Conjecture Atlas

Proof source

A quantitative strain gap and the failure of fixed-control approximation on the supplied fan

Complete proof text in the unsigned reading edition.

Download sourceBrowse research records

Reading edition of the complete source note. Original labels and explicit equation tags are retained; references use source titles where compiled numbering is unavailable. Current result status is recorded separately in the project register.

Contents of this source
  1. A quantitative strain gap and the failure of fixed-control approximation on the supplied fan

A quantitative strain gap and the failure of fixed-control approximation on the supplied fan

Keep the five planes, trace data, fields bib_i, strains ZiZ_i, and edge stress ss of PK. Equip N=R4N=\mathbf R^4 and its subspaces with their Euclidean norms. Write BRi=Pi∩BR(0),DRij=Sij∩BR(0),s(c)=∑i<jsijcij.(KG1)B_R^i=P_i\cap B_R(0),\qquad D_R^{ij}=S_{ij}\cap B_R(0),\qquad \mathfrak s(c)=\sum_{i<j}s_{ij}c_{ij}. \tag{KG1} All tensor-measure total variations below use the Euclidean Frobenius norm. Changing to the fixed lattice calibrations changes only the constants.

The two external analytic inputs are the classical Korn–Poincare estimate in BD\mathrm{BD} and the bounded BD\mathrm{BD} trace map on a bounded Lipschitz domain. Here Eu=sym⁡DuEu=\operatorname{sym}Du denotes the full distributional symmetric derivative. In the form needed on a three-ball, there is an affine Killing field KK such that ∥u−K∥L1(BR)≤CR ∣Eu∣(BR).(KG2)\|u-K\|_{L^1(B_R)}\leq C R\,|Eu|(B_R). \tag{KG2} The trace estimate on a half-ball is ∥γu∥L1(∂BR+)≤C(R−1∥u∥L1(BR+)+∣Eu∣(BR+)).(KG3)\|\gamma u\|_{L^1(\partial B_R^+)} \leq C\bigl(R^{-1}\|u\|_{L^1(B_R^+)} +|Eu|(B_R^+)\bigr). \tag{KG3} We use KG2 only for continuous BD\mathrm{BD} fields on a closed ball; hence they are also in L2L^2. KG2 follows by Holder and scaling from the L3/2L^{3/2} estimate stated as Theorem A.1 in Friedrich’s paper cited below. KG3 follows by scaling from Babadjian’s Theorem 3.2; for a continuous field its trace is its usual boundary restriction. No L1L^1 estimate for the full gradient is used.

The finite Killing stress is stable in the strain-measure norm

Theorem (Quantitative obstruction). There is a constant C>0C>0, depending only on the fixed five-plane configuration, such that the following holds. Let gig_i be continuous covector fields on BRi‾\overline{B_R^i}, belonging to BD(BRi)\mathrm{BD}(B_R^i). Assume that their tangential traces agree on every pair disk: gi(x)∣Sij=gj(x)∣Sij(x∈DRij).(KG4)g_i(x)|_{S_{ij}}=g_j(x)|_{S_{ij}} \quad(x\in D_R^{ij}). \tag{KG4} Then R3∣s(c)∣≤C∑i=04∣Zi L3−Egi∣(BRi).(KG5)R^3|\mathfrak s(c)| \leq C\sum_{i=0}^4 |Z_i\,\mathcal L^3-Eg_i|(B_R^i). \tag{KG5} In particular, when c01=1c_{01}=1 and the other edge data are zero, the sum on the right has a strictly positive lower bound proportional to R3R^3. The conclusion applies to the restrictions of any single smooth ambient covector field.

Proof. First take R=1R=1. Put ui=bi−giu_i=b_i-g_i and ηi=∣Eui∣(B1i)\eta_i=|Eu_i|(B_1^i). The full identity Ebi=ZiL3Eb_i=Z_i\mathcal L^3 gives exactly the error measure appearing in KG5. By KG2 choose Ki(x)=ki+Ωi(x,⋅),Ωi∈⋀2Pi∗,ri=ui−Ki,K_i(x)=k_i+\Omega_i(x,\cdot),\qquad \Omega_i\in\bigwedge^2P_i^*, \qquad r_i=u_i-K_i, with ∥ri∥L1(B1i)≤Cηi\|r_i\|_{L^1(B_1^i)}\leq C\eta_i. Its symmetric derivative is still EuiEu_i. Apply KG3 to either half-ball cut out by SijS_{ij}. Restriction can only decrease the L1L^1 norm and the variation of the derivative measure, so ∥ri∣D1ij∥L1(D1ij)≤Cijηi.(KG6)\|r_i|_{D_1^{ij}}\|_{L^1(D_1^{ij})} \leq C_{ij}\eta_i. \tag{KG6} Continuity identifies this trace from either side. The finitely many constants can be replaced by their maximum.

Let θ^ij\widehat\theta_{ij} be the alternating form on SijS_{ij} with θ^ij(aij,bij)=1\widehat\theta_{ij}(a_{ij},b_{ij})=1. Denote the coefficient vectors of Ωi\Omega_i by ωi\omega_i, as in PK2–PK5. On D1ijD_1^{ij}, the common gg trace cancels. Using PK9 gives the exact identity of tangential covectors (ri−rj)∣Sij=(cij−(Dω)ij)θ^ij(x,⋅)−(ki−kj)∣Sij.(KG7)(r_i-r_j)|_{S_{ij}} =\bigl(c_{ij}-(D\omega)_{ij}\bigr) \widehat\theta_{ij}(x,\cdot) -(k_i-k_j)|_{S_{ij}}. \tag{KG7} The disk is symmetric under x↦−xx\mapsto-x. For any linear map LL and constant vector aa, the triangle inequality gives ∫D1ij∣Lx∣ dx≤∫D1ij∣Lx−a∣ dx.(KG8)\int_{D_1^{ij}}|Lx|\,dx \leq\int_{D_1^{ij}}|Lx-a|\,dx. \tag{KG8} Indeed integrate 2∣Lx∣≤∣Lx−a∣+∣−Lx−a∣2|Lx|\leq|Lx-a|+|-Lx-a|. The constant κij=∫D1ij∣θ^ij(x,⋅)∣ dx\kappa_{ij}=\int_{D_1^{ij}} |\widehat\theta_{ij}(x,\cdot)|\,dx is strictly positive. KG6–KG8 therefore imply ∣cij−(Dω)ij∣≤C′(ηi+ηj).(KG9)|c_{ij}-(D\omega)_{ij}| \leq C'(\eta_i+\eta_j). \tag{KG9} Pair this inequality with the finite stress PK6. Since sTD=0s^TD=0, it gives ∣s(c)∣≤C∑iηi|\mathfrak s(c)|\leq C\sum_i\eta_i. There is no bound or convergence assumption on the translations kik_i or rotations Ωi\Omega_i; the symmetric-disk argument removes the translations and the stress removes all rotations.

For general RR, replace gi(x)g_i(x) on the unit ball by R−1gi(Rx)R^{-1}g_i(Rx). The target bib_i is homogeneous of degree one, and the variation of its strain error rescales by R−3R^{-3}. The unit-ball result yields KG5. ◻

Smooth full-exact forms are included

Let a=d′β+d′′ζa=d'\beta+d''\zeta be a smooth real (1,1)(1,1)-form on an ambient neighborhood of the closed ball. Write β=∑pj d′′xj\beta=\sum p_j\,d''x_j and ζ=∑qj d′xj\zeta=\sum q_j\,d'x_j. In the usual superform signs, aij=∂ipj−∂jqi,sym⁡a=e(B),B=∑j(pj−qj) dxj.(KG10)a_{ij}=\partial_i p_j-\partial_j q_i, \qquad \operatorname{sym}a=e(B), \qquad B=\sum_j(p_j-q_j)\,dx_j. \tag{KG10} Restriction to a linear plane commutes with this identity. Consequently KG5 gives ∑i∫BRi∣Zi−ιi∗sym⁡a∣ dL3≥C−1R3∣s(c)∣.(KG11)\sum_i\int_{B_R^i} |Z_i-\iota_i^*\operatorname{sym}a|\,d\mathcal L^3 \geq C^{-1}R^3|\mathfrak s(c)|. \tag{KG11} This lower bound holds for all such aa, without bounds on their coefficients or derivatives.

A fixed finite positive current cannot dominate vanishing errors

We make the current-order conclusion precise. Work in an ambient open neighborhood OO of BR(0)‾\overline{B_R(0)}. For a smooth full-exact aa, form the fixed-fan current Ea=∑iιi∗[Zi−ιi∗sym⁡a].(KG12)\mathcal E_a=\sum_i\iota_{i*} [Z_i-\iota_i^*\operatorname{sym}a]. \tag{KG12} Let Q\mathcal Q be a fixed real current of bidegree (2,2)(2,2) which is nonnegative on compactly supported strongly positive smooth real (2,2)(2,2) tests. For each fixed constant strongly positive test form, its evaluation on Q\mathcal Q is thus a locally finite positive Radon measure. Say that −ϵQ≤Ea≤ϵQ-\epsilon\mathcal Q\leq\mathcal E_a\leq \epsilon\mathcal Q when both differences are nonnegative on all these tests. This definition is sufficient for the result; no identification of the strong, middle and weak positivity cones is made.

Corollary (No vanishing fixed-control error on PK10). If s(c)≠0\mathfrak s(c)\ne0, there is no sequence of smooth full-exact (1,1)(1,1)-forms aja_j and numbers ϵj↓0\epsilon_j\downarrow0 satisfying −ϵjQ≤Eaj≤ϵjQ(KG13)-\epsilon_j\mathcal Q\leq\mathcal E_{a_j} \leq\epsilon_j\mathcal Q \tag{KG13} on OO, for one fixed Q\mathcal Q as above.

Proof. For each PiP_i, choose an orthonormal basis of its three-dimensional tangent space. The six scalar quadratic probes in the directions eae_a and ea+ebe_a+e_b, 1≤a<b≤31\leq a<b\leq3, determine every symmetric matrix entry by polarization. For each direction choose a constant decomposable positive ambient (2,2)(2,2)-form whose restriction to PiP_i pairs with that quadratic probe, including its fixed positive normalization. Such a form is obtained from two covectors spanning the perpendicular two-plane and their double-primed copies. Write these thirty forms as Ψiν\Psi_{i\nu}.

For each Ψiν\Psi_{i\nu}, KG13 dominates the corresponding scalar signed measure by ϵj Q(Ψiν)\epsilon_j\,\mathcal Q(\Psi_{i\nu}). Let Ai=Pi∖⋃k≠iPkA_i=P_i\setminus\bigcup_{k\ne i}P_k. The five intrinsic three-dimensional error measures are mutually singular: their intersections have three-dimensional measure zero. Restricting the scalar measure inequality to AiA_i therefore isolates exactly the desired scalar probe of the ii-th strain error. No other source plane contributes on this Borel set.

Choose a nonnegative compact cutoff χ\chi in OO which is one on BR(0)‾\overline{B_R(0)}. Polarization and finite-dimensional norm comparison give a constant C0C_0 such that ∑i∫BRi∣Zi−ιi∗sym⁡aj∣ dL3≤C0ϵj∑i,ν⟨Q,χΨiν⟩.(KG14)\sum_i\int_{B_R^i} |Z_i-\iota_i^*\operatorname{sym}a_j|\,d\mathcal L^3 \leq C_0\epsilon_j \sum_{i,\nu}\langle\mathcal Q,\chi\Psi_{i\nu}\rangle. \tag{KG14} Every term in the finite sum is finite and nonnegative, and the sum is independent of jj. Its right side tends to zero, contrary to KG11. ◻

What this rules out.

The same fixed supplied fan has all the local primitives and fixed-fan d′d′′d'd''-closure proved in PK, yet even arbitrarily large smooth ambient full-exact coefficients cannot approach its strain in the above fixed-positive-control order. This is a quantitative obstruction on retained real source planes. No original analytic zero realizing PK10 has been constructed, and KG does not refute the original analytic all-refinement criterion. Genuine actual unit refinements can change the incident spaces and the associated stress; that operation is not represented by a real subdivision of this fixed fan.

Primary analytic sources.

M. Friedrich, A Korn–Poincare-type inequality for special functions of bounded deformation, arXiv:1503.06755v3, Theorem A.1, states the classical BD\mathrm{BD} Korn–Poincare estimate used in KG2 in arbitrary dimension. J.-F. Babadjian, Traces of functions of bounded deformation, arXiv:1308.5497v2, Theorem 3.2, proves the bounded trace map on Lipschitz domains and its agreement with continuous boundary restrictions. Only these two analytic estimates are imported; KG5–KG14 are the finite-configuration and current-order deductions given here.

Source text and edition identity

This reading copy omits personal attribution and places existing mathematical question marks over their relation symbols. The mathematical content and source status are retained.

Reading-copy SHA-256 700b64c92016f927a301b2d902267cc1e7568168030090b4613d9209b932345d

Frozen-original SHA-256 700b64c92016f927a301b2d902267cc1e7568168030090b4613d9209b932345d

\section{A quantitative strain gap and the failure of fixed-control approximation on the supplied fan}
\label{sec:r13-killing-stress-l1-gap}

Keep the five planes, trace data, fields \(b_i\), strains \(Z_i\),
and edge stress \(s\) of PK.  Equip \(N=\mathbf R^4\) and its
subspaces with their Euclidean norms.  Write
\[
 B_R^i=P_i\cap B_R(0),\qquad
 D_R^{ij}=S_{ij}\cap B_R(0),\qquad
             \mathfrak s(c)=\sum_{i<j}s_{ij}c_{ij}.
 \tag{KG1}
\]
All tensor-measure total variations below use the Euclidean
Frobenius norm.  Changing to the fixed lattice calibrations
changes only the constants.

The two external analytic inputs are the classical
Korn--Poincare estimate in \(\mathrm{BD}\) and the bounded
\(\mathrm{BD}\) trace map on a bounded Lipschitz domain.
Here \(Eu=\operatorname{sym}Du\) denotes the full distributional
symmetric derivative.  In the form needed on a three-ball,
there is an affine Killing field \(K\) such that
\[
 \|u-K\|_{L^1(B_R)}\leq C R\,|Eu|(B_R).
 \tag{KG2}
\]
The trace estimate on a half-ball is
\[
 \|\gamma u\|_{L^1(\partial B_R^+)}
       \leq C\bigl(R^{-1}\|u\|_{L^1(B_R^+)}
                              +|Eu|(B_R^+)\bigr).
 \tag{KG3}
\]
We use KG2 only for continuous \(\mathrm{BD}\) fields on a
closed ball; hence they are also in \(L^2\).  KG2 follows by
Holder and scaling from the \(L^{3/2}\) estimate stated as
Theorem A.1 in Friedrich's paper cited below.  KG3 follows
by scaling from Babadjian's Theorem 3.2; for a continuous
field its trace is its usual boundary restriction.  No
\(L^1\) estimate for the full gradient is used.

\subsection{The finite Killing stress is stable in the strain-measure norm}

\begin{theorem}[Quantitative obstruction]
\label{thm:r13-strain-measure-gap}
There is a constant \(C>0\), depending only on the fixed
five-plane configuration, such that the following holds.
Let \(g_i\) be continuous covector fields on
\(\overline{B_R^i}\), belonging to \(\mathrm{BD}(B_R^i)\).
Assume that their tangential traces agree on every pair disk:
\[
                  g_i(x)|_{S_{ij}}=g_j(x)|_{S_{ij}}
                           \quad(x\in D_R^{ij}).
 \tag{KG4}
\]
Then
\[
 R^3|\mathfrak s(c)|
       \leq C\sum_{i=0}^4
             |Z_i\,\mathcal L^3-Eg_i|(B_R^i).
 \tag{KG5}
\]
In particular, when \(c_{01}=1\) and the other edge data
are zero, the sum on the right has a strictly positive lower
bound proportional to \(R^3\).  The conclusion applies to
the restrictions of any single smooth ambient covector field.
\end{theorem}

\begin{proof}
First take \(R=1\).  Put
\(u_i=b_i-g_i\) and
\(\eta_i=|Eu_i|(B_1^i)\).  The full identity
\(Eb_i=Z_i\mathcal L^3\) gives exactly the error measure
appearing in KG5.  By KG2 choose
\[
 K_i(x)=k_i+\Omega_i(x,\cdot),\qquad
              \Omega_i\in\bigwedge^2P_i^*,
 \qquad r_i=u_i-K_i,
\]
with \(\|r_i\|_{L^1(B_1^i)}\leq C\eta_i\).
Its symmetric derivative is still \(Eu_i\).
Apply KG3 to either half-ball cut out by \(S_{ij}\).
Restriction can only decrease the \(L^1\) norm and the
variation of the derivative measure, so
\[
       \|r_i|_{D_1^{ij}}\|_{L^1(D_1^{ij})}
                           \leq C_{ij}\eta_i.
 \tag{KG6}
\]
Continuity identifies this trace from either side.  The
finitely many constants can be replaced by their maximum.

Let \(\widehat\theta_{ij}\) be the alternating form on
\(S_{ij}\) with
\(\widehat\theta_{ij}(a_{ij},b_{ij})=1\).  Denote the
coefficient vectors of \(\Omega_i\) by \(\omega_i\), as in
PK2--PK5.  On \(D_1^{ij}\), the common \(g\) trace cancels.
Using PK9 gives the exact identity of tangential covectors
\[
 (r_i-r_j)|_{S_{ij}}
 =\bigl(c_{ij}-(D\omega)_{ij}\bigr)
                 \widehat\theta_{ij}(x,\cdot)
                    -(k_i-k_j)|_{S_{ij}}.
 \tag{KG7}
\]
The disk is symmetric under \(x\mapsto-x\).  For any
linear map \(L\) and constant vector \(a\), the triangle
inequality gives
\[
       \int_{D_1^{ij}}|Lx|\,dx
                     \leq\int_{D_1^{ij}}|Lx-a|\,dx.
 \tag{KG8}
\]
Indeed integrate
\(2|Lx|\leq|Lx-a|+|-Lx-a|\).
The constant
\(\kappa_{ij}=\int_{D_1^{ij}}
                  |\widehat\theta_{ij}(x,\cdot)|\,dx\)
is strictly positive.  KG6--KG8 therefore imply
\[
       |c_{ij}-(D\omega)_{ij}|
                          \leq C'(\eta_i+\eta_j).
 \tag{KG9}
\]
Pair this inequality with the finite stress PK6.  Since
\(s^TD=0\), it gives
\( |\mathfrak s(c)|\leq C\sum_i\eta_i\).
There is no bound or convergence assumption on the translations
\(k_i\) or rotations \(\Omega_i\); the symmetric-disk
argument removes the translations and the stress removes
all rotations.

For general \(R\), replace \(g_i(x)\) on the unit ball by
\(R^{-1}g_i(Rx)\).  The target \(b_i\) is homogeneous of
degree one, and the variation of its strain error rescales
by \(R^{-3}\).  The unit-ball result yields KG5.
\end{proof}

\subsection{Smooth full-exact forms are included}

Let \(a=d'\beta+d''\zeta\) be a smooth real \((1,1)\)-form
on an ambient neighborhood of the closed ball.  Write
\(\beta=\sum p_j\,d''x_j\) and
\(\zeta=\sum q_j\,d'x_j\).  In the usual superform signs,
\[
 a_{ij}=\partial_i p_j-\partial_j q_i,
 \qquad
 \operatorname{sym}a=e(B),
       \qquad B=\sum_j(p_j-q_j)\,dx_j.
 \tag{KG10}
\]
Restriction to a linear plane commutes with this identity.
Consequently KG5 gives
\[
 \sum_i\int_{B_R^i}
       |Z_i-\iota_i^*\operatorname{sym}a|\,d\mathcal L^3
                    \geq C^{-1}R^3|\mathfrak s(c)|.
 \tag{KG11}
\]
This lower bound holds for all such \(a\), without bounds
on their coefficients or derivatives.

\subsection{A fixed finite positive current cannot dominate vanishing errors}

We make the current-order conclusion precise.  Work in an
ambient open neighborhood \(O\) of \(\overline{B_R(0)}\).
For a smooth full-exact \(a\), form the fixed-fan current
\[
 \mathcal E_a=\sum_i\iota_{i*}
       [Z_i-\iota_i^*\operatorname{sym}a].
 \tag{KG12}
\]
Let \(\mathcal Q\) be a fixed real current of bidegree
\((2,2)\) which is nonnegative on compactly supported
strongly positive smooth real \((2,2)\) tests.  For each
fixed constant strongly positive test form, its evaluation
on \(\mathcal Q\) is thus a locally finite positive Radon
measure.  Say that
\(-\epsilon\mathcal Q\leq\mathcal E_a\leq
\epsilon\mathcal Q\) when both differences are
nonnegative on all these tests.  This definition is sufficient
for the result; no identification of the strong, middle and
weak positivity cones is made.

\begin{corollary}[No vanishing fixed-control error on PK10]
\label{cor:r13-no-fixed-control-on-supplied-five-plane-fan}
If \(\mathfrak s(c)\ne0\), there is no sequence of smooth
full-exact \((1,1)\)-forms \(a_j\) and numbers
\(\epsilon_j\downarrow0\) satisfying
\[
             -\epsilon_j\mathcal Q\leq\mathcal E_{a_j}
                                      \leq\epsilon_j\mathcal Q
 \tag{KG13}
\]
on \(O\), for one fixed \(\mathcal Q\) as above.
\end{corollary}

\begin{proof}
For each \(P_i\), choose an orthonormal basis of its
three-dimensional tangent space.  The six scalar quadratic
probes in the directions \(e_a\) and \(e_a+e_b\),
\(1\leq a<b\leq3\), determine every symmetric matrix
entry by polarization.  For each direction choose a constant
decomposable positive ambient \((2,2)\)-form whose restriction
to \(P_i\) pairs with that quadratic probe, including its
fixed positive normalization.  Such a form is obtained from
two covectors spanning the perpendicular two-plane and their
double-primed copies.  Write these thirty forms as
\(\Psi_{i\nu}\).

For each \(\Psi_{i\nu}\), KG13 dominates the corresponding
scalar signed measure by
\(\epsilon_j\,\mathcal Q(\Psi_{i\nu})\).
Let
\(A_i=P_i\setminus\bigcup_{k\ne i}P_k\).
The five intrinsic three-dimensional error measures are
mutually singular: their intersections have three-dimensional
measure zero.  Restricting the scalar measure inequality to
\(A_i\) therefore isolates exactly the desired scalar
probe of the \(i\)-th strain error.  No other source plane
contributes on this Borel set.

Choose a nonnegative compact cutoff \(\chi\) in \(O\)
which is one on \(\overline{B_R(0)}\).  Polarization and
finite-dimensional norm comparison give a constant \(C_0\)
such that
\[
 \sum_i\int_{B_R^i}
       |Z_i-\iota_i^*\operatorname{sym}a_j|\,d\mathcal L^3
 \leq C_0\epsilon_j
       \sum_{i,\nu}\langle\mathcal Q,\chi\Psi_{i\nu}\rangle.
 \tag{KG14}
\]
Every term in the finite sum is finite and nonnegative,
and the sum is independent of \(j\).  Its right side tends
to zero, contrary to KG11.
\end{proof}

\paragraph{What this rules out.}
The same fixed supplied fan has all the local primitives and
fixed-fan \(d'd''\)-closure proved in PK, yet even arbitrarily
large smooth ambient full-exact coefficients cannot approach
its strain in the above fixed-positive-control order.  This
is a quantitative obstruction on retained real source planes.
No original analytic zero realizing PK10 has been constructed,
and KG does not refute the original analytic all-refinement
criterion.  Genuine actual unit refinements can change the
incident spaces and the associated stress; that operation is
not represented by a real subdivision of this fixed fan.

\paragraph{Primary analytic sources.}
M. Friedrich, \emph{A Korn--Poincare-type inequality for special
functions of bounded deformation}, arXiv:1503.06755v3,
Theorem A.1, states the classical \(\mathrm{BD}\)
Korn--Poincare estimate used in KG2 in arbitrary dimension.
J.-F. Babadjian, \emph{Traces of functions of bounded deformation},
arXiv:1308.5497v2, Theorem 3.2, proves the bounded trace map
on Lipschitz domains and its agreement with continuous
boundary restrictions.  Only these two analytic estimates
are imported; KG5--KG14 are the finite-configuration and
current-order deductions given here.