--- title: 'Fuglede-type Estimates: A Quantitative Overview' url: https://www.emergentmind.com/topics/fuglede-type-estimates type: topic --- # Fuglede-type Estimates: A Quantitative Overview Searching arXiv for the cited papers and closely related work on Fuglede-type estimates. “Fuglede-type estimates” designates a family of quantitative statements rather than a single theorem. In current usage, the phrase covers at least four distinct lineages: approximation and lower-bound results for the Fuglede–Kadison determinant in finite von Neumann algebra settings; exact determinant computations connected with Lehmer’s problem and \(L^2\)-torsion; norm inequalities of Fuglede–Putnam type for nonnormal operators, often after Aluthge transform; and geometric stability inequalities modeled on Fuglede’s second-order analysis of perturbations of the ball. A separate but related lineage attached to Fuglede’s name concerns the spectral-set/tiling conjecture in finite vector spaces, where the relevant estimates are Fourier-analytic and combinatorial rather than determinant-theoretic [1001.0419], [2202.03877], [2406.15922], [1112.1302], [2304.12157], [1902.02936]. ## 1. Terminological scope and basic invariants The determinant-theoretic core is the Fuglede–Kadison determinant. For a countable discrete group \(\Gamma\), with group von Neumann algebra \(\mathcal N\Gamma\subset B(\ell^2(\Gamma))\) and tracial state \[ \operatorname{tr}_{\mathcal N\Gamma}(T)=\langle T\delta_e,\delta_e\rangle, \] an invertible \(f\in \mathcal N\Gamma\) has \[ \det_{\mathcal N\Gamma}(f)=\exp\!\bigl(\operatorname{tr}_{\mathcal N\Gamma}(\log |f|)\bigr) =\exp\!\Bigl(\tfrac12\,\operatorname{tr}_{\mathcal N\Gamma}\bigl(\log(f^*f)\bigr)\Bigr). \] In a general tracial \(W^*\)-probability space \((M,\tau)\), the corresponding notation in free-probabilistic contexts is \[ \Delta(T)=\exp\bigl(\tau(\log|T|)\bigr). \] A second invariant appearing in recent work is the capacity of a positive linear map \(\Phi:M_m(\mathbb C)\to M_m(\mathbb C)\), \[ \mathrm{cap}(\Phi)=\inf\Bigl\{\det(\Phi(b))\ \Bigm|\ b>0,\ \det(b)=1\Bigr\}, \] which is positive exactly when \(\Phi\) is rank non-decreasing [1001.0419], [2406.15922]. In operator theory, the phrase “Fuglede-type” refers instead to extensions or quantitative refinements of the classical Fuglede–Putnam implication. If \(A=U|A|\) and \(B=V|B|\), then \[ \Com(A,B)=\{X:AX=XB\}, \] and \((A,B)\) has the FP-property when \[ \Com(A,B)\subseteq \Com(A^*,B^*). \] The relevant transformed object is the Aluthge transform \[ \widetilde C=|C|^{1/2}U|C|^{1/2}. \] This is analytically unrelated to the determinant line, although both are described as “Fuglede-type” in the literature [1112.1302]. The geometric literature uses the phrase in yet another sense. For perturbations \(B_h\) of the unit-volume ball, one studies lower bounds for energies such as \(P+\varepsilon\,\mathrm{Cap}^{-1}\) or \(P-c\,\lambda_1\), with the inequality structure modeled on second-order isoperimetric expansions. By contrast, the finite-field Fuglede conjecture concerns whether a set is spectral if and only if it tiles by translation; there the basic estimates are Fourier zeros, equidistribution on parallel hyperplanes, and incidence bounds on affine lines and planes [2304.12157], [1902.02936]. | Lineage | Central object | Representative estimate | |---|---|---| | Fuglede–Kadison approximation | finite compressions \(f_{F_n}\) | normalized \(\log|\det(S_n)|\) converges to \(\log\det_{\mathcal N\Gamma}(f)\) | | Free-group determinant theory | \(\Id+\sum \zeta_iR_{x_i}\) | exact FK determinants and upper bounds for Lehmer’s constants | | Free probability and capacity | \(S=\sum a_i\otimes s_i\) | \(\Delta(S)=\mathrm{cap}(\Phi)^{1/(2m)}e^{-1/2}\) | | Fuglede–Putnam theory | \((A,B)\), \((\widetilde A,\widetilde B)\) | Schatten-norm lower bounds for \(\widetilde A^*X-X\widetilde B\) | | Convex-shape stability | \(P+\varepsilon \mathrm{Cap}^{-1}\), \(P-c\lambda_1\) | quadratic control by \(\|h\|_{H^1(\partial B)}^2\) | | Finite-field Fuglede conjecture | \(E\subset \mathbb F_p^3\) | plane/line concentration bounds exclude intermediate spectral sizes | ## 2. Approximation of finite compressions and amenable-group entropy For countable amenable \(\Gamma\), Li proved a finite-compression approximation theorem for \(\det_{\mathcal N\Gamma}(f)\) that is the prototypical determinant-theoretic Fuglede-type estimate [1001.0419]. If \(\{F_n\}\) is a Følner sequence, \(P_n:\ell^2(\Gamma)\to \ell^2(F_n)\) is the orthogonal projection, and \[ f_{F_n}=P_n f P_n\in B(\ell^2(F_n))\simeq M_{|F_n|}(\mathbb C), \] then \(f_{F_n}\) need not be invertible. The theorem therefore allows controlled perturbations. Precisely, if \(f\in \mathbb Z\Gamma\) is invertible in \(\mathcal N\Gamma\), then for every \(\varepsilon>0\) and every operator-norm bound \(K<\infty\), there exists \(\delta>0\) such that whenever \(n\) is large and \[ S_n\in B(\ell^2(F_n)),\qquad \|S_n\|\le K,\quad \|S_n^{-1}\|\le K,\quad \mathrm{rank}(S_n-f_{F_n})\le \delta |F_n|, \] one has \[ \Bigl|\frac1{|F_n|}\log|\det(S_n)|-\log\det_{\mathcal N\Gamma}(f)\Bigr|<\varepsilon. \] Thus the normalized finite-dimensional determinant is asymptotically rigid under small-rank perturbation, provided uniform norm and inverse-norm bounds are retained. The proof has four standard components in this circle of ideas. First, one reduces to the positive case and then returns to the general case by polar-decomposition and adjoint arguments. Second, Lück’s trace approximation theorem is applied to polynomials \(Q\), giving \[ \operatorname{tr}_{\mathcal N\Gamma}\!\bigl(Q(f^*f)\bigr) = \lim_{n\to\infty}\frac1{|F_n|}\,\mathrm{Tr}\bigl(Q(f_{F_n}^*f_{F_n})\bigr), \] after which polynomial approximation of \(\log\) transfers the limit to the determinant. Third, noninvertibility of \(f_{F_n}\) is “corralled” into a subspace of rank \(o(|F_n|)\), and a perturbation on that subspace produces an invertible \(S_n\). Fourth, a rank estimate combined with Weyl-type inequalities shows that the determinant error is negligible after division by \(|F_n|\). The theorem subsumes classical one- and multi-variable Mahler-measure regimes. For \(\Gamma=\mathbb Z\), a Laurent polynomial \(f=\sum_{k=-m}^m a_k u^k\in \mathbb Z[u,u^{-1}]\) with no zeros on the unit circle is invertible in \(L(\mathbb Z)\), and \(\det_{L(\mathbb Z)}(f)\) is the classical Mahler measure; the compressions are Toeplitz matrices and recover Szegő’s strong limit theorem. For \(\Gamma=\mathbb Z^d\), a torus-zero-free Laurent polynomial gives the higher-dimensional Mahler measure and multi-Toeplitz compressions. The same determinant approximation is the analytic side of Li’s entropy identity. For the algebraic action on \[ X_f=\{x\in (\mathbb R/\mathbb Z)^\Gamma:\ f*x=0\}, \] the main dynamical theorem is \[ h_{\mathrm{top}}(X_f)=\log\det_{\mathcal N\Gamma}(f). \] The normalized determinants \(\frac1{|F_n|}\log|\det(S_n)|\) simultaneously approximate the Fuglede–Kadison determinant and the exponential growth rate of \((F_n,\varepsilon)\)-separated orbit sets. Possible directions of further generalization explicitly indicated in this work are replacing amenable \(\Gamma\) by sofic groups, extending to other traces such as crossed products or groupoid von Neumann algebras, and allowing more general perturbations with subtler rank-versus-norm trade-offs. ## 3. Exact determinant formulas over free groups and Lehmer-type bounds A second determinant lineage replaces asymptotic approximation by exact evaluation. Ben Aribi studied Fuglede–Kadison determinants over non-cyclic free groups and connected them to Lück’s Lehmer constants [2202.03877]. For a bounded \(G\)-equivariant operator \(A\) on \(\ell^2(G)\), \[ \det_{N(G)}(A)=\lim_{\varepsilon\to 0^+}\exp\!\Bigl(\tfrac12\,\tr_G\!\bigl(\ln(A^*A+\varepsilon\,\Id)\bigr)\Bigr), \] and if \(A\) is positive and injective this simplifies to \(\exp(\tr_G(\ln A))\). The key exact computation concerns free generators \(x,y\) of \(F_2\). The paper proves \[ \det_{N(F_2)}(\Id+R_x+R_y)=\frac{2}{\sqrt3}\approx 1.1547. \] More generally, the analysis treats operators of the form \[ A=\Id+\sum_{i=1}^{d-1}\zeta_i R_{x_i},\qquad |\zeta_i|=1, \] by relating \(\det_{N(G)}(A)\) to trace generating series and to random walks on Cayley graphs. The decisive combinatorial input is closed-path counting on the \(d\)-regular tree, following techniques of Bartholdi and Dasbach–Lalin. The resulting integral evaluation uses explicit algebraic generating functions and elementary antiderivatives involving \(\mathrm{artanh}\) and logarithms. These exact formulas feed directly into Lehmer-type questions. The paper recalls the four constants \[ \Lambda(G)\le \Lambda^w(G)\le \Lambda^w_1(G),\ \Lambda_1(G), \] defined as infima of \(\det_{N(G)}(A)>1\) over different classes of integer-matrix operators. Since free groups satisfy the Strong Atiyah conjecture, the exact \(F_2\) computation yields \[ \Lambda^w_1(F_2)=\frac{2}{\sqrt3}<1.1763. \] By subgroup induction, any torsion-free group containing a copy of \(F_2\) satisfies \[ 1\le \Lambda(G),\Lambda^w(G),\Lambda_1(G),\Lambda^w_1(G)\le \frac{2}{\sqrt3}. \] Because \(2/\sqrt3<\mathcal M(L)=1.17628\dots\), the paper concludes that \(\mathcal M(L)\) cannot serve as a universal lower gap for such groups. The same paper also imports \(L^2\)-torsion. For a finite-volume hyperbolic \(3\)-manifold \(M\), \[ T^{(2)}(M)=\exp\!\bigl(\tfrac{\mathrm{vol}(M)}{6\pi}\bigr), \] and a Dehn-surgery gluing formula gives \[ T^{(2)}(N,Q)=T^{(2)}(M) \] for the induced epimorphism \(Q:\pi_1(N)\to \pi_1(M)\). In favorable CW-presentations, \(T^{(2)}(M)\) is itself a single Fuglede–Kadison determinant of a \(1\times1\) operator over \(\mathbb Z[\pi_1(M)]\). Hence the Lehmer constants of \(\pi_1(M)\) are bounded above by \(\exp(\mathrm{vol}(M)/6\pi)\), and for small-volume fillings of the Whitehead link this upper bound is strictly below \(2/\sqrt3\). ## 4. Matrix-valued semicircular elements, capacity, and universal lower bounds Recent free-probabilistic work gives a closed-form Fuglede–Kadison determinant formula in terms of capacity [2406.15922]. In a tracial \(W^*\)-probability space \((M,\tau)\), a matrix-valued semicircular element is \[ S=\sum_{i=1}^n a_i\otimes s_i\in M_m(\mathbb C)\,\overline\otimes\, M, \] where \(a_i\in M_m(\mathbb C)\) are fixed and \(s_i\) are freely independent standard semicircular elements. Its covariance map is \[ \Phi(b)=\sum_{i=1}^n a_i b a_i^*, \qquad \Phi^*(b)=\sum_{i=1}^n a_i^* b a_i. \] The central theorem is \[ \Delta(S)=\exp\bigl(\tau(\log|S|)\bigr)=\mathrm{cap}(\Phi)^{1/(2m)}\,e^{-1/2}, \] equivalently \[ \tau(\log|S|)=\frac1{2m}\log\mathrm{cap}(\Phi)-\frac12. \] The proof splits into an indecomposable case and a general rank-non-decreasing case. If \(\Phi\) is indecomposable, Gurvits’s operator-scaling theorem gives a unique minimizer \(b>0\), \(\det b=1\), for the capacity problem; scaling \(S\mapsto b^{1/2}Sb^{-1/2}\) makes the covariance doubly stochastic, and in that normalization the operator-valued free-probability analysis shows that \(S\) has the standard semicircle law on \([-2,2]\), hence determinant \(e^{-1/2}\). In the general case, one perturbs \[ \Phi_t=\Phi+t\,\mathrm{tr}_m(\cdot)\,1 \] by a depolarizing channel, proves the formula for \(\Phi_t\), and then lets \(t\to 0\) using semi-continuity of \(\Delta\) and an approximate minimizer for \(\mathrm{cap}(\Phi)\). The same paper sharpens a capacity lower bound for integer data. When \[ \Phi(b)=\sum_{i=1}^n a_i b a_i^*,\qquad a_i\in M_m(\mathbb Z), \] and \(\Phi\) is rank non-decreasing, Garg–Gurvits–Oliveira–Widgerson had shown \(\mathrm{cap}(\Phi)\ge m^{-2m}\). Mai and Speicher improve this to the dimension-independent estimate \[ \mathrm{cap}(\Phi)\ge 1. \] The argument embeds the associated linear pencil into \[ T=\sum_{i=1}^n a_i\otimes u_i\in M_m(\mathbb C)\,\overline\otimes\,L(\mathbb F_n), \] uses the Determinantal Conjecture as proved for free groups by Elek–Szabó to obtain \(\Delta(T)\ge 1\), and then combines a conjugation argument with the AM–GM bound \(\Delta(C)\le \tau(C)\). Several consequences are immediate. First, \[ \Delta(S)=0\iff \mathrm{cap}(\Phi)=0, \] so \(S\) is invertible as an unbounded operator if and only if \(\Phi\) is rank non-decreasing; this recovers analytically the noncommutative Edmonds problem. Second, for integer data and invertible \(S\), \[ \Delta(S)\ge e^{-1/2}, \] a universal lower bound independent of \(m\) and of the entries of the \(a_i\). Third, if \(S\ge 0\) has spectrum in \([0,L]\), then for every \(\varepsilon>0\), \[ \int_0^\varepsilon 1\,d\mu_S(t)\le \frac{\log L-\log \Delta(S)}{\log(L/\varepsilon)}, \] so a positive lower bound on \(\Delta(S)\) yields uniform control of spectral mass near zero. This suggests a precise infinite-dimensional analogue of the principle that an integer matrix with determinant at least \(1\) cannot have too many very small eigenvalues. ## 5. Fuglede–Putnam-type inequalities and the Aluthge transform In operator theory, “Fuglede-type” refers to consequences of the Fuglede–Putnam implication beyond the normal setting [1112.1302]. Let \[ A=U|A|,\qquad B=V|B| \] be polar decompositions on Hilbert spaces \(\mathscr H_1,\mathscr H_2\), and let \(\widetilde C=|C|^{1/2}U|C|^{1/2}\) be the Aluthge transform. The classical theorem asserts that if \(A\) and \(B\) are normal, then \[ AX=XB\Longrightarrow A^*X=XB^*. \] Moslehian and Nabavi Sales study how this property behaves under Aluthge transform. If \(A\) and \(B\) are invertible and \((A,B)\) has the FP-property, then \((\widetilde A,\widetilde B)\) also has the FP-property. Conversely, if \(A,B\) are invertible, the spectra of \(U\) and \(V\) are contained in some open semicircle, and \((\widetilde A,\widetilde B)\) has the FP-property, then \((A,B)\) has the FP-property. They also prove the commutant relations \[ \Com(A,B)\subseteq \Com(\widetilde A,\widetilde B), \] and, when \(A\) is invertible, \[ \Com(A,B)= \Com(\widetilde A,\widetilde B). \] The mechanism is functional calculus on \(|A|\) and \(|B|\), together with the intertwining relations produced by the FP-property. The quantitative estimate is a Schatten-norm inequality. If \[ \Re(U|A|^{1/2})\ge a>0,\qquad \Re(V|B|^{1/2})\ge a>0,\qquad U^*X=XV, \] then for any \(1\le p\le \infty\), \[ \|\widetilde A^*X-X\widetilde B\|_p \ge 2a\,\bigl\|\,|B|^{1/2}X-X|B|^{1/2}\bigr\|_p. \] The proof proceeds by first obtaining a single-operator estimate under the commutation condition \(U^*X=XU\), then embedding the pair into the block operators \[ T=\begin{pmatrix}A&0\\0&B\end{pmatrix}, \qquad Y=\begin{pmatrix}0&X\\X^*&0\end{pmatrix}, \] and finally reading off the relevant block inequality. This line should not be conflated with Fuglede–Kadison determinant estimates. The common feature is the extension of a Fuglede-origin theorem to nonnormal or perturbed settings, but the objects, techniques, and conclusions are different: one concerns trace-log invariants in von Neumann algebras, the other commutant structure and Schatten-norm control for bounded operators. ## 6. Geometric stability and finite-field analogues The geometric literature uses “Fuglede-type” for second-order stability inequalities around the ball under convexity constraints [2304.12157]. For the capacity functional, let \(B\subset \mathbb R^n\) be the unit-volume ball and write a small radial perturbation as \[ B_h=\{t x(1+h(x)):\ x\in \partial B,\ t\in [0,1]\}, \] with \(h\in W^{1,\infty}(\partial B)\), \(|B_h|=|B|\), and barycenter at \(0\). Then there exist \(\eta>0\) and \(\varepsilon_0>0\) such that for \(\|h\|_{W^{1,\infty}}\le \eta\) and \(0<\varepsilon<\varepsilon_0\), \[ P(B_h)+\varepsilon\,\mathrm{Cap}(B_h)^{-1}\ge P(B)+\varepsilon\,\mathrm{Cap}(B)^{-1}, \] with equality only when \(h\equiv 0\). Equivalently, with \(R(K)=\mathrm{Cap}(K)^{-1}\), \[ R(B_h)-R(B)=\mathrm{Cap}(B)^{-1}-\mathrm{Cap}(B_h)^{-1}\ge C\|h\|_{H^1(\partial B)}^2. \] The crucial weak IT estimate is \[ \mathrm{Cap}(B_h)-\mathrm{Cap}(B)\le C_n\|h\|_{H^1(\partial B)}^2, \] while perimeter contributes the classical second-order expansion \[ P(B_h)=P(B)+P'_B(0)[\xi_h]+\tfrac12 P''_B(0)[\xi_h,\xi_h]+o(\|h\|_{H^1}^2), \] with \[ P'_B(0)[\xi]=\int_{\partial B}H(\xi\cdot n),\qquad P''_B(0)[\xi,\xi]=\int_{\partial B} |\nabla_\tau(\xi\cdot n)|^2-(n-1)(\xi\cdot n)^2. \] The proof strategy combines a radial homeomorphism \(\phi_h\), the capacitary test function \(u_B(x)=\min\{1,|x|^{2-n}\}\), and sharp isoperimetric control of perimeter. For the first Dirichlet eigenvalue, the same paper studies \[ J_c(K)=P(K)-c\,\lambda_1(K) \] among convex unit-volume sets. It identifies the exact threshold \[ c^*=\frac{N(N+1)p_N}{4\,l_N(l_N-N)\,\omega_N^{(N+1)/N}} \] for local minimality of the ball. If \(00\) such that every convex \(K\) with \(|K|=1\) and \(|K\Delta B|\le \delta(c)\) satisfies \[ J_c(K)\ge J_c(B), \] with equality only for \(K=B\). If \(c>c^*\), there are smooth convex perturbations \(K_j\to B\) with \(J_c(K_j)