x−y∣<η⟹∥ex−−ey−∥<γM.(25)\lvert x-y\rvert< \eta\quad\Longrightarrow\quad\lVert e_x^- - e_y^-\rVert< \frac{\gamma}{M}. \tag*{(25)}

Suppose that such a YY has ∥Y∥∥Y−1∥≤M\lVert Y\rVert\lVert Y^{-1}\rVert\le M. After a positive scalar rescaling, we may assume

∥Y∥=∥Y−1∥≤M.(26)\lVert Y\rVert= \lVert Y^{-1}\rVert\le\sqrt{M}. \tag*{(26)}

Since the eigenvalues in (23) are distinct, the intertwining equation makes YY act pointwise:

(Yu)(p)=Gpu(p),Gp∈GLq(C),p∈E.(27)(Yu)(p) = G_pu(p), \qquad G_p \in GL_q(\mathbb{C}), \quad p \in E. \tag*{(27)}

In particular, epHY−1=Gp−1epKe_p^H Y^{-1} = G_p^{-1}e_p^K. Equation (22) gives

b∥Gp−1∥=∥Gp−1epK∥≤a∥Y−1∥≤aM.b\lVert G_p^{-1}\rVert= \lVert G_p^{-1}e_p^K\rVert\le a\lVert Y^{-1}\rVert\le a\sqrt{M}.

Consequently,

∥Gpξ∥≥γM∥ξ∥(ξ∈Cq).(28)\lVert G_p\xi\rVert\ge\frac{\gamma}{\sqrt{M}}\lVert\xi\rVert\qquad(\xi\in\mathbb{C}^q). \tag*{(28)}

Let CE:Cq→HE(q)C_E : \mathbb{C}^q \to H_E^{(q)} send ξ\xi to the constant function with value ξ\xi. Since ∥CE∥≤1\lVert C_E\rVert\le1, the operator U=LE(q)YCEU = L_E^{(q)}YC_E maps Cq\mathbb{C}^q into H−\mathcal{H}_- and has norm at most M\sqrt{M}. Define the matrix-valued function Φ\Phi by

Φ(x)ξ=(Uξ)(x).(29)\Phi(x)\xi= (U\xi)(x). \tag*{(29)}

It is continuous and odd, and Φ(p)=Gp\Phi(p) = G_p for p∈Ep \in E. Under the Hilbert space identification H−(q)=H−⊗Cq\mathcal{H}_-^{(q)} = \mathcal{H}_- \otimes\mathbb{C}^q, componentwise evaluation is ex−⊗Iqe_x^- \otimes I_q. Hence

∥(ex−−ey−)⊗Iq∥=∥ex−−ey−∥.\left\lVert(e_x^- - e_y^-) \otimes I_q \right\rVert= \left\lVert e_x^- - e_y^- \right\rVert.

It follows that

∥Φ(x)−Φ(y)∥≤M∥ex−−ey−∥.(30)\|\Phi(x)-\Phi(y)\| \le\sqrt{M}\left\|e_x^- - e_y^-\right\|. \tag*{(30)}

Since qq is odd, det⁡Φ\det\Phi is a continuous odd complex-valued function. Borsuk–Ulam gives a point zz where Φ(z)\Phi(z) is singular. Choose a unit vector ξ∈ker⁡Φ(z)\xi\in\ker\Phi(z) and a point p∈E∪(−E)p \in E \cup(-E) with ∣p−z∣<η|p-z| < \eta. At this point, Φ(p)\Phi(p) is GxG_x or −Gx-G_x for some x∈Ex \in E. Equations (28), (25), and (30) give the contradiction

γM≤∥Φ(p)ξ∥=∥(Φ(p)−Φ(z))ξ∥<MγM.\frac{\gamma}{\sqrt{M}} \le\|\Phi(p)\xi\| = \|(\Phi(p)-\Phi(z))\xi\| < \sqrt{M}\frac{\gamma}{M}.

An invertible block intertwiner

The intertwining equation for primitive square roots requires equal diagonal blocks; see [19 Lemma 4.9]. We choose those blocks, and the off-diagonal entries, so that the inverse can be written explicitly.

Lemma 3.2. Let ZZ, WW, VV be commuting bounded operators on a Hilbert space such that

Z2+WV=I,Z^2+WV=I,

and suppose that R=I+WVR=I+WV is invertible. Set

D=−V2R−1,F=(ZWDWZ).D=-V^2R^{-1}, \qquad F=\begin{pmatrix} Z & W \\ DW & Z \end{pmatrix}.

Then,

F−1=(RZ−RWVWVRZ),FJ2(D)=J2(D)F.(31)F^{-1}=\begin{pmatrix} RZ & -RW \\ VWV & RZ \end{pmatrix}, \qquad FJ_2(D)=J_2(D)F. \tag*{(31)}

If a bounded involution Γ\Gamma anticommutes with ZZ, WW, VV, then DD commutes with Γ\Gamma, while FF and F−1F^{-1} interchange the two eigenspaces of Γ⊕Γ\Gamma\oplus\Gamma.

Proof. All the operators in the block entries commute. The relation Z2=I−WVZ^2=I-WV gives

R(Z2−DW2)=(I+WV)(I−WV)+(WV)2=I.R(Z^2-DW^2)=(I+WV)(I-WV)+(WV)^2=I.

Multiplying the two block matrices in either order gives the identity, with lower-left inverse entry −RDW=V2W=VWV-RDW=V^2W=VWV. Comparing the products with J2(D)J_2(D) proves the intertwining identity. If Γ\Gamma anticommutes with ZZ, WW, VV, then RR, R−1R^{-1}, DD commute with Γ\Gamma, and each entry of FF and F−1F^{-1} anticommutes with Γ\Gamma. This proves the last assertion. □

We realise these operators on H\mathcal{H}. For a smooth scalar function gg, let MgM_g denote multiplication by gg, and put

Z=Mx3,W=Mx1+ix2,V=Mx1−ix2.(32)Z=M_{x_3}, \qquad W=M_{x_1+ix_2}, \qquad V=M_{x_1-ix_2}. \tag*{(32)}

They are bounded, commute, and satisfy Z2+WV=IZ^2+WV=I by the equation of the sphere. All three anticommute with Γ\Gamma. Here VV is multiplication by the conjugate coordinate, not the Hilbert space adjoint of WW.

The operator RR is multiplication by 1+x12+x221+x_1^2+x_2^2. Its reciprocal is smooth, so R−1R^{-1} is a bounded multiplier. This argument does not require RR to be self-adjoint for the Sobolev inner product. The operator DD is multiplication by the smooth even function

d(x)=−(x1−ix2)21+x12+x22.(33)d(x)=-\frac{(x_1-\mathrm{i}x_2)^2}{1+x_1^2+x_2^2}. \tag*{(33)}

Lemma 3.2 therefore gives a bounded invertible operator

F:H+(2)⟶H−(2).(34)F:H_+^{(2)}\longrightarrow H_-^{(2)}. \tag*{(34)}

Both FF and its inverse have smooth multiplication operators as entries. They are fixed independently of the finite set EE.

We now pass this operator and its inverse to the finite-dimensional quotients.

Proposition 3.3. There is C≥1C\ge1 such that, for each finite nonempty EE containing no antipodal pair, an invertible operator

XE:HE(2)⟶KE(2)X_E:H_E^{(2)}\longrightarrow K_E^{(2)}

satisfies ∥XE∥,∥XE−1∥≤C\lVert X_E\rVert,\lVert X_E^{-1}\rVert\le C and

XEAE(λ)(2)=BE(λ)(2)XE(35)X_EA_E(\lambda)^{(2)}=B_E(\lambda)^{(2)}X_E \tag*{(35)}

for every scalar function λ\lambda on EE. Moreover,

XEJ2(AE(d∣E))=J2(BE(d∣E))XE,(36)X_EJ_2(A_E(d|_E))=J_2(B_E(d|_E))X_E, \tag*{(36)}

and

∥AE(d∣E)∥,∥BE(d∣E)∥≤∥D∥.(37)\lVert A_E(d|_E)\rVert,\lVert B_E(d|_E)\rVert\le\lVert D\rVert. \tag*{(37)}

Proof. Multiplication preserves vanishing on EE. Thus, FF maps the restriction kernel in H+(2)H_+^{(2)} into that in H−(2)H_-^{(2)}, and F−1F^{-1} maps the latter kernel back into the former. The two inclusions show that FF maps the first kernel onto the second. Hence, the induced quotient maps are inverses. By the quotient norms,

∥XE∥≤∥F∥,∥XE−1∥≤∥F−1∥.\lVert X_E\rVert\le\lVert F\rVert,\qquad\lVert X_E^{-1}\rVert\le\lVert F^{-1}\rVert.

For example, ∥XEu∥\lVert X_Eu\rVert is at most ∥Ff∥\lVert Ff\rVert for every lift ff of uu; taking the infimum over lifts gives the first bound. The inverse is treated in the same way.

At each p∈Ep\in E, the matrix defining XEX_E commutes with λ(p)I2\lambda(p)I_2. This proves (3.18), even when λ\lambda is specified only on EE. The identity with J2(D)J_2(D) in (31) passes to the quotients and gives (36). Finally, DD preserves both parity spaces and both restriction kernels. Its quotient operators have norm at most ∥D∥\lVert D\rVert, proving (37). □\square

The distinction between the two intertwining equations will be used below. Equation (35) allows any finite choice of eigenvalues. Equation (36) requires the particular multiplier DD.

Counterexamples on Hilbert space

We first choose the finite sets and recall how disjoint spectra control the intertwiners of an orthogonal sum. Both constructions will then use the operators from Section 3.

Finite sets and spectral separation

The finite sets must become dense after adjoining their antipodes. We also choose them so that no eigenvalue occurs in two different summands.

Let

U={x∈S2:0<x3<1}.U=\{x\in\mathbb{S}^{2}:0<x_{3}<1\}.

Choose finite nonempty sets Em⊂UE_m\subset U such that

the covering radii of Em∪(−Em) tend to zero,(38)\text{the covering radii of }E_m\cup(-E_m)\text{ tend to zero,} \tag*{(38)}

and all third coordinates of all selected points are distinct. Such sets can be chosen inductively. For a prescribed radius, compactness of the sphere and density of U∪(−U)U\cup(-U) give a finite antipodal net with smaller radius. Perturb its points within UU to avoid the third coordinates chosen earlier and to make the new coordinates distinct. Only finitely many values are excluded, and every nonempty open subset of UU contains an interval of possible third coordinates. The perturbations can be made small enough to retain the prescribed covering radius.

Suppose that AmA_m and BmB_m act on finite-dimensional Hilbert spaces, their norms are uniformly bounded, and

σ(Am)=σ(Bm)=Σm,Σm∩Σn=∅(m≠n).(39)\sigma(A_m)=\sigma(B_m)=\Sigma_m,\qquad\Sigma_m\cap\Sigma_n=\varnothing\quad(m\ne n). \tag*{(39)}

If Y(⨁mAm)=(⨁mBm)YY(\bigoplus_m A_m)=(\bigoplus_m B_m)Y, then its blocks satisfy

YnmAm=BnYnm.Y_{nm}A_m=B_nY_{nm}.

For n≠mn\ne m, the Sylvester–Rosenblum theorem [21] gives Ynm=0Y_{nm}=0. Thus, YY is block diagonal. If YY is invertible, its diagonal blocks are invertible and

∥Ym∥≤∥Y∥,∥Ym−1∥≤∥Y−1∥.(40)\lVert Y_m\rVert\le\lVert Y\rVert,\qquad\lVert Y_m^{-1}\rVert\le\lVert Y^{-1}\rVert. \tag*{(40)}

The same conclusion holds after a common finite amplification. This is the orthogonal-sum version of the spectral separation used in [19, Lemma 2.16]. Only disjointness of the finite spectra is needed; their union need not have isolated points.

Similar primitive square roots

We are now ready to prove Theorem 1.1. The multiplier DD gives uniformly bounded finite-dimensional operators, and the maps XEX_E give a similarity of their primitive square roots.

Proof of Theorem 1.1. On HEmH_{E_m} and KEmK_{E_m}, respectively, set

Am=AEm(d∣Em),Bm=BEm(d∣Em),A_m=A_{E_m}(d|_{E_m}), \qquad B_m=B_{E_m}(d|_{E_m}),

and define

T0=⨁mAm on ⨁mHEm,S0=⨁mBm on ⨁mKEm.(41)T_0=\bigoplus_m A_m\ \text{on}\ \bigoplus_m H_{E_m}, \qquad S_0=\bigoplus_m B_m\ \text{on}\ \bigoplus_m K_{E_m}. \tag*{(41)}

Both sums are bounded by (37). Proposition 3.3 also shows that

X=⨁mXEm,X−1=⨁mXEm−1X=\bigoplus_m X_{E_m}, \qquad X^{-1}=\bigoplus_m X_{E_m}^{-1}

are bounded. Reordering the orthogonal sums and applying (36) gives

XJ2(T0)=J2(S0)X.(42)XJ_2(T_0)=J_2(S_0)X. \tag*{(42)}

Taking squares gives similarity of the doubles. Repetition gives similarity at every even multiplicity.

For x∈Ux\in U, the function in (33) satisfies

∣d(x)∣=1−x322−x32.(43)|d(x)|=\frac{1-x_3^2}{2-x_3^2}. \tag*{(43)}

This is strictly decreasing as x3x_3 increases from zero to one. Hence, the values d(x)d(x) at all selected points are nonzero and pairwise distinct, including values from different sets. The finite blocks in (41) therefore satisfy (39).

Suppose that an invertible YY intertwines an odd amplification of T0T_0 and S0S_0. Spectral separation gives invertible restrictions YmY_m satisfying (40). Choose M≥1M\ge1 with M≥∥Y∥∥Y−1∥M\ge\lVert Y\rVert\lVert Y^{-1}\rVert. By (38), every sufficiently large EmE_m satisfies the net condition in Proposition 3.1. Hence,

∥Ym∥∥Ym−1∥>M,\lVert Y_m\rVert\lVert Y_m^{-1}\rVert>M,

contrary to (40). We have proved

T0(q)∼S0(q)⟺q is even.(44)T_0^{(q)}\sim S_0^{(q)}\quad\Longleftrightarrow\quad q\ \text{is even}. \tag*{(44)}

The point spectrum of either operator is exactly ⋃md(Em)\bigcup_m d(E_m). Indeed, the components of an eigenvector must solve the corresponding finite-block eigenvalue equations. Since all the eigenvalues are distinct, every eigenspace is one-dimensional. The eigenvectors span each finite block, so their linear span is dense in each direct sum.

Neither operator is compact. Fix a point on the equator and choose points in Em∪(−Em)E_m \cup(-E_m) converging to it. Their third coordinates tend to zero, so the corresponding eigenvalue moduli tend to 1/21/2 by (43). For either operator, choose unit eigenvectors in these distinct finite blocks. They form an orthonormal sequence whose images do not tend to zero. The compact-operator criterion in [3 Chapter II, §4] excludes compactness.

To obtain the assertion about all primitive roots, we use two identities from [19]: similarity of AA and BB implies similarity of Jk(A)J_k(A) and Jk(B)J_k(B) by Lemma 2.4(c), and

Jk(Jn(A))≃uJkn(A)J_k(J_n(A)) \simeq_u J_{kn}(A)

by Proposition 2.5(a). Applying these facts to (42) proves similarity of the primitive roots at every even order. At an odd order nn, such a similarity would give T0(n)∼S0(n)T_0^{(n)} \sim S_0^{(n)} after taking nnth powers, contrary to (44).

For t>0t>0, let Lt=diag⁡(t−1/2I,I)L_t=\operatorname{diag}(t^{-1/2}I,I). The scalar-rescaling identity recorded in [19 Remark 5.3] is

J2(tA)=t1/2LtJ2(A)Lt−1.(45)J_2(tA)=t^{1/2}L_tJ_2(A)L_t^{-1}. \tag*{(45)}

Choose t>0t>0 with tmax⁡{∥T0∥,∥S0∥}<εt\max\{\lVert T_0\rVert,\lVert S_0\rVert\}<\varepsilon, and put T=tT0T=tT_0, S=tS0S=tS_0. The intertwining equations for finite amplifications are unchanged by a common nonzero scalar. Equation (45) preserves the similarity of the primitive square roots. The preceding even-order argument and the odd-power argument then give (1.5) for this scaled pair. The assertions about point spectra, eigenspaces, and noncompactness follow from those proved above. Finally, identify the two separable infinite-dimensional Hilbert spaces unitarily with HH. □

Compact operators with prescribed singular value bounds

For the compact pair, we keep the quotient spaces and the similarities of the doubles, but choose positive eigenvalues separately on each finite set. Small scalar multiples will give the singular-value bounds in Theorem 1.2.

Proof of Theorem 1.2. Choose any injective function

λm:Em⟶(0,∞)\lambda_m:E_m\longrightarrow(0,\infty)

on each finite set, and now put

Am=AEm(λm),Bm=BEm(λm).A_m=A_{E_m}(\lambda_m),\qquad B_m=B_{E_m}(\lambda_m).

Their norms are finite, although they may grow with mm. Proposition 3.3 gives the same uniformly bounded similarities of their doubles, while Proposition 3.1 still applies at every odd multiplicity.

Write nm=∣Em∣n_m = |E_m|, N0=0N_0 = 0, and Nm=n1+⋯+nmN_m = n_1 + \cdots+ n_m. Choose positive numbers tmt_m inductively so that

tmmax⁡{∥Am∥,∥Bm∥}<aNm(46)t_m\max\{\|A_m\|,\|B_m\|\}<a_{N_m} \tag*{(46)}

and the finite sets tmλm(Em)t_m\lambda_m(E_m) are pairwise disjoint. At each step, (46) leaves a nonempty interval of choices, whereas spectral disjointness excludes only finitely many ratios of previous eigenvalues to the values of λm\lambda_m. Hence, both requirements can be satisfied.

Define

K=⨁mtmAm,L=⨁mtmBm.(47)K=\bigoplus_m t_m A_m,\qquad L=\bigoplus_m t_m B_m. \tag*{(47)}

By (46), the tails tend to zero in operator norm, so KK and LL are compact. For Nm−1<j≤NmN_{m-1}<j\leq N_m, the first m−1m-1 blocks have total rank less than jj, and the remaining tail has norm at most

sup⁡k≥maNk=aNm≤aj.\sup_{k\geq m}a_{N_k}=a_{N_m}\leq a_j.

The singular-value approximation principle [22 Chapter 1],

sj(R)=inf⁡{∥R−F∥:rank⁡F<j},s_j(R)=\inf\{\|R-F\|:\operatorname{rank}F<j\},

therefore gives (1.6) for both operators.

The maps XEmX_{E_m} and their inverses are uniformly bounded. Their direct sums are bounded inverses of one another, and (3.18), applied to tmλmt_m\lambda_m, gives K(2)∼L(2)K^{(2)}\sim L^{(2)}. Repetition gives all even multiplicities. If an invertible YY intertwined an odd amplification, the disjoint finite spectra would make it block diagonal. Its restrictions would satisfy (4.3), whereas Proposition 3.1 contradicts those bounds for sufficiently large mm. This proves (1.7).

Every finite block is invertible. Thus, K,LK,L are injective, and their ranges contain all finitely supported vectors, so both ranges are dense. Their eigenvectors span each finite block and have dense linear span. By the Riesz–Schauder theorem [3 Chapter VII, §7], every nonzero spectral value of a compact operator is an eigenvalue. The block equations identify these values as

⋃mtmλm(Em).\bigcup_m t_m\lambda_m(E_m).

They are positive and distinct. Each belongs to a single diagonalizable finite block. The same block equations for powers of K−μIK-\mu I and L−μIL-\mu I show that the generalized eigenspace at each nonzero eigenvalue μ\mu is one-dimensional. Hence, each such eigenvalue is algebraically simple. Unitary identifications put both operators on HH. □\square

For p>0p>0, a compact operator RR belongs to Sp\mathcal{S}_p when ∑jsj(R)p<∞\sum_j s_j(R)^p<\infty; see [22 Chapter 1]. Choosing a sequence whose positive powers are all summable gives the following consequence.

Corollary 4.1. There are injective compact operators K,LK,L with dense range, belonging to every Schatten class, such that K(2)∼L(2)K^{(2)}\sim L^{(2)} but K≁LK\not\sim L.

Proof. This follows from Theorem 1.2 by taking aj=2−ja_j=2^{-j}. □\square

For the compact pair, J2(K)≁J2(L)J_2(K)\not\sim J_2(L) by [19 Theorem 4.10]. The two constructions use the same quotient spaces and the same doubled intertwiners. Changing the eigenvalues and choosing small weights preserves (3.18), but not the additional identity (3.19) required for primitive square roots.

Matrix size and bounded products

We first compare the similarities of the finite-dimensional pairs before and after amplification. We then use those pairs to decide cancellation in bounded products of matrices and in finite type I von Neumann algebras.

The following corollary records the finite-dimensional pairs obtained from Propositions 3.1 and 3.3.

Corollary 5.1. There is C≥1C\ge1 with the following property. For every M≥1M\ge1, there are invertible diagonalizable matrices A,BA,B of the same size, with distinct eigenvalues and ∥A∥,∥B∥≤1\lVert A\rVert,\lVert B\rVert\le1, and an invertible XX such that

XJ2(A)=J2(B)X,∥X∥,∥X−1∥≤C,(48)XJ_2(A)=J_2(B)X,\qquad\lVert X\rVert,\lVert X^{-1}\rVert\le C, \tag*{(48)}

while every invertible YY satisfying YA(q)=B(q)YYA^{(q)}=B^{(q)}Y for an odd positive integer qq has

∥Y∥∥Y−1∥>M.(49)\lVert Y\rVert\lVert Y^{-1}\rVert>M. \tag*{(49)}

The matrix size may depend on MM.

Proof. Take the spaces HEm,KEmH_{E_m},K_{E_m} and multiplication by d∣Emd|_{E_m}. For sufficiently large mm, Proposition 3.1 gives (5.2) simultaneously for all odd qq. The eigenvalues are distinct and nonzero by (4.6) and the choice of the sets. Proposition 3.3 gives operator norms at most ∥D∥\lVert D\rVert and an intertwiner of the primitive square roots whose norm and inverse norm are uniformly bounded. Choose one fixed scalar c>0c>0 such that c∥D∥≤1c\lVert D\rVert\le1, and multiply both operators by cc. This leaves the intertwining equations for finite amplifications unchanged. By (45), the primitive-root intertwiner is conjugated by LcL_c. Its norm and inverse norm are multiplied by at most ∥Lc∥∥Lc−1∥\lVert L_c\rVert\lVert L_c^{-1}\rVert, a fixed number. Thus, (48) holds with a constant independent of mm and MM. Represent the two finite-dimensional Hilbert spaces in orthonormal bases to obtain the stated matrices. □\square

Taking squares in (48) gives s(A(2),B(2))≤C2s(A^{(2)},B^{(2)})\le C^2. Equation (49) with q=1q=1, together with Theorem 1.3, proves (1.11) with t0=C2t_0=C^2. Thus, the failure of a bound independent of matrix size occurs even for invertible diagonalizable contractions with simple spectra, and even when their primitive square roots have uniformly bounded similarities. Taking M>C2M>C^2 in Corollary 5.1 also gives s(A(2),B(2))<s(A,B)s(A^{(2)},B^{(2)})<s(A,B), so the equality suggested in [13 Remark 4.2] does not hold in general.

Let (dj)j≥1(d_j)_{j\ge1} be a sequence of positive integers. In the theorem below,

M=∏j≥1Mdj(C)(50)\mathcal{M}=\prod_{j\ge1}M_{d_j}(\mathbb{C}) \tag*{(50)}

denotes the von Neumann algebra of all sequences (Aj)(A_j) with sup⁡j∥Aj∥<∞\sup_j\lVert A_j\rVert<\infty, equipped with the supremum norm. This is the bounded product, not the norm-closed direct sum. An element (Yj)(Y_j) is invertible in M\mathcal{M} precisely when every YjY_j is invertible and both sup⁡j∥Yj∥\sup_j\lVert Y_j\rVert and sup⁡j∥Yj−1∥\sup_j\lVert Y_j^{-1}\rVert are finite.

The reduction for the full sequence of matrix sizes at multiplicity two is given in [13 Corollary 4.6]. We now decide cancellation for each sequence of sizes and include the primitive-square-root question.

Theorem 5.2. For the algebra in (50), the following are equivalent.

(i) sup⁡jdj<∞\sup_j d_j<\infty.

(ii) For all A,B∈MA,B\in\mathcal{M}, similarity of A(2)A^{(2)} and B(2)B^{(2)} in M2(M)M_2(\mathcal{M}) implies similarity of AA and BB in M\mathcal{M}.

(iii) For all A,B∈MA,B\in\mathcal{M}, similarity of J2(A)J_2(A) and J2(B)J_2(B) in M2(M)M_2(\mathcal{M}) implies similarity of AA and BB in M\mathcal{M}.

When these conditions hold, cancellation is valid for every finite amplification and every primitive root.

The fixed-size theorem proves one direction. For the converse, the matrices in Corollary 5.1 fit into a subsequence of factors of increasing size.

Proof. Assume (i), and choose NN with dj≤Nd_j \le N for every jj. Suppose that XA(q)=B(q)XXA^{(q)} = B^{(q)}X for an invertible X∈Mq(M)X \in M_q(\mathcal{M}), and put t=∥X∥∥X−1∥t = \lVert X\rVert\lVert X^{-1}\rVert. Evaluation in the jjth factor gives

s(Aj(q),Bj(q))≤t.s\left(A_j^{(q)},B_j^{(q)}\right) \le t.

By Theorem 1.3,

s(Aj,Bj)≤K≔max⁡1≤n≤Nθn,q(t)<∞.s(A_j,B_j) \le K \coloneqq\max_{1 \le n \le N} \theta_{n,q}(t) < \infty.

For each jj, choose YjAj=BjYjY_jA_j = B_jY_j with condition number less than K+1K+1, and balance YjY_j by a positive scalar. Then,

∥Yj∥, ∥Yj−1∥<K+1(j≥1).\lVert Y_j\rVert,\ \lVert Y_j^{-1}\rVert< \sqrt{K+1} \qquad(j \ge1).

Thus, Y=(Yj)Y=(Y_j) is invertible in M\mathcal{M} and satisfies YA=BYYA=BY. This proves (ii) and cancellation at every finite multiplicity. Taking powers gives cancellation for every primitive root. In particular, (ii) implies (iii) directly by squaring.

Suppose now that (dj)(d_j) is unbounded. For each m≥1m \ge1, choose matrices Am,Bm∈Mkm(C)A_m,B_m \in M_{k_m}(\mathbb{C}) and an intertwiner XmX_m as in Corollary 5.1, with M=mM=m. Choose distinct indices jmj_m inductively so that djm≥kmd_{j_m} \ge k_m. In the jmj_mth factor set

A^jm=Am⊕0djm−km,B^jm=Bm⊕0djm−km,\widehat{A}_{j_m}=A_m\oplus0_{d_{j_m}-k_m}, \qquad\widehat{B}_{j_m}=B_m\oplus0_{d_{j_m}-k_m},

and put A^j=B^j=0\widehat{A}_j=\widehat{B}_j=0 in the other factors. These sequences define contractions A^,B^∈M\widehat{A},\widehat{B}\in\mathcal{M}.

A permutation of the direct summands identifies J2(A^jm)J_2(\widehat{A}_{j_m}) with

J2(Am)⊕J2(0djm−km),J_2(A_m)\oplus J_2(0_{d_{j_m}-k_m}),

and gives the same identification for B^jm\widehat{B}_{j_m}. The map Xm⊕IX_m\oplus I intertwines these two operators. Its norm and inverse norm are at most CC, after increasing the fixed constant to at least one. In the other factors use the identity. The resulting sequence is an invertible element of M2(M)M_2(\mathcal{M}), so

J2(A^)∼J2(B^)in M2(M).(51)J_2(\widehat{A})\sim J_2(\widehat{B}) \quad\text{in } M_2(\mathcal{M}). \tag*{(51)}

If A^\widehat{A} and B^\widehat{B} were similar in M\mathcal{M}, their coordinate similarities would have uniformly bounded condition numbers. In the jmj_mth factor, invertibility of AmA_m and BmB_m separates their spectra from the additional zero block. The Sylvester–Rosenblum theorem therefore forces any such similarity to restrict to a similarity between AmA_m and BmB_m, with no increase in its norm or inverse norm. By (49), the condition number of this restriction is greater than mm. This contradicts the uniform bound. Thus, (5.4) holds without similarity of A^\widehat{A} and B^\widehat{B}, which disproves (iii). We have shown that (iii) implies (i), completing the proof. □\square

Finite type I von Neumann algebras

The centres of the homogeneous summands do not change the conclusion. We use the central decomposition of a finite type I von Neumann algebra [14 §6.6], in the form

M=∏n∈JMn(An),(52)\mathcal{M}=\prod_{n\in J}M_n(\mathcal{A}_n), \tag*{(52)}

where J⊆NJ\subseteq\mathbb{N} and each An\mathcal{A}_n is a nonzero abelian von Neumann algebra. The product in (52) is again the bounded product. We exclude the zero algebra, so JJ is nonempty.

Theorem 5.3. Let M\mathcal{M} be a nonzero finite type I von Neumann algebra, with decomposition (52). The following conditions are equivalent.

(i) The set JJ is bounded.

(ii) For all A,B∈MA,B\in\mathcal{M}, similarity of A(2)A^{(2)} and B(2)B^{(2)} in M2(M)M_2(\mathcal{M}) implies similarity of AA and BB in M\mathcal{M}.

(iii) For all A,B∈MA,B\in\mathcal{M}, similarity of J2(A)J_2(A) and J2(B)J_2(B) in M2(M)M_2(\mathcal{M}) implies similarity of AA and BB in M\mathcal{M}.

When these conditions hold, cancellation is valid for every finite amplification and every primitive root.

To prove the converse, it is important not to infer nonsimilarity in M\mathcal{M} merely from nonsimilarity in a subalgebra. The abelian centres provide homomorphisms back to the scalar matrix factors.

Proof. If JJ is bounded, it is finite. Apply Theorem 2.3 in each of the finitely many homogeneous summands. The resulting similarities and their inverses form bounded families, so they give a similarity in M\mathcal{M}. This proves cancellation for every finite amplification. Taking powers gives the assertion for primitive roots. As before, (ii) implies (iii) by squaring.

Suppose that JJ is unbounded, and put

N=∏n∈JMn(C).\mathcal{N}=\prod_{n\in J}M_n(\mathbb{C}).

By Theorem 5.2, there are A,B∈NA,B\in\mathcal{N} whose primitive square roots are similar in M2(N)M_2(\mathcal{N}) but which are not similar in N\mathcal{N}. Let

ι:N⟶M,ι((An))=(An⊗1An)\iota:\mathcal{N}\longrightarrow\mathcal{M},\qquad \iota((A_n))=(A_n\otimes1_{\mathcal{A}_n})

be the isometric unital inclusion. It carries the given similarity to one between J2(ι(A))J_2(\iota(A)) and J2(ι(B))J_2(\iota(B)). For each nn, choose a character χn\chi_n on An\mathcal{A}_n. Such a character exists because An\mathcal{A}_n is a nonzero unital abelian C∗C^*-algebra; see [4 §I.3]. The characters are not assumed to be normal. Applying χn\chi_n entrywise gives a unital ∗*-homomorphism

π:M⟶N,π((Tn))=((id⁡Mn⊗χn)(Tn))n∈J.\pi:\mathcal{M}\longrightarrow\mathcal{N},\qquad\pi((T_n))=\left((\operatorname{id}_{M_n}\otimes\chi_n)(T_n)\right)_{n\in J}.

Each coordinate map is contractive, so π\pi is well defined on the bounded product, and πι=id⁡N\pi\iota=\operatorname{id}_{\mathcal{N}}. If an invertible Y∈MY\in\mathcal{M} intertwined ι(A)\iota(A) and ι(B)\iota(B), then π(Y)\pi(Y) would intertwine AA and BB and would have inverse π(Y−1)\pi(Y^{-1}). This contradicts their nonsimilarity in N\mathcal{N}. Hence, (iii) fails when JJ is unbounded, proving the theorem. □

References

Department of Mathematics, Faculty of Sciences of Monastir, University of Monastir, Tunisia

Email address: [email protected]

Email address: [email protected]

References

  1. [1]E. A. Azoff, Test problems for operator algebras, Trans. Amer. Math. Soc. 347 (1995), no. 8, 2989–3001.DOI
  2. [2]H. Bercovici, Three test problems for quasisimilarity, Canad. J. Math. 39 (1987), no. 4, 880–892. doi:10.4153/CJM-1987-043-x.DOI
  3. [3]J. B. Conway, A Course in Functional Analysis, 2nd ed., Graduate Texts in Mathematics, vol. 96, Springer-Verlag, New York, 1990.
  4. [4]K. R. Davidson, C*-Algebras by Example, Fields Institute Monographs, vol. 6, American Mathematical Society, Providence, RI, 1996.DOI
  5. [5]K. R. Davidson and D. A. Herrero, The Jordan form of a bitriangular operator, J. Funct. Anal. 94 (1990), no. 1, 27–73. doi:10.1016/0022-1236(90)90027-I.DOI
  6. [6]D. Deckard and C. Pearcy, On matrices over the ring of continuous complex-valued functions on a Stonian space, Proc. Amer. Math. Soc. 14 (1963), no. 2, 322–328.DOI
  7. [7]D. Deckard and C. Pearcy, On continuous matrix-valued functions on a Stonian space, Pacific J. Math. 14 (1964), no. 3, 857–869. doi:10.2140/pjm.1964.14.857.DOI
  8. [8]D. W. Hadwin, Nonseparable approximate equivalence, Trans. Amer. Math. Soc. 266 (1981), no. 1, 203–231. doi:10.1090/S0002-9947-1981-0613792-6.
  9. [9]A. Hatcher, Algebraic Topology, Cambridge University Press, Cambridge, 2002.
  10. [10]B. Hou and C. Jiang, The Jordan decomposition and Kaplansky’s second test problem for Hermitian holomorphic vector bundles, preprint, 2025, arXiv:2503.03415v1.arxiv.org/abs/2503.03415
  11. [11]C. Jiang, Similarity classification of Cowen–Douglas operators, Canad. J. Math. 56 (2004), no. 4, 742–775. doi:10.4153/CJM-2004-034-8.DOI
  12. [12]C. Jiang, X. Guo, and K. Ji, K-group and similarity classification of operators, J. Funct. Anal. 225 (2005), no. 1, 167–192. doi:10.1016/j.jfa.2004.12.008.DOI
  13. [13]C. Jiang, M. Ma, R. Shi, and Y. Zhang, Kaplansky’s second test problem in operator algebras, preprint, 2026, arXiv:2607.20593v1.
  14. [14]R. V. Kadison and J. R. Ringrose, Fundamentals of the Theory of Operator Algebras. Vol. II: Advanced Theory, Graduate Studies in Mathematics, vol. 16, American Mathematical Society, Providence, RI, 1997.
  15. [15]R. V. Kadison and I. M. Singer, Three test problems in operator theory, Pacific J. Math. 7 (1957), no. 2, 1101–1106. doi:10.2140/pjm.1957.7.1101.DOI
  16. [16]I. Kaplansky, Infinite Abelian Groups, University of Michigan Press, Ann Arbor, MI, 1954.
  17. [17]H. Krause, Krull–Schmidt categories and projective covers, Expo. Math. 33 (2015), no. 4, 535–549. doi:10.1016/j.exmath.2015.10.001.DOI
  18. [18]J. M. Lee, Introduction to Smooth Manifolds, 2nd ed., Graduate Texts in Mathematics, vol. 218, Springer, New York, 2013.
  19. [19]L. W. Marcoux, H. Radjavi, S. Troitscheit, and Y. Zhang, Stability relations for Hilbert space operators and a problem of Kaplansky, Math. Z. 308 (2024), no. 3, Paper No. 39, 46 pp. doi:10.1007/s00209-024-03590-w; arXiv:2306.11202v1.DOI
  20. [20]L. W. Marcoux, P. Sarkowicz, and Y. Zhang, Kaplansky’s problem and unitary orbits in matrix amplifications, J. Operator Theory, to appear, arXiv:2508.13834.
  21. [21]M. Rosenblum, On the operator equation BX − XA = Q, Duke Math. J. 23 (1956), no. 2, 263–269. doi:10.1215/S0012-7094-56-02324-9.DOI
  22. [22]B. Simon, Trace Ideals and Their Applications, 2nd ed., Mathematical Surveys and Monographs, vol. 120, American Mathematical Society, Providence, RI, 2005.
  23. [23]M. E. Taylor, Partial Differential Equations I: Basic Theory, 2nd ed., Applied Mathematical Sciences, vol. 115, Springer, New York, 2011.

Paper details

vlnc:W_gdwtfr57
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