OctoInformBeta

OpenAI preprint, September 2026 · §5. A complete metric with nonnegative Ricci curvature

Proposition 5.1: Radial distance and positive subunit volume ratio persist across prescribed transmission histories in ℝ16\mathbb R^{16}

A common scale controls complete Berger metrics.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

In Cartesian dimension n=16n=16, with link dimension m=n−1=15m=n-1=15, the construction combines a radial scale with pulsed Hopf anisotropy. The geometric question is whether one can choose the scale and a starting threshold before choosing the subsequent controls, while retaining nonnegative Ricci curvature and completeness.

A smooth Riemannian metric gg on ℝn\mathbb R^n is a smooth positive definite symmetric matrix function G(x)G(x).✦ Write r(x)2=x⋅xr(x)^2=x\cdot x✦, r(x)=x⋅xr(x)=\sqrt{x\cdot x}✦, and t=log⁡rt=\log r away from the origin.✦ Riemannian path length and intrinsic distance are defined by (1), where the infimum is over continuously differentiable curves with the indicated endpoints: Lg(γ)=∫01γ̇(s)TG(γ(s))γ̇(s)ds,dg(x,y)=infγ(0)=xγ(1)=yLg(γ).L_g(\gamma)=\int_0^1 \sqrt{\dot\gamma(s)^T G(\gamma(s))\dot\gamma(s)}\,ds, \qquad d_g(x,y)=\inf_{\substack{\gamma(0)=x\\\gamma(1)=y}}L_g(\gamma).(1) A smooth Riemannian metric gg on ℝn\mathbb R^n is complete when every sequence that is Cauchy for dgd_g converges for dgd_g. The metric is Euclidean near the origin when G=IG=I on some ball r<εr<\varepsilon, with ε>0\varepsilon>0.✦

With GkhG^{kh} denoting an entry of G−1G^{-1}, eie_i the ith standard coordinate vector, and coordinate differentiation given by ∂if=Df[ei]\partial_i f=Df[e_i], the Christoffel symbols and Ricci tensor are the expressions in (2): Γijk=12∑hGkh(∂iGjh+∂jGih−∂hGij),(Ric⁡g)ij=∑k(∂kΓijk−∂jΓikk)+∑k,h(ΓijkΓkhh−ΓikhΓjhk).\begin{aligned} \Gamma^k_{ij} &=\tfrac12\sum_h G^{kh} (\partial_iG_{jh}+\partial_jG_{ih}-\partial_hG_{ij}),\\ (\operatorname{Ric}_g)_{ij} &=\sum_k(\partial_k\Gamma^k_{ij}-\partial_j\Gamma^k_{ik}) +\sum_{k,h}(\Gamma^k_{ij}\Gamma^h_{kh} -\Gamma^h_{ik}\Gamma^k_{jh}). \end{aligned}(2) A smooth Riemannian metric gg on ℝn\mathbb R^n has nonnegative Ricci curvature when Ric⁡g(x)(v,v)≥0\operatorname{Ric}_g(x)(v,v)\geq0 for every x,v∈ℝnx,v\in\mathbb R^n.

The intrinsic ball volume and asymptotic volume ratio use Cartesian Lebesgue measure and the Euclidean unit ball volume ωn\omega_n, as in (3): Vg(R)=∫{x:dg(0,x)<R}det⁡G(x)dx,ωn=vol⁡{x:r(x)<1},AVR⁡(g)=limR→∞Vg(R)ωnRn.\begin{aligned} V_g(R)&=\int_{\{x:d_g(0,x)<R\}}\sqrt{\det G(x)}\,dx, &\omega_n&=\operatorname{vol}\{x:r(x)<1\},\\ \operatorname{AVR}(g)&=\lim_{R\to\infty} \frac{V_g(R)}{\omega_nR^n}. \end{aligned}(3) The volumes and the limit are taken in the extended nonnegative reals.

The Scale and the Controls

Let η\eta denote the smooth transition function. For a real parameter μ\mu, the scale cutoff χ\chi, decay weight ww, integrated weight WW, total weight TT, radial scale aμa_\mu and its logarithmic slope hμh_\mu are defined by (4): χ(t)=η(t−5/2),w(t)=χ(t)(1+t)−5/4,W(t)=∫2tw(s)ds,T=W(4)+4⋅5−1/4,aμ(t)=e−μW(t),hμ(t)=−μw(t).\begin{aligned} \chi(t)&=\eta(t-5/2),& w(t)&=\chi(t)(1+t)^{-5/4},& W(t)&=\int_2^t w(s)\,ds,\\ T&=W(4)+4\cdot5^{-1/4},& a_\mu(t)&=e^{-\mu W(t)},& h_\mu(t)&=-\mu w(t). \end{aligned}(4) The fixed anisotropy q*q_*, limiting scale a*a_* and inverse squared scale β\beta are the constants in (5); they are distinct from the varying functions aμ(t)a_\mu(t) and q(t)q(t): q*=101100,a*=99853100000q*1/15,β=a*−2.q_*=\frac{101}{100},\qquad a_*=\sqrt{\frac{99853}{100000}\,q_*^{1/15}},\qquad \beta=a_*^{-2}.(5)

A smooth chronological control word consists of a nonempty ordered list of orthogonal complex structures JiJ_i, 0≤i<p0\leq i<p, and smooth amplitude functions Ai:ℝd→ℝA_i:\mathbb R^d\to\mathbb R, with a radius ρ>0\rho>0.✦ An orthogonal complex structure is a real matrix JJ satisfying J2=−IJ^2=-I and JTJ=IJ^TJ=I.✦ The standard one on ℝ16\mathbb R^{16} is J0=(0−I8I80)J_0=\left(\begin{smallmatrix}0&-I_8\\I_8&0\end{smallmatrix}\right). The control parameter ball is K=B¯ℝd(0,ρ)K=\overline B_{\mathbb R^d}(0,\rho).✦ An admissible amplitude vector ss in ℝp\mathbb R^p is a vector belonging to 𝒜={0}∪{(Ai(x))i=0p−1:x∈K}\mathcal A=\{0\}\cup\{(A_i(x))_{i=0}^{p-1}:x\in K\}.✦

The period with index j≥0j\geq0 has pulse duration PjP_j, dwell duration DjD_j and start time TjT_j given by (6): Pj=(j+1)6,Dj=(j+1)7,Tj=10+∑i=0j−1(Di+4Pi).P_j=(j+1)^6,\qquad D_j=(j+1)^7,\qquad T_j=10+\sum_{i=0}^{j-1}(D_i+4P_i).(6) Thus this index jj denotes the paper’s period j+1j+1.

For an amplitude history sj∈ℝps_j\in\mathbb R^p, the smooth steps σi\sigma_i, bumps bib_i, unit packet scalar zjz_j and individual packet stretches QjiQ_{ji} are (7), with M=2p+1M=2p+1: σi(u)=η(u−(2i+1)),bi(u)=σi′(u),zj(u)=M∑i=0p−1sjibi(Mu),Qji(u)=1+Pj−1Msjibi(Mu).\begin{aligned} \sigma_i(u)&=\eta(u-(2i+1)),& b_i(u)&=\sigma_i'(u),\\ z_j(u)&=M\sum_{i=0}^{p-1}s_{ji}b_i(Mu),& Q_{ji}(u)&=1+P_j^{-1}M s_{ji}b_i(Mu). \end{aligned}(7) The scalar anisotropy beginning at j0j_0 is defined by (8), using a smooth dwell window Ωj\Omega_j and transition ramp ff: f(u)=η(2u−1/2),Ωj(t)=f(t−Tj−PjPj)−f(t−Tj−2Pj−DjPj),q(t)=1+∑j≥j0[Pj−1zj(t−TjPj)+(q*−1)Ωj(t)].\begin{aligned} f(u)&=\eta(2u-1/2),\\ \Omega_j(t)&=f\!\left(\frac{t-T_j-P_j}{P_j}\right) -f\!\left(\frac{t-T_j-2P_j-D_j}{P_j}\right),\\ q(t)&=1+\sum_{j\geq j_0} \left[P_j^{-1}z_j\!\left(\frac{t-T_j}{P_j}\right) +(q_*-1)\Omega_j(t)\right]. \end{aligned}(8) The diagonal anisotropy q0q_0 is the scalar profile given by (8) with every sj=0s_j=0.✦ A selector dd is a function with d(u)∈{0,…,p−1}d(u)\in\{0,\ldots,p-1\} that selects the active bump by satisfying bi(Mu)=0b_i(Mu)=0 for i≠d(u)i\ne d(u). At every uu, either dd is constant on a neighborhood or all unit packet scalars vanish on a neighborhood, for every amplitude vector. The global angular selector is the selected slot d((t−Tj)/Pj)d((t-T_j)/P_j) in an active pulse and is absent elsewhere.✦ The corresponding complex structure J(t)J(t) is Jd((t−Tj)/Pj)J_{d((t-T_j)/P_j)} in that pulse and J0J_0 elsewhere.✦ For amplitudes s∈ℝps\in\mathbb R^p, the control packet scalar at time uu is ∑i=0p−1sibi(u)\sum_{i=0}^{p-1}s_i b_i(u).✦ The period anisotropy increment for period j+1j+1 is Pj−1zj((t−Tj)/Pj)+(q*−1)Ωj(t)P_j^{-1}z_j((t-T_j)/P_j)+(q_*-1)\Omega_j(t).✦ The active pulse index at logarithmic radius tt is the index of the active pulse containing tt, when there is one, among the pulses beginning at the threshold j0j_0; it is absent outside those pulses.✦ The control packet duration for pp chronological slots is M=2p+1M=2p+1.✦

The Metric and the Radial Equations

The horizontal scale cc, anisotropy half slope vv and horizontal logarithmic slope uu are defined by (9): c(t)=aμ(t)2q(t)−1/m,v(t)=q′(t)2q(t),u(t)=hμ(t)−v(t)m.c(t)=a_\mu(t)^2q(t)^{-1/m},\qquad v(t)=\frac{q'(t)}{2q(t)},\qquad u(t)=h_\mu(t)-\frac{v(t)}m.(9) For x≠0x\ne0, the radial unit vector is er=x/re_r=x/r and the Hopf unit vector is eJ=J(log⁡r)x/re_J=J(\log r)x/r. The Cartesian Berger metric gg on ℝn\mathbb R^n is the metric with coefficient function GG in (10), where ⊗\otimes is the outer product: G(0)=I,G(x)=c(t)I+(1−c(t))er⊗er+c(t)(q(t)−1)eJ⊗eJ,t=log⁡r(x).G(0)=I,\qquad G(x)=c(t)I+(1-c(t))e_r\otimes e_r +c(t)(q(t)-1)e_J\otimes e_J,\quad t=\log r(x).(10)✦ A Cartesian Berger metric gg on ℝn\mathbb R^n has positive smooth profiles c,qc,q, both equal to 11 below some logarithmic radius, and an angular selector JJ satisfying the following local condition: at every tt, either JJ is constant on a neighborhood or q=1q=1 on a neighborhood.✦

The radial, horizontal and Hopf Ricci numerators Rr,Rh,RvR_r,R_h,R_v are the scalar expressions in (11); dots denote differentiation in tt: Rr=−[m(u+u̇+u2)+v+v̇+2uv+v2],Rh=n−2q−c[u̇+(n−2+mu+v)(1+u)],Rv=(n−2)q2−cq[u̇+v̇+(n−2+mu+v)(1+u+v)].\begin{aligned} R_r&=-\bigl[m(u+\dot u+u^2)+v+\dot v+2uv+v^2\bigr],\\ R_h&=n-2q-c\bigl[\dot u+(n-2+m u+v)(1+u)\bigr],\\ R_v&=(n-2)q^2-cq\bigl[\dot u+\dot v +(n-2+m u+v)(1+u+v)\bigr]. \end{aligned}(11)

For degree ℓ\ell, let VℓV_\ell be the real vector space of homogeneous polynomials in nn variables of degree ℓ\ell annihilated by the Euclidean polynomial Laplacian Δ=∑i∂i2\Delta=\sum_i\partial_i^2, with the Fischer inner product, and put Nℓ=dim⁡VℓN_\ell=\dim V_\ell and Bℓ=ℓ(ℓ+m−1)B_\ell=\ell(\ell+m-1). The bounded Hopf derivative on VℓV_\ell is DJP=∑i,hJihxh∂iPD_J P=\sum_{i,h}J_{ih}x_h\partial_iP.✦ The standard Hopf square is DJ02D_{J_0}^2; write bℓib_{\ell i} for the vectors of an orthonormal eigenbasis bℓb_\ell and dℓid_{\ell i} for their eigenvalues. Write LL for the number of spectral blocks; their degrees are 2,…,L+12,\ldots,L+1, so their block index is ℓ−2\ell-2. For a real matrix MM on ℝn\mathbb R^n, the infinitesimal coordinate motion on real polynomials is the derivative along the linear vector field MxMx, given by P↦∑i,hMihxh∂iPP\mapsto\sum_{i,h}M_{ih}x_h\partial_iP.✦

The positive angular operator AℓA_\ell, radial coefficient α\alpha, drift p0(t)p_0(t) and limiting indicial root θℓ\theta_\ell have the forms in (12): Aℓ(t)=c(t)−1[BℓI+(1−q(t)−1)DJ(t)2],α(t)=aμ(t)−2,p0(t)=14+15hμ(t),θℓ=−7+49+βBℓ.\begin{aligned} A_\ell(t)&=c(t)^{-1} \bigl[B_\ell I+(1-q(t)^{-1})D_{J(t)}^2\bigr],\\ \alpha(t)&=a_\mu(t)^{-2},& p_0(t)&=14+15h_\mu(t),& \theta_\ell&=-7+\sqrt{49+\beta B_\ell}. \end{aligned}(12) Here the indicial equation is θℓ(θℓ+14)=βBℓ\theta_\ell(\theta_\ell+14)=\beta B_\ell. The pulse system uses exponent 1/151/15, limiting drift 1414, lower drift bound 1010 and coefficient bound 22. The radial history has the standard Hopf squares, central degrees ℓ\ell and central drift 1414, with Bℓ=ℓ(ℓ+14)B_\ell=\ell(\ell+14). The background and controlled Hopf squares are symmetric with quadratic forms between −Bℓ∥v∥2-B_\ell\|v\|^2 and zero, the radial coefficient is nonnegative, the drift satisfies 10≤p0(t)≤1410\leq p_0(t)\leq14 at every logarithmic radius, and the endpoint anisotropy is at least one.

The angular packet in (13) defines the angular coefficient of a control history before the selector dd chooses its active bump: Fℓ(q,J)=q1/mBℓI+(q1/m−q1/m−1)DJ2,𝒫jℓ(u)=BℓI+∑i[Fℓ(Qji(u),Ji)−BℓI],𝒜ℓ(t)=α(t)[Fℓ(q0(t),J0)+∑j≥j0(𝒫jℓ(t−TjPj)−BℓI)].\begin{aligned} F_\ell(q,J)&=q^{1/m}B_\ell I +(q^{1/m}-q^{1/m-1})D_J^2,\\ \mathcal P_{j\ell}(u)&=B_\ell I+ \sum_i\bigl[F_\ell(Q_{ji}(u),J_i)-B_\ell I\bigr],\\ \mathcal A_\ell(t)&=\alpha(t)\left[ F_\ell(q_0(t),J_0)+\sum_{j\geq j_0} \left(\mathcal P_{j\ell}\!\left(\frac{t-T_j}{P_j}\right)-B_\ell I\right) \right]. \end{aligned}(13) The exact Berger angular coefficient Fℓ(q,J)F_\ell(q,J) is the operator on VℓV_\ell in (13), with background BℓIB_\ell I and angular direction DJ2D_J^2.✦ The period angular increment in period j+1j+1 is the operator 𝒫jℓ((t−Tj)/Pj)−BℓI\mathcal P_{j\ell}((t-T_j)/P_j)-B_\ell I on VℓV_\ell. The active angular increment is the period angular increment for j≥j0j\geq j_0 and the zero operator for j<j0j<j_0.✦

The regular displacement operator Xℓ(t)X_\ell(t) is the value component of the first-order phase equation in (14): ddt(YZ)=(0I𝒜ℓ(t)−p0(t)I)(YZ),(Y(0),Z(0))=(I,ℓI),Xℓ(t)=Y(t).\frac{d}{dt}\binom{Y}{Z} =\begin{pmatrix}0&I\\\mathcal A_\ell(t)&-p_0(t)I\end{pmatrix}\binom{Y}{Z}, \qquad (Y(0),Z(0))=(I,\ell I),\qquad X_\ell(t)=Y(t).(14)✦ The evolution is the Peano-Baker series, whose zeroth term is the initial vector and whose next term is the integral from 00 to tt of the coefficient applied to the preceding term. For the scalar equation y″+p0y′=λyy''+p_0y'=\lambda y, the displacement propagators U,VU,V are the solutions with initial data (1,0)(1,0) and (0,1)(0,1). For period jj and degree ℓ\ell, the scalar Riccati baseline bjℓb_{j\ell} is the logarithmic derivative y′/yy'/y of the scalar solution with initial data (1,θℓ)(1,\theta_\ell). In period jj, its coefficient is α(Tj+s)Bℓ\alpha(T_j+s)B_\ell and its drift is p0(Tj+s)p_0(T_j+s), evaluated at s=max⁡(0,t−Tj)s=\max(0,t-T_j); denote it by bjℓ(t)b_{j\ell}(t).✦ For an operator coefficient A(t)A(t) on a real normed space EE and a scalar drift p(t)p(t), the first-order block operator on E×EE\times E sends (y,z)(y,z) to (z,A(t)y−p(t)z)(z,A(t)y-p(t)z).✦

The diagonal leg coefficient is λℓi(t)=α(t)[q0(t)1/mBℓ+(q0(t)1/m−q0(t)1/m−1)dℓi]\lambda_{\ell i}(t)=\alpha(t)[q_0(t)^{1/m}B_\ell+ (q_0(t)^{1/m}-q_0(t)^{1/m-1})d_{\ell i}]. Let Ujℓ,VjℓU_{j\ell},V_{j\ell} be the diagonal matrices of its scalar propagators from Tj+PjT_j+P_j to Tj+1T_{j+1}, with drift p0p_0. The whole-period value transmission is 𝒯jℓ=[Xℓ(Tj+1)Xℓ(Tj)−1]bℓ\mathcal T_{j\ell}=[X_\ell(T_{j+1})X_\ell(T_j)^{-1}]_{b_\ell}, where brackets mean matrix representation in the orthonormal basis bℓb_\ell.

Fix target matrices Sℓ∈SL⁡(Nℓ,ℝ)S_\ell\in\operatorname{SL}(N_\ell,\mathbb R). An admitted transmission history FF is a chronological control word, a starting index j0j_0, admissible amplitudes sjs_j vanishing for j<j0j<j_0, and a selector as above, with Qji(u)∈[1/2,3/2]Q_{ji}(u)\in[1/2,3/2] for j≥j0j\geq j_0, q(t)∈[1/2,2]q(t)\in[1/2,2], and Rr,Rh,Rv≥0R_r,R_h,R_v\geq0 everywhere, such that each active period satisfies (15) for some zjℓ∈(1/2,3/2)z_{j\ell}\in(1/2,3/2): 𝒯jℓ=exp⁡(∫TjTj+Pjbjℓ(t)dt)zjℓ(Ujℓ+θℓVjℓ)Sℓ,j≥j0,2≤ℓ≤L+1.\mathcal T_{j\ell} =\exp\!\left(\int_{T_j}^{T_j+P_j}b_{j\ell}(t)\,dt\right) z_{j\ell}\,(U_{j\ell}+\theta_\ell V_{j\ell})S_\ell, \qquad j\geq j_0,\quad 2\leq\ell\leq L+1.(15)✦ Its metric is denoted gFg_F.

The Claim

Theorem 1. Let TT be the total scale weight. For any real a0a_0 with 0<a0<10<a_0<1 and e−T/4≤a0e^{-T/4}\leq a_0, there is a scale parameter satisfying 0<μ≤14,limt→∞aμ(t)=a0,e−μT=a0.0<\mu\leq\tfrac14,\qquad \lim_{t\to\infty}a_\mu(t)=a_0,\qquad e^{-\mu T}=a_0.(16)✦ For every real μ\mu and every t≥4t\geq4, the logarithmic scale tail is log⁡(aμ(t)e−μT)=4μ(1+t)−1/4.\log\!\left(\frac{a_\mu(t)}{e^{-\mu T}}\right) =4\mu(1+t)^{-1/4}.(17)✦ Under the same hypotheses, the scale slope is hμ(t)=−μ(1+t)−5/4.h_\mu(t)=-\mu(1+t)^{-5/4}.(18)✦ Choose 0<μ≤1/40<\mu\leq1/4 with aμ(t)→a*a_\mu(t)\to a_*. Fix a chronological control word. There is an integer jgeomj_{\mathrm{geom}} such that, whenever j0≥jgeomj_0\geq j_{\mathrm{geom}} and the amplitude history is admissible, the following bounds hold at every logarithmic radius: 12≤q(t)≤q*+12,Rr(t)≥0,Rh(t)≥0,Rv(t)≥0.\tfrac12\leq q(t)\leq q_*+\tfrac12,\qquad R_r(t)\geq0,\qquad R_h(t)\geq0,\qquad R_v(t)\geq0.(19)✦ More generally, let μ≥0\mu\geq0 satisfy aμ(t)→a*a_\mu(t)\to a_*, let qq be smooth with values in [1/2,2][1/2,2], and let JJ be locally constant wherever qq is not identically 11 on a neighborhood. Suppose c=q=1c=q=1 for t≤0t\leq0. For the Cartesian Berger metric gg associated with these profiles, g is complete and Euclidean near 0,0<AVR⁡(g)<1,dg(0,x)=r(x)(x∈ℝ16).g\text{ is complete and Euclidean near }0,\qquad 0<\operatorname{AVR}(g)<1,\qquad d_g(0,x)=r(x)\quad(x\in\mathbb R^{16}).(20)✦ For the fixed scale parameter supplied by the scale choice with a0=a*a_0=a_*, every admitted transmission history FF satisfies gF is complete and Euclidean near 0,0<AVR⁡(gF)<1,dgF(0,x)=r(x)(x∈ℝ16).g_F\text{ is complete and Euclidean near }0,\qquad 0<\operatorname{AVR}(g_F)<1,\qquad d_{g_F}(0,x)=r(x)\quad(x\in\mathbb R^{16}).(21)✦ Its Ricci tensor satisfies Ric⁡gF(x)(v,v)≥0(x,v∈ℝ16).\operatorname{Ric}_{g_F}(x)(v,v)\geq0 \qquad(x,v\in\mathbb R^{16}).(22)✦

In the paper, the result reads: for the fixed spectral data and pulse family there are a smooth a:ℝ→(0,1]a:\mathbb R\to(0,1] and a threshold jgeomj_{\mathrm{geom}}, independent of subsequent controls, such that every j0≥jgeomj_0\geq j_{\mathrm{geom}} and every parameter sequence (hj)(h_j) in KK define a smooth complete metric g=dr2+r2𝒢(a(log⁡r),q(log⁡r),J(log⁡r))g=dr^2+r^2\mathcal G(a(\log r),q(\log r),J(\log r)) on ℝm+1\mathbb R^{m+1} with nonnegative Ricci curvature, where a=1a=1 for t≤2t\leq2 and decreases to a*a_*, for some μ>0\mu>0 and all t≥4t\geq4 one has a′/a=−μ(1+t)−5/4a'/a=-\mu(1+t)^{-5/4} and log⁡(a/a*)=4μ(1+t)−1/4\log(a/a_*)=4\mu(1+t)^{-1/4}, the metric is Euclidean near 00, dg(0,x)=r(x)d_g(0,x)=r(x), and dvol⁡g=rma(log⁡r)mdrdvol⁡g0d\operatorname{vol}_g=r^m a(\log r)^m\,dr\,d\operatorname{vol}_{g_0}, with 𝒢\mathcal G the Berger link metric and g0g_0 the round link metric.

What Is Checked

The scale selection (16) and tail identities (17), (18) are stated over the explicit weight and primitive in (4). The curvature threshold (19) is stated over an arbitrary chronological control word and its admissible amplitude vectors. The general geometry assertion (20) concerns Cartesian Berger coefficients, while (21) and (22) concern histories satisfying the packet, selection and transmission conditions.

The Cartesian Berger tensor is positive definite for positive c,qc,q.✦ Its entries are smooth when c,qc,q are smooth and round near the origin and JJ satisfies the local round condition.✦ For an admitted history, c,qc,q are smooth and positive and equal 11 for t≤0t\leq0; at each tt, JJ is locally constant or q=1q=1 on a neighborhood.✦ Smoothness concerns these metric coefficients and does not require a continuous label JJ.

For every real scale parameter μ\mu, the horizontal scale functional is positive whenever the anisotropy profile is positive at every logarithmic radius.✦ If that profile is also smooth, the horizontal scale functional is smooth.✦

The radial coefficient α\alpha is smooth✦, satisfies |α(t)|≤2|\alpha(t)|\leq2✦, equals 11 for t≤0t\leq0✦ and tends to β\beta✦; the drift equals 1414 for t≤0t\leq0✦.

For a symmetric operator on a finite-dimensional real inner-product space, its matrix in an orthonormal eigenbasis is the diagonal matrix of its eigenvalues.✦